On September 1, 2026, The Descartes Systems Group acquired Extensiv for approximately US$120 million, satisfied from cash on hand.

Extensiv, formerly 3PL Central, is a warehouse and order management platform for third-party logistics providers and the brands they serve. It appeared on the Inc. 5000 five consecutive years. Its last disclosed growth rate was 189% over three years. It had four products, more than 25,000 logistics professionals on the platform, eleven years of institutional capital behind it, and, on my estimate, somewhere between $90 and $100 million of revenue.

It sold for about 1.2 times revenue.

Everything below is an attempt to explain how a real software company with real customers and a decade of compounding growth got priced like a services rollup.

The Number

Separate what’s certain from what isn’t.

We don’t have any real revenue data. No audited revenue figure for Extensiv has ever been public, and that is the single biggest caveat in this piece. What we have instead is the Inc. 5000, which ranks companies on percentage revenue growth over a trailing three-year window and requires them to submit real figures. Extensiv submitted five years running: No. 4150 in 2020, No. 4051 in 2021, No. 3812 in 2022, No. 2816 in 2023, No. 3038 in 2024, with 75% three-year growth on the 2021 list and 189% on the 2023 list, measured 2019 to 2022 under Inc.’s stated methodology.

Chain those rates together and a path falls out: roughly $20M in 2018, $23M in 2019, $31M in 2020, $46M in 2021, and $66M in 2022. Grow that at a decelerating high-single-digit clip through the freight recession and you land at $90 to $100 million by 2026. These are self-reported, unaudited percentages run through my arithmetic. Treat them as a shape rather than a fact.

At $100 million of revenue, $120 million is 1.2x. At my mid case of $90 million, it’s 1.3x. Even if I’m overshooting badly and the real number is $80 million, it’s 1.5x. There is no version of this where a growing vertical SaaS business sold at a normal software multiple.

On August 24, 2026, eight days before Extensiv, Descartes acquired Tai, an AI-powered transportation management platform for freight brokers, for about $100 million.

Roughly the same check for a young single-product freight tool as for a twenty-year-old platform with four products, a five-year Inc. 5000 streak, thousands of brands and roughly 2,000 3PL warehouses running on it.

Descartes did not misprice anything. Descartes paid what a business is worth when the seller’s alternative is refinancing.

Which is the actual subject of this piece.

Act One: The Warehouse in Commerce

3PL Central was founded in 2006, and the founding story is the most durable kind a software company can have. An operator built the tool it couldn’t buy.

Per the Los Angeles Business Journal, “3PL was founded in 2006 by JAM-N Logistics in Commerce to fulfill logistics challenges inside the company.” A third-party warehouse went looking for a cloud WMS to run its own floor, didn’t find one worth buying, and wrote it.

That’s the same shape as Weave’s attic, and historically it’s the best single predictor that the product actually works. It did.

The company later moved to El Segundo. Its first acquisition came a decade in: Traker Systems of Temecula, bought in 2016.

By then the money had already arrived.

Act Two: The Money Arrives

On December 3, 2015, growth equity firm Mainsail Partners invested $17.5 million, the company’s first outside capital after nine years of bootstrapping. In 2018 Mainsail brought in Andy Lloyd, formerly of NetSuite, as CEO. That is the standard move when a sponsor decides a founder-run company needs to become a platform.

Then came 2021, and 2021 was very good.

In April, Mainsail led a $45 million recapitalization. The numbers Lloyd gave the LA Business Journal explain why anyone was willing. Order volume through the platform jumped 46% year over year in 2020, and 63% in the first quarter of 2021 against the same quarter a year earlier.

Hold that sentence. It is the most important one in this piece, and not for the reason it was written.

Three acquisitions followed in nine months.

April 7, 2021: Skubana. Multichannel order management, New York. Capital Clarity was Skubana’s exclusive financial advisor, and terms were not disclosed. Lloyd told the LABJ the company did not use the Mainsail round to pay for it. This became Extensiv Order Manager.

September 1, 2021: Scout Software. A cloud WMS for private and company-operated warehouses. Terms undisclosed.

December 1, 2021: CIO Technologies, including CartRover. The integration layer connecting more than 100 shopping carts and marketplaces. CIO’s CEO Gunnar Fredlund noted they already supported integrations for 3PL Warehouse Manager and Skubana and shared hundreds of common customers. Terms undisclosed. This became Extensiv Integration Manager.

DC Velocity summed up the pace at the third deal: three acquisitions in six months following a $45 million injection from its private equity owner.

In 2022 the whole thing was rebranded from 3PL Central to Extensiv.

Revenue roughly tripled in three years. What the company did not have was more equity it wanted to sell.

Act Three: The Loan

In December 2022, Runway Growth Capital closed an $80 million senior secured first-lien term loan. Runway Growth Finance Corp.’s Q4 2022 portfolio update describes the “completion of a new $80 million senior secured, first-lien term loan commitment to Extensiv (formerly known as ‘3PL Central’), funding $65 million at close.” Runway added roughly $3 million more in the first quarter of 2023 and about half a million later, bringing funded principal to roughly $68.5 million.

Lloyd’s framing was the standard one and an entirely reasonable one: non-dilutive capital for new product launches and a continued acquisition strategy.

And the company acted on it immediately. Six weeks later, in January 2023, Extensiv named Aaron Stead president and added two vice presidents. The release says the roles were added, in so many words, “to execute its aggressive growth plans following the $80 million secured term loan.” The debt wasn’t incidental to the strategy. It was the strategy.

Be fair about what was and wasn’t unreasonable here.

None of it was unreasonable in December 2022. Growth debt is the rational instrument for a sponsor-backed company that has just tripled revenue and doesn’t want to price another equity round into a collapsing multiple environment. Mainsail had already put in $62.5 million. Raising more at 2022 marks would have been expensive. Borrowing was not. Every growth-stage board in the country ran that trade that year.

Then watch what happened to the terms.

Act Four: The Line in the Filing

Here is the sentence that explains the multiple. It comes from Runway Growth Finance Corp.’s first-quarter 2026 report, accepted by EDGAR on May 7, 2026:

3PL Central LLC (dba Extensiv). Senior Secured. SOFR+7.00%, 9.00% floor, 5.10% ETP. Initial Acquisition Date 11/9/2022. Maturity Date 6/30/2026.

The loan came due on June 30, 2026. The company sold on September 1, 2026.

Sixty-three days.

Three things in that line matter, and the third matters most.

One, the spread. SOFR plus 7.00% with a 9.00% floor is not the pricing you’d expect on a healthy, growing software credit. For scale, look at what else sat in Runway’s schedule that same quarter: Digicert at SOFR+5.75% with a 6.50% floor, Zinnia at SOFR+6.50% with a 7.50% floor. Extensiv is priced well above both.

Two, the end-of-term payment. ETPs are one-time fees stated as a percentage of principal. Runway’s book that quarter is full of them in the 1% to 3.5% range. Five percent is what a lender charges when it is being asked for something.

Three, the maturity. The loan originated November 9, 2022 and was due June 30, 2026. A December 2022 press release announcing an $80 million growth facility for product launches and an acquisition strategy does not describe a three-and-a-half-year instrument. Somewhere between origination and the first quarter of 2026, that maturity got pulled forward.

Be precise about the limits of what I can prove. I have verified the Q1 2026 terms directly from the filing. I have not been able to pull the original November 2022 schedule line to show you the before and the after side by side. Secondary summaries of earlier Runway filings report materially softer original terms, something near SOFR plus 4.50% with a 6.50% floor and a 2.00% ETP, which would mean the spread rose 250 basis points, the floor rose 250, and the ETP more than doubled. I’d treat those “from” figures as unconfirmed until somebody reads the FY2023 and FY2024 schedules line by line. What is not in doubt is where the loan ended up, and when it was due.

One thing cuts the other way, and it belongs here. The loan was never placed on non-accrual. It remained a Non-Control/Non-Affiliate debt investment in Systems Software through the first quarter of 2026. Runway’s second-quarter 2026 net realized loss of $45.3 million was attributable to Marley Spoon and Blueshift, not to Extensiv.

So here is the honest characterization. This is not a Pluralsight story, where lenders took the keys. This is a borrower that renegotiated with its lender from a position of weakness, paid for the accommodation in spread and fees, accepted a shorter runway to a hard date, and then hit that date.

Act Five: The Volume Stopped

Go back to the sentence I asked you to hold. Order volumes up 46%, then 63%.

Extensiv’s revenue was never purely contractual. A 3PL WMS priced against warehouses, orders and throughput doesn’t only need its customers to renew. It needs its customers’ customers to keep shipping boxes. That is a fundamentally different exposure from a seat-based enterprise subscription, and it is the exposure the 2022 loan was underwritten against, at the exact peak of pandemic e-commerce volume.

Then the freight recession arrived, and Extensiv documented its own undoing better than any outside analyst could have.

From the company’s 4th Annual 3PL Warehouse Benchmark Report, published in October 2023 on data from more than 240 3PL warehouses:

In 2021, 55% of warehouses operated at 90%+ capacity, growing to 59% of 3PLs in 2022. In 2023, the percentage of warehouses operating at 90%+ capacity nearly halved to 33%.

The same report found that 22% of 3PLs reported stagnant or declining order volumes, that the number of flat or declining warehouses more than doubled year over year, and that 70% were fighting higher labor costs.

Read what that report says about the company that published it. Extensiv’s customers didn’t churn en masse. They didn’t have to. They just stopped filling their warehouses. Capacity utilization nearly halved in a single year, and a business that bills against throughput cannot decouple itself from that. The revenue didn’t leave. It thinned.

This is the third category, and I don’t think I’ve written it clearly enough before.

Instructure is a system of record. A university does not switch its gradebook. Gross retention in the mid to high 90s, switching costs measured in academic years, which is why $2.275 billion of acquisition debt sits on it as a staircase.

Weave is a system of action. The front desk. Phones, texts, reminders. Real switching costs, a thinner floor, 89% gross retention.

Extensiv is a system of throughput. Its customers are genuinely sticky, because ripping out the WMS that runs your warehouse is arguably harder than replacing a phone system, and its logo retention was almost certainly fine. But stickiness protects you from churn, not from volume. You can hold 95% of your customers and still watch revenue flatten, because the meter your pricing runs on belongs to somebody else’s demand curve.

That is the thinnest floor of the four, and it is the one nobody prices correctly, because on a spreadsheet it looks like the strongest.

A company growing 15% can service $68.5 million of debt at SOFR plus 7.00%. A company growing 6% can service it too. It just can’t do anything else, and it certainly can’t refinance it on good terms.

That is why the multiple was 1.2x. Not because the product was bad. Because the seller had a date.

Act Six: The Quietest Exit in the Series

One leadership change matters. On October 3, 2024, Aaron Stead was promoted from President to CEO, and Andy Lloyd moved to the board. Board chairman Jason Payne called it “a significant milestone for Extensiv.”

Read the praise in that release closely. Stead had delivered “impressive growth, increased customer sentiment, and expanding margins.“ Margin language is what boards reach for when growth language has stopped being available. Twenty months before the debt came due, the company was already being run for cash.

Twenty-three months later, Descartes announced the acquisition. Approximately US$120 million, all cash, from cash on hand. Not pending. Done.

Now look at what isn’t in that announcement.

No earn-out. No holdback, no contingent consideration, no “up to $X million including performance-based payments.” That’s notable because Descartes uses that structure constantly. A clean cash number with no performance tail is what you write when nobody is being asked to stay and hit a target.

No quote from Edward Ryan, Descartes’ CEO, who is quoted in most Descartes deal announcements, including the Tai release eight days earlier.

No quote from Aaron Stead. No quote from Mainsail Partners. No advisors named on either side.

The only two quotes are Descartes general managers. Mikel Richardson, GM of Ecommerce Operations, on the network: “3PLs are under constant pressure to fulfill faster, scale flexibly, and support the evolving needs of modern brands.” And Scott Sangster, GM of Logistics Services Providers: “We’re excited to welcome Extensiv’s customers, partners, and employees.”

That is an announcement written entirely from the buyer’s side of the table.

Be precise about what’s established here. An absence of quotes is not by itself evidence of a distressed sale. Sellers decline to comment for all kinds of reasons, and Descartes writes short releases. But across five of these pieces, this is the first deal announcement in which no representative of the company being sold, not the CEO, not the chairman, not the sponsor who held it for eleven years, said anything at all.

The Waterfall

Arithmetic, labeled as inference throughout, because none of the parties disclosed any of it.

In: $17.5 million in 2015 plus $45 million in 2021 equals $62.5 million of disclosed equity, before whatever the founders and co-investors held and before eleven years of holding cost. Plus $68.5 million of funded senior debt.

Out: approximately $120 million, all cash.

The senior secured first-lien loan gets repaid first. Principal, interest at a 9.00% floor, and a 5.10% end-of-term payment on the principal amount. Call it $72 to $80 million. Transaction expenses come next.

That leaves roughly $35 to $48 million of residual value for everyone below the debt: Mainsail, its co-investors, the founders, and every employee holding options.

Against $62.5 million in.

If that’s directionally right, and it’s an estimate built on a disclosed loan balance, disclosed terms and an undisclosed payoff, then Mainsail Partners held this company for nearly eleven years, through five Inc. 5000 appearances, three acquisitions, a rebrand and a tripling of revenue, and got back materially less money than it put in. Employee equity was almost certainly worth nothing.

And the lender was made whole.

That’s the version of this story that matters. Nobody did anything reckless. There was no Vista-style paying for a company that doesn’t exist yet. There was no seven-fund club opening a trapdoor. There was a reasonable amount of reasonably priced senior debt, taken by a competent board at a moment when it was the cheapest capital available, secured against revenue that everyone believed would keep compounding, including the lender, and including me if I’d been in that room.

Then the boxes stopped moving. The terms got worse. The maturity moved closer. And on June 30, 2026, the bill arrived.

Act Seven: The Buyer

Descartes is not an ambulance the way Francisco Partners was for Weave. Descartes is a machine.

Eight days before Extensiv, on August 24, 2026, it acquired Tai for about US$100 million. It has bought this shelf before, acquiring the UK e-commerce WMS Peoplevox in 2020. Extensiv slots alongside it and feeds the Global Logistics Network, which is the asset Descartes is actually compounding. Every acquisition adds nodes, and Extensiv adds roughly 2,000 3PL warehouses and their brand customers.

Descartes reports its fiscal second quarter on September 10, which is the next place any of this gets tested.

What I’d Watch

1. Runway’s third-quarter 2026 filings. The quarter containing the September 1 close will show a realized gain or loss and a final value on the 3PL Central line. Repaid at or near par means this was an M&A payoff and my read is right. A realized loss means the sale didn’t cover the debt, and the story is worse than the press release suggests. Watch RWAY’s EDGAR page.

2. The original November 2022 loan schedule. Somebody should read the 3PL Central line in the FY2023 and FY2024 Consolidated Schedules of Investments. That gives you the repricing quarter by quarter and the date the maturity moved. It is the difference between “the terms drifted” and “there was an amendment, and here is when.”

3. Descartes’ fiscal Q2 2027 call on September 10. The purchase price allocation will disclose acquired revenue and deferred revenue. That is the first real revenue number for Extensiv, ever, and it will either confirm or embarrass the 1.2x in my headline.

4. Whether Order Manager survives. Skubana was the largest tuck-in and the most duplicative with what Descartes already owns. Product rationalization is where the value of a $120 million acquisition either shows up or doesn’t.

5. The El Segundo office. Nothing was disclosed about employees or facilities. Descartes typically keeps acquired products and rebrands them, as with Descartes Peoplevox and Descartes MacroPoint. Whether a Descartes Extensiv keeps a Southern California presence is the first tell about whether this was bought as a product or as a customer list.

The Thing I Actually Learned

I started this series believing the story was that private equity kills companies. Instructure taught me the better version: leverage is a multiplier, and what it multiplies is whether the revenue will still be there in five years. Weave taught me you don’t need leverage at all, because a mispricing and enough time will open the same trapdoor.

Extensiv taught me the third one, and it’s the one I’m least comfortable with. The debt doesn’t have to fail for you to lose. It just needs a date on it.

Nobody defaulted. Nobody took the keys. Runway Growth Capital underwrote a real business at reasonable terms, monitored it, repriced it when the numbers softened, shortened the runway rather than forcing a default, never marked it to non-accrual, and got repaid. By every standard applied to lenders, that is a well-managed loan. Nobody is going to write about the Extensiv credit as a cautionary tale, because from the lender’s chair there isn’t one.

And it still compressed a $90 to $100 million revenue software company into a 1.2x exit, because $68.5 million of senior claim sitting in front of you with a maturity attached converts every strategic question into a financing question. You don’t sell when the price is right. You sell when the note is due.

Which brings me to the part I’d tell any founder reading this, and the reason I wrote about a California company in a series about Utah ones.

Ask what your revenue is a function of before you borrow against it. Extensiv’s revenue was a function of how many boxes American consumers ordered, a variable nobody at Extensiv could influence, forecast or hedge. That is not a criticism of the business. The business is good, which is why Descartes bought it. It is a criticism of underwriting throughput as though it were contract.

Weave’s floor was 89% gross retention. Instructure’s floor was a gradebook nobody can switch. Extensiv’s floor was other people’s demand, and other people’s demand doesn’t sign a contract.

The most telling detail in this entire story is a number from Extensiv’s own research department. Warehouses running at 90% or higher capacity fell from 59% to 33% in a single year. The company published that. It put out a press release about it, framed as an industry trend, while its own loan was being repriced.

It was both.

If you worked at Extensiv or 3PL Central through the 2021 acquisitions, the 2022 rebrand, the Runway financing, or the sale to Descartes, I’d like to hear from you. Two things in particular: what the Skubana, Scout and CartRover purchase prices actually were, and when the Runway loan was amended.

A Note on What Isn’t Known

• No audited revenue figure for Extensiv has ever been public. The $90 to $100 million estimate, and therefore the 1.2x in the headline, is modeled from self-reported, unaudited Inc. 5000 growth percentages. It is the load-bearing estimate in this piece and it deserves the scrutiny.

• The original December 2022 loan terms are unconfirmed. I verified the Q1 2026 line directly. The “from” figures need somebody to open the earlier 10-Ks.

• The loan payoff amount is undisclosed. My $72 to $80 million range applies the stated 9.00% floor and 5.10% ETP to roughly $68.5 million of funded principal.

• The Skubana, Scout and CartRover purchase prices were never disclosed.

• Whether a sale process ran, and who else looked, is unknown. Extensiv was private and Descartes is the acquirer, so there is no proxy and no “Background of the Merger” section coming. That is the structural reason a private-to-strategic sale is harder to write about than a take-private, and it is why the Runway filings carry so much of the weight here.

• The waterfall is arithmetic, not disclosure. Mainsail has said nothing about its return, and may never.

Domo burned roughly seven dollars to buy one dollar of recurring revenue and sold its operating business to Progress Software. Pluralsight went from a $3.5 billion unicorn to a lender takeover that zeroed out Vista Equity Partners — killed not by its product but by a $1.175 billion recurring-revenue loan held by seven private credit funds who could sit in one room and act as one. Instructure took the same ownership churn, the same layoffs, the same debt-funded acquisitions, and lived, because Canvas is a near-monopoly on a system of record and the loan underneath it was a broadly syndicated term loan B with hundreds of holders.

Same ladder. Different rung material. Leverage isn’t a verdict. It’s a multiplier.

Then, on August 18, 2026, Francisco Partners agreed to take Weave Communications private for $7.40 a share in cash — roughly $650 million, a 34% premium to the unaffected close of $5.53.

And the pattern broke. Not because Weave is healthier than the others. Because by the time the private equity firm showed up, the damage had already been done — in public, without leverage, in full view of everyone.

Weave IPO’d at $24 a share. It’s selling at $7.40. That’s a 69% per-share decline — and because Weave never split its stock and was never recapitalized, that is precisely what a Day One IPO buyer who held to the end will have lost.

Measured at the company level rather than the share level, the number is softer and more interesting. Weave went public at roughly $1.5 billion of equity value and is selling for about $650 million — a 57% decline. The gap between 69% and 57% is dilution: the share count grew about a quarter, from roughly 63.6 million just after the IPO to 79.9 million by June 2026, through stock compensation and shares issued for an acquisition.

Now hold either number against the business. Revenue over that same period went from $115.9 million to a guided $275 million. The company more than doubled while its equity value more than halved.

Which gives you the whole story in one line: Weave went public at somewhere around ten to thirteen times revenue and is being sold at roughly two. Nothing else here is as large as that number.

You cannot blame private equity for it. Nobody borrowed anything. As of June 30, 2026, Weave had no outstanding debt at all — nothing drawn on its $50 million revolver — and about $78 million in cash and investments.

So this one is a different kind of piece. The first three were about what leverage did to a company. This one is about what happened before leverage — and about the question the merger proxy hasn’t answered yet.

Act One: The Attic

Brandon Rodman’s mother went back to school in middle age to get a dental hygiene license. His father ran a general contracting business. He grew up in Eugene, Oregon, went to BYU in 1999, served a mission in Romania, and spent six years knocking doors in summer sales — a job he describes, correctly, as a job about failure.

In 2008 his employer offered him a large pay cut or the door. In September 2008 he and his wife agreed to put $5,000 toward a startup, with an explicit stop-loss: if the money ran out before revenue came in, he’d go get a job.

The idea came from two things colliding. He’d heard about a dental office manager booking appointments from home during maternity leave. And he knew from summer sales how cheap it had become to spin up a call center off a website and a spreadsheet.

Dental offices can’t reach lapsed patients because they call during business hours, when those patients are at work. Dentists don’t want to make evening calls because dentists don’t want to work evenings. That gap was the entire business. Hand over your list of overdue patients; Recall Solutions would call them at night and book them.

It ran out of the finished attic above the garage. Employees came through the front door every morning, said hi to his wife and daughter, and went upstairs. It got hot up there.

Two things then happened that turned a service business into a software company.

First, the math stopped working. At fifty practices, regenerating call lists ate a couple days every month. Rodman asked what that looked like at five hundred practices. Second, they tried text messages and discovered something that now sounds obvious and in 2009 did not: most patients would rather text than talk. Send me what you need and I’ll handle it when I have a minute.

By 2011 they were building software — a phone system with texting, wired into the practice’s patient records. And “recall” is dental jargon for the six-month cleaning call, a name that pigeonholed them, so in 2010 they’d already renamed to Weave, for weaving the phone, the texting, and the patient data together.

The first version wasn’t built in-house. In 2010 a gas leak evacuated the building next door and Rodman ended up in the parking lot with two engineers, Gabe and Jordan Gunderson, whose dev shop Izeni turned out to know how to build phone systems. They built the first version of Weave. Gabe Gunderson would later co-found Rodman’s next company.

By 2013 Weave was nearly out of money. Rodman pitched every angel and venture firm in Utah and got no from all of them. As a last-ditch effort they applied to Y Combinator, flew to Mountain View, sat through a ten-minute interview, and got the acceptance call driving to the Oakland airport. No Utah company had ever been accepted to YC before. Within a month they’d raised a million dollars. Within two months they went from asking whether to shut down to not needing the money.

Rodman’s read on that is the sharpest line in his own telling: nothing about the business had changed. What changed was who believed it could be big.

Geoff Ralston, then a YC partner, later admitted the application hadn’t impressed him — and noted that before the batch even started there was cleanup to do, because the original entity was a complicated LLC that needed unwinding, and there was a cash crunch to resolve before Demo Day.

Act Two: The Year They Almost Sold It for Nothing

Here is the part of the Weave story nobody has reported, and it’s the hinge.

On IPO day in November 2021, Pelion Venture Partners managing partner Blake Modersitzki posted publicly that Rodman had reminded him of something: they almost sold Weave in 2016 for next to nothing. Modersitzki credited Rodman and then-CFO Alan Taylor with the decision to take the risk and keep building instead.

The would-be buyer has never been named. Neither has the price.

The timing fits the balance sheet. Weave had raised $17 million in a Series B-1 in November 2015 and wouldn’t close its $23 million B-2 until September 2017. 2016 was the trough between them.

Sit with the counterfactual for a second, because it reframes everything that follows. A company that came within a decision of selling for scraps in 2016 was valued at $970 million three years later and $1.5 billion two years after that. The distance between “next to nothing” and a billion dollars was one refusal to sell, made by a founder and a CFO.

That founder would be removed four years later. That’s not irony. It’s the mechanism.

Act Three: The Money Arrives

The rounds, in order:

• Seed — January 2014, $1M (Y Combinator, SV Angel, Initialized)

• Series A — June 2014, $5M, led by A.Capital Ventures

• Series B-1 — November 2015, $17M, led by Crosslink

• Series B-2 — September 2017, $23M, led by Catalyst Investors

• Series C — December 2018, $37.5M, led by Lead Edge Capital

• Series D — October 2019, $70M, led by Tiger Global, at a $970M post-money valuation

Source: Crunchbase data compiled by TechBuzz at the time of the IPO.

Disclosed primary capital into the company: $153.5 million. You’ll see $168 million in some analyst writeups, and Rodman’s own bio claims more than $300 million. The bigger numbers include secondary purchases — shares moving between investors, putting nothing in the company’s bank account.

The Series D is the number everybody quotes. Tiger Global led $70 million at a $970 million post-money valuation in October 2019.

Which sets up a sentence I badly wanted to write and won’t: Weave is selling for less than Tiger paid. True on its face. Not a fair comparison. A venture post-money is the price of the most senior, most protected shares applied to every share in the company — and Weave’s preferred carried a liquidation preference of about $159 million that the founders’ and employees’ shares did not. Gornall and Strebulaev put the average overstatement at 48%, with common shares overvalued by 56%. The common was never worth $970 million. The IPO is the honest baseline, because a public price is the only number in this story that everyone actually had to pay.

Still — October 2019 was the peak. Weave was near $100 million in ARR, had gone from about 300 employees to over 550 in a single year, and Rodman had been named Utah Business CEO of the Year and ranked the #4 CEO in America by Glassdoor with a 99% approval rating. He’d built a culture people wrote about publicly and unprompted, under a hashtag — #peoplenotemployees — that went on billboards.

Nine months later, the board took his job.

Act Four: The Removal, and the Bell

Rodman’s own account, published in Utah Business two years later:

We had just raised our Series D round at a $970-million valuation... Within a couple of months of raising that round, some of our board members invited me in and said, “We think it’s time to bring in a new CEO.”

The timing in that sentence doesn’t quite work — the Series D closed in October 2019 and the change was announced in July 2020, nine months later. The likeliest reading is that the conversation started early and the execution took until mid-2020, as IPO preparation accelerated and COVID arrived.

The sequence:

• July 2020: Rodman moves from CEO to Chairman. Jeff Lyman (CPO) and Marty Smuin (COO) serve as interim co-CEOs.

• December 1, 2020: Roy Banks is named CEO — a payments-industry veteran, formerly a CEO Partner at Tritium Partners and president of Truckstop’s LoadPay unit. He was hired to take the company public.

• June 2021: Rodman resigns as Chairman and from the board entirely. Clint Berry leaves at the same time.

Rodman’s public framing has been consistently gracious — he says it wasn’t the decision he’d have made but it was theirs to make, and that he’s grateful to the investors who believed. He’s also been honest about the cost: he’d put everything into it, his family was inside it, it was his identity whether he wanted it to be or not, and having it taken away was hard.

What is not in the public record: which board members drove it, and why. No triggering event has ever been reported. The investor base by then was Tiger Global, Pelion, Lead Edge, Catalyst, Crosslink, and Bessemer. This remains, six years later, the single most under-reported thing about Weave.

Then came November 11, 2021.

Weave listed on the NYSE. Roy Banks rang the opening bell surrounded by the C-suite and the board. Not one of the three founders was on the podium.

Rodman posted a photograph of himself and his wife on the sidewalk outside, under the enormous blue Weave banner draped across the exchange. Passersby were taking pictures of the banner, he wrote, without knowing that the two people standing there had started the company. He described the bell ringing inside while people celebrated on television.

Co-founder Clint Berry stated publicly that Rodman hadn’t even been invited.

The reaction from the Utah and YC worlds was immediate and unusually blunt. Garry Tan — then managing partner at Initialized Capital, an early Weave investor, and a YC partner when Weave went through the program — posted that new management had replaced the founder and lacked the basic decency to invite the founders who’d exceeded growth goals for twelve consecutive quarters, and told the new management to have some class. Jon Bradshaw wrote that the founders took the company 99% of the way and new management came in and took the glory. Weber State’s Alex Lawrence wrote that he couldn’t be excited about a landmark Utah IPO when the founders were outside taking their own photos.

Pelion’s Modersitzki took the other tack, publicly thanking all three founders — and it was in that post that the 2016 near-sale surfaced.

Weave, in its post-IPO quiet period, said only that it was extremely grateful for its founders’ contributions and that they were the heroes of the company’s story.

Be precise about what’s established here: “not invited” is a claim made by a co-founder and corroborated by the founder’s own account of standing outside. The company never confirmed it and never denied it. That distinction matters, and most retellings collapse it.

One more thread from that reporting, which I’d treat as unverified: an anonymous source told TechBuzz that more than 60% of the employees present at the October 2019 Series D had left before the November 2021 IPO. Weave said that figure was inaccurate — though TechBuzz’s own review of LinkedIn found leadership turnover had been very high. Named senior departures in 2021 alone included the VP of Product, the Chief Architect, and the Director of IT Operations.

Act Five: The Trapdoor Nobody Financed

Here is where this story diverges hard from the other three.

Weave priced its IPO at $24 on November 10, 2021 — below its marketed $25-to-$28 range — selling 5 million shares for $120 million. Goldman Sachs, BofA and Citigroup led. The company itself acknowledged demand wasn’t strong.

On November 11 it opened at $21.80 and closed at $18.79. That’s roughly a 22% first-day drop, which one analyst tracking the cohort called the largest IPO-day decline among software companies. Weave was barely above $100 million in ARR — one of the smallest software IPOs on record, at the exact top of the 2021 multiple cycle.

By mid-2022 the stock was around $3.50, with an intraday low of $2.91 on July 1, 2022. At $3.50 that’s an equity value of roughly $230 million — but Weave was sitting on well over $100 million of cash and short-term investments, so the enterprise value was only about $145 million.

Strip that down and look at what it means. Eight months after raising $120 million in an IPO, the market was valuing Weave’s entire operating business — the software, the 27,000 locations, the brand, the fourteen years of work — at barely more than the cash it had just handed the company.

Nothing about that required debt. There was no sponsor, no covenant, no lender club. A software company that had never turned a GAAP profit was priced at peak-cycle multiples, the cycle turned, and the multiple did the work that leverage does elsewhere. In Pluralsight’s case seven private credit funds opened the floor. Here it was the entire public market, and it opened faster.

The management churn kept pace. Roy Banks stepped down on August 15, 2022, citing health and family reasons — the only stated reason in the record, and I’ll take it at face value. But the governance-tracking service exechange noted the context without editorializing: the announcement followed a 71% decline in the share price since December 2021. Brett White — a board member since mid-2020, President and COO since April 2022, formerly COO and CFO at Mindbody — took over as interim in August and permanently on October 4, 2022.

Two CEOs hired to run a public company. One gone within nine months of the listing.

And then — this is the part that makes Weave hard to write about — the business actually got good.

Revenue, growth, and free cash flow by year:

• 2021 — $115.9M, up 45%. Free cash flow negative.

• 2022 — $142.1M, up 23%. Gross margin 66.2% in Q4. Free cash flow −$15.9M.

• 2023 — $170.5M, up 19.9%. Gross margin 69.1% in Q4. Free cash flow turns positive at +$6.5M.

• 2024 — $204.3M, up 19.9%. Gross margin 72.1%. Free cash flow +$10.4M.

• 2025 — $239.0M, up 17.0%. Gross margin 72.1%. Free cash flow +$12.9M.

• 2026 (guided) — $273–275M, up roughly 15%.

Gross margin climbed roughly six hundred basis points. Free cash flow swung from negative sixteen million to positive thirteen. Q2 2026 delivered $67.5 million in revenue, up 15.5%, with a GAAP net loss of just $4.3 million and non-GAAP net income of $3.3 million. White’s strategic claim on that call was that Weave owns the most defensible position in SMB healthcare — the communication layer between the patient and the practice.

Weave has still never posted a GAAP-profitable year. But by 2026 it was a real, cash-generating, 72%-margin vertical software business.

So why did it get cheap?

Because of one number the press releases lead with less often.

Dollar-based net retention was 124% in 2021. It was 99% at the end of 2022. It was 93% at the end of 2025. Gross retention over the same window went from 94% to 89%.

Read that gross number again. Roughly eleven percent of Weave’s revenue base walks out the door every year before a single upsell. Everything above the waterline — the growth, the margin expansion, the cash flow — is being produced by a sales organization refilling a leaking tank faster than it drains.

And this is exactly where the Instructure comparison earns its keep.

Canvas is a system of record. A university does not switch its gradebook. Gross retention in the mid-to-high 90s, switching costs measured in academic years, and that is why $2.275 billion of acquisition debt sits on it as a staircase rather than a trapdoor.

Weave is a system of action. It’s the front desk — the phone, the texts, the reminders, the payment request. Its switching costs are more real than skeptics allow, because Weave replaces the practice’s actual phone system and hardware, which is a genuinely miserable thing to rip out. That’s a better moat than Solutionreach or Podium ever had.

But it is not the gradebook. It sits on top of the practice management system rather than being one — and the PMS vendors know it. Henry Schein owns Dentrix, the leading dental practice management system. Henry Schein One also owns Lighthouse 360 and Demandforce, which compete directly with Weave. The company whose database Weave depends on also sells Weave’s competitors and can bundle them.

That is the structural difference between 89% gross retention and 95% gross retention. And 89% is a much thinner floor to put leverage on.

Act Six: Four Months

The end came fast, and the sequence is worth laying out precisely.

March 28, 2026: Weave signs a cooperation agreement with activist investors Engine Capital, L.P. and 2717 Partners LP. The board expands from eight seats to ten. Edward Robson — founder of 2717 Partners, previously at Siris Capital and The Riverside Company — and Ryan Dubin join as Class III directors. The company agrees to search for a third independent director with software operating experience, agrees not to re-nominate certain incumbents, and caps the board at ten without the investors’ consent. Olshan Frome Wolosky, the most active law firm in shareholder activism, represented both investors.

And the board forms a new Finance Committee — chaired by David Silverman, with Tyler Newton, Robson and Dubin — explicitly focused on long-term shareholder value.

Here’s the detail I keep turning over. Engine Capital’s Q1 2026 13F shows a brand-new position of 1,051,525 shares, worth about $4.86 million — roughly 1.3% of the company. A newly established, roughly one-percent stake converted into two board seats and a new finance committee inside a single quarter. A private equity veteran joined a public board that then ran a sale process. I found no public 13D, no letter, no campaign deck demanding a sale. Whatever pressure existed, it was applied privately.

August 6, 2026: Weave reports a Q2 beat — and discloses that a go-to-market reorganization produced a bookings shortfall from May through July that will hit second-half revenue. The stock falls 8.33% after hours to $6.11.

August 18, 2026 — twelve days later: Weave announces it is selling itself to Francisco Partners at $7.40 a share.

A note on that $650 million, since I’ve been picky about measures elsewhere: Weave’s own release describes it as “an aggregate equity valuation of approximately $650 Million.” That’s a fully diluted equity value — $7.40 against the 79.9 million shares outstanding is about $591 million, and the difference is options, RSUs and PRSUs that accelerate on a change of control. Net of roughly $78 million in cash and no debt, the enterprise value is closer to $570 million.

Board chair Stuart C. Harvey Jr. said the board conducted a thorough evaluation of strategic alternatives and spoke with a number of strategic and financial parties. The board approved unanimously. Jefferies advised Weave; Orrick was Weave’s counsel; Kirkland & Ellis represented Francisco Partners. The acquisition entities are Willow Parent, LLC and Willow Merger Sub, Inc. The deal is expected to close in Q4 2026, and Weave will stay in Lehi and keep its name.

One clause in that press release is genuinely unusual and I want to flag it before anyone writes the cynical version of this story:

As of the date of the Agreement, no executive officer has entered into any agreement with Francisco Partners to roll over equity, invest alongside the buyer, or acquire an equity interest in the surviving company.

No management rollover. No SC 13E-3 going-private filing required. This is not, on its face, a management buyout dressed up as an arm’s-length deal. That distinguishes it sharply from the Instructure process, where the CEO’s $22 million accelerated equity package sat in the middle of the frame the entire time.

No sale process leaked before the announcement. I checked Reuters, Bloomberg, Axios, PE Hub, Semafor and The Information; Axios Pro covered it only after the fact. No competing bidder has been named.

Act Seven: The Buyer

Francisco Partners is one of the most active private equity investors in healthcare IT, with two decades in the sector. Its healthcare portfolio, past and present, includes Availity, eSolutions, GoodRx, Kyruus, QGenda, Zocdoc, Office Ally, CoverMyMeds, Edifecs, and Nextech. Ezra Perlman, FP’s Co-President, is the named executive on the Weave deal and on most of that list.

Two facts about the buyer belong in this story.

First, FP is deploying at scale right now. In July 2026 it closed roughly $21 billion — Fund VIII at $16.4 billion against a $14 billion target, plus Agility IV at $4.6 billion — the largest fundraise in its 27-year history. Eight days before the Weave deal, on August 10, it agreed to buy the Canadian payments processor Moneris from BMO and RBC for approximately C$2.0 billion. Weave is the SMB-healthcare node in a payments consolidation strategy, which is exactly how FP described it: AI, payments, revenue cycle management.

Second — and this is the callback — Francisco Partners bid for Instructure in 2024 and lost to KKR. It was the underbidder on the last company I wrote about. Two years later it has Weave.

FP’s healthcare track record is, on the evidence, good. It sold QGenda to Hearst in August 2024 in a deal valued at up to $3 billion; industry reporting put FP’s return at roughly 15x on an eight-year hold. That is not a firm with a Pluralsight-shaped hole in its history.

But every one of those outcomes was determined by something not yet public here.

What I’d Watch

1. The financing structure. This is the entire piece.

As of today, Weave has filed a press-release 8-K and soliciting materials. The merger agreement itself, the go-shop provision if there is one, the termination fee, the debt commitment letters, the arrangers, and any ratings action on the borrower entity are all undisclosed. They’ll come with the Item 1.01 8-K and the preliminary proxy — watch Weave’s EDGAR page.

When they do, ask the Pluralsight-versus-Instructure question directly. A broadly syndicated term loan B, sold to CLOs and loan funds with hundreds of holders, is a staircase. A concentrated private-credit club deal — or worse, a recurring-revenue loan lending against ARR — is what took Pluralsight, and on a business with 89% gross retention it would be an aggressive thing to underwrite. FP has $21 billion of fresh equity and doesn’t need to lever this hard. Whether it does anyway is the story.

2. Gross retention, not net retention.

GRR is the load-bearing number. 94% to 89% in three years is the trend that made this company cheap. Under private ownership it disappears from public view — but you’ll see it indirectly in whether FP is still adding locations, and in whether Weave stays a buyer of AI companies or becomes a seller of product lines.

3. The “Background of the Merger” section in the DEFM14A.

This will tell you when Francisco Partners first made contact, how many parties were approached, whether Engine and 2717 initiated the process or arrived after it started, and whether any strategic buyer — Henry Schein, EverCommerce, Tebra, Birdeye — ever showed up. It will also contain the Item 402(t) golden parachute table for Brett White. Given the Instructure process, that section is where the uncomfortable facts live.

4. TrueLark.

Weave bought TrueLark (legally Vidurama, Inc.) on May 16, 2025 for about $35 million — roughly $25 million cash and $10 million in equity, with 931,438 shares issued and a $2.3 million indemnification holdback. Key TrueLark employees got performance RSUs of up to a million shares vesting on revenue milestones at the first and second anniversaries. In the Q3 2025 10-Q those PRSUs are described as vesting on conditions; by 2026 they’re described as awards that “were scheduled to vest.” I want to be careful — that’s a tense change in a footnote, not a disclosure — but it reads like the first milestone wasn’t hit. TrueLark powers the AI Receptionist product that is the entire growth narrative FP just bought. Whether those people stay through a take-private is an early tell.

5. A dividend recapitalization.

Same tell as Instructure. If FP borrows against Weave to pay itself before there’s an exit, the story changes character and you can read this whole deal backwards.

The Thing I Actually Learned

I started this series believing the story was private equity kills companies. Instructure taught me the more useful version: leverage is a multiplier, and what it multiplies is whether the revenue will still be there in five years.

Weave taught me something narrower and worse. You don’t need leverage to open a trapdoor. You just need a mispricing and enough time.

Nobody borrowed against Weave. There was no sponsor, no covenant, no seven funds in a room. It carried no debt at all. And a company with real customers, improving margins and positive cash flow still lost 69% of its share price and more than half its equity value between its IPO and its sale — while more than doubling revenue. The mechanism was a public market that priced $100 million of ARR at somewhere north of ten times revenue in November 2021 and, four years later, would pay two.

Which means the extraction event in this story wasn’t the private equity deal. It was the IPO — and the people who paid for it were the 2021 buyers, the employees holding RSUs struck against a $1.5 billion valuation, and, in a different currency, the three founders who weren’t on the podium.

Francisco Partners is not the villain here. It’s the ambulance. It is paying a 34% premium for a company the public market already broke.

But notice what that means for the series thesis, because it doesn’t retire it. It re-arms it. Weave has precisely the profile the thesis flags as dangerous — sub-100% net retention, sub-90% gross retention, no GAAP profit, a system of action rather than a system of record, and a practice management vendor that owns both its data layer and its competitors. The floor is thin. It just hasn’t had any weight put on it yet.

That’s the fork. FP can run the QGenda playbook — invest in payments and RCM, arrest the retention slide, sell it to a strategic in five years at a multiple of what it paid. Or it can put a Pluralsight-shaped capital structure on an 89%-gross-retention business and find out what happens.

Nothing in the public record tells you which. Everything in the merger proxy will.

The most Utah part of this story is still the beginning: $5,000, an attic that got hot, and a decision in 2016 not to sell for next to nothing — a decision that created about a billion dollars of value and cost the man who made it his job.

Ask me again after the proxy.

If you worked at Weave through the 2020 leadership change or the IPO, or you were in the room for the 2016 sale conversation, I’d like to hear from you. I’m particularly interested in who the 2016 buyer was.

A Note on Where the Founders Went

Brandon Rodman launched Previ in January 2022, a fintech built around getting employees at small companies the kind of negotiated pricing that only big-company employees get. It raised $6.5 million from Sound Ventures and Ryan Smith, and now includes a product called Rally.ai. In 2025 he pitched it from a stage inside Weave’s building, to four hundred HR leaders, at his old company.

Clint Berry co-founded Kolla, which builds API and data-warehouse infrastructure on top of dental practice management systems — with Jason Newman, Weave’s former Chief Architect. They are building the integration layer that constrained Weave, as an independent product.

Jared Rodman mentors founders at the Convoi accelerator in Utah.

Utah Business wrote about Rodman’s removal, alongside other founders who lost their companies, in “Falling from grace” in August 2023.

Nvidia invests in OpenAI. OpenAI buys Nvidia chips. Nvidia books the revenue, the stock rises, and the higher valuation funds the next investment. Repeat with CoreWeave, with xAI, with Anthropic. Money leaves Nvidia’s balance sheet, travels in a circle, and comes back as revenue.

It’s a good story. It’s also, in its broad strokes, true. The deals are real, the dollar figures are enormous, and they are increasingly disclosed in SEC filings rather than whispered about. In September 2025, Nvidia and OpenAI announced that Nvidia “intends to invest up to $100 billion in OpenAI progressively as each gigawatt is deployed.” A month later, AMD issued OpenAI a warrant for up to 160 million AMD shares — roughly a tenth of the company, exercisable at a penny a share — vesting as OpenAI deploys AMD hardware. In April 2026, Amazon committed up to $25 billion more to Anthropic, on top of $8 billion already invested, while Anthropic committed to spend more than $100 billion on AWS over ten years.

So when people ask “is this circular financing a bubble?”, I understand the impulse.

I just think it’s the wrong question — and that asking it this way makes people argue past each other for hours without noticing they’re answering two entirely different things.

The two questions hiding inside one

When you ask whether an infrastructure boom is “real,” you are actually asking two separable questions:

One: will this buildout leave behind durable, useful capital? Data centers, power infrastructure, chips, the accumulated engineering knowledge of how to operate them at scale.

Two: will the people financing it get their money back?

The entire circular-financing debate proceeds as though these have the same answer. They frequently don’t. And the historical record is not ambiguous on this point — it’s close to emphatic.

Exhibit A: the fiber that nobody used

Between roughly 1997 and 2001, telecom carriers laid an extraordinary quantity of optical fiber across the United States. The most commonly cited figures — repeated widely enough in the financial press that I’d treat them as good-order-of-magnitude estimates rather than audited data — put it around 80 million miles, of which something like 85 to 95 percent sat unused. “Dark,” in the industry’s term.

The financial outcome was carnage. Corning went from roughly $100 a share in 2000 to roughly a dollar in 2002. Ciena’s revenue fell from about $1.6 billion to about $300 million.

The infrastructure outcome was the modern internet.

That dark fiber didn’t evaporate. It sat in the ground, was bought out of bankruptcy for cents on the dollar, and became the physical substrate for YouTube in 2005, AWS in 2006, and Netflix streaming in 2007 — businesses whose economics only worked once bandwidth had collapsed in price, which happened precisely because it had been catastrophically overbuilt by people who lost everything building it.

Answer to question one: yes, spectacularly. Answer to question two: no, catastrophically. Same event.

(A useful piece of hygiene here: one of the load-bearing justifications for that buildout — the claim that internet traffic was doubling every hundred days — was simply false. AT&T’s Andrew Odlyzko measured actual backbone traffic and found it doubling roughly once a year. Bubbles run on numbers nobody checks. Worth remembering when we get to the current cycle’s numbers.)

Exhibit B: the railroad built on outright fraud

The Union Pacific’s construction was financed through Crédit Mobilier, a construction company owned by the railroad’s own promoters, which contracted with the railroad at inflated prices and distributed discounted stock to sitting congressmen. This is not a case where circularity resembled self-dealing. It was self-dealing, and it became one of the defining corruption scandals of the nineteenth century.

The railroad got built anyway.

Robert Fogel’s The Union Pacific Railroad: A Case in Premature Enterprise (1960) reached the conclusion that gives this whole framework its shape: the UP was socially profitable despite being privately premature. Built too early, financed corruptly, ruinous for many of its investors — and worth it, in aggregate, to the society that got a transcontinental railroad out of the deal.

Fogel also supplies the necessary corrective against getting too romantic about this. His broader estimate of the social savings from all American railroads came in at under five percent of 1890 GNP. Enormously valuable, not civilizationally indispensable. That’s the honest version, and I’d rather cite the honest version.

Exhibit C: China’s solar industry, losing money on purpose

More recent, and closer to the current situation in that it involves manufacturing capacity rather than track or fiber. Chinese solar manufacturers, propelled by state-directed capital, built overwhelming global capacity and then destroyed their own margins doing it. Reuters reported the industry lost around $60 billion in 2024; the China Photovoltaic Industry Association noted at a December 2024 conference that 39 of 121 listed solar companies posted net losses over the first three quarters, and that dozens of firms had delisted, gone bankrupt, or been acquired.

China now dominates global solar manufacturing.

Value created for the world: immense. Value created for the shareholders who funded it: substantially negative.

So why is the SaaS comparison the wrong one?

This was my own starting hypothesis, and I want to be straightforward that I think it was wrong — or at least that it was answering a different question than the one I thought it was.

The intuition goes: 2010s SaaS was circular too. Venture-funded startups bought each other’s software, vendor venture arms invested in their own customers, and the whole ecosystem passed the same VC dollars around. It worked out. Therefore AI’s circularity is normal.

The first half of that is broadly right. The problem is the conclusion, because the structural difference between SaaS and AI is the whole ballgame:

SaaS was asset-light. AI is asset-heavy.

When a venture-funded startup churned off Datadog, Datadog’s growth decelerated. That hurt. It did not leave Datadog holding a warehouse of rapidly depreciating physical equipment purchased on the assumption that the customer would still exist. Software gross margins absorb demand shocks in a way that a data center full of GPUs on a five-year depreciation schedule simply cannot.

The 2022 SaaS drawdown was severe — multiple compression, growth deceleration, a lot of destroyed paper wealth. But no major SaaS vendor failed because its startup customers stopped paying. That survivability wasn’t luck. It was a direct consequence of not having sunk hundreds of billions into depreciating hardware.

Which means SaaS is the wrong analog. There’s a much better one, and it’s less comfortable.

The right analog: 1999 telecom vendor financing

In the late 1990s, equipment makers discovered that their customers — the competitive local exchange carriers building out new networks — couldn’t afford their equipment. So the equipment makers lent them the money to buy it.

Contemporaneous reporting from 2001 put Lucent’s financing commitments in the range of $7–8 billion, Nortel’s at about $3.1 billion committed with $1.4 billion actually drawn, and Cisco’s at roughly $2.4 billion in customer loans. A McKinsey estimate later cited around $25.6 billion in combined exposure across nine suppliers by the end of 2000.

Then the customers went bankrupt. By some counts, 24 of the 30 largest publicly traded telecom carriers filed.

The instructive part is what happened to the vendors’ accounting. In May 2004, the SEC announced that Lucent had “fraudulently and improperly recognized approximately $1.148 billion of revenue and $470 million in pre-tax income” during fiscal 2000, and imposed a $25 million penalty — notably, for lack of cooperation with the investigation. Nine Lucent officers were charged.

This is the pattern worth internalizing: vendor financing lets you book revenue today against a receivable that may not exist tomorrow. The revenue is recognized at shipment. The risk sits quietly on the balance sheet. When it goes bad, it comes back through the income statement.

That’s the correct historical rhyme for what’s happening now. Not SaaS.

But — and this genuinely matters — the mechanism today is different

Here’s where I have to argue against my own analogy, because the strongest counterargument is a good one.

Lucent’s loans sat on Lucent’s balance sheet, which is why its customers’ failures nearly became its own. Nvidia’s arrangements are structurally different in two ways:

They’re mostly equity, not trade credit. An equity loss is capped at carrying value and lands in non-operating income. It never inflated revenue in the first place, so it can’t reverse through revenue. This is meaningfully safer than a vendor loan.

A large share of the financing is third-party capital. Nvidia has reportedly signed memoranda with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to build financing platforms — vehicles funded by Wall Street rather than by Nvidia’s balance sheet. The Financial Times‘s Lex column has argued these are better described as “back-to-back” rather than circular: Nvidia books real revenue and holds a distinct asset at each step, rather than the money dissolving into itself.

There’s a related point about Oracle that gets consistently lost. Oracle appears in every circular-financing diagram, but Oracle is not receiving Nvidia’s capital — it’s borrowing from the bond market. Oracle disclosed $523 billion in total remaining performance obligations as of its Q2 FY2026, up 438% year over year, and separately raised $30 billion in investment-grade bonds and mandatory convertible preferred stock within days of announcing its intent. That’s a debt-funded buildout, not a circular one. Lumping it in overstates how much of the loop is actually Nvidia’s own money.

And the most-cited quantification of the circularity — NewStreet Research’s estimate, reported by Fortune in September 2025, that for every $10 billion Nvidia invests in OpenAI it sees about $35 billion in GPU purchases or lease payments — comes from a subscription research boutique that hasn’t published its methodology. It may well be right. It is also unverifiable from outside, partly by design: Nvidia’s 10-Q discloses two direct customers at 23% and 16% of revenue, but explicitly defines direct customers as “add-in board manufacturers, distributors, ODMs, OEMs, and system integrators.” The end customers are invisible in the filings. Nobody outside Nvidia can actually trace this.

Jensen Huang’s position, to a Bloomberg interviewer earlier this year: these are “generational companies,” Nvidia’s stake is “a small percentage of the amount of money that they ultimately have to go raise,” and the idea that it’s circular is “ridiculous.”

Where the line actually is between “aggressive” and “fraud”

This distinction gets blurred constantly, and it shouldn’t be, because there’s a clean legal test.

Round-tripping is when two parties transact reciprocally, in secret, for the purpose of inflating both sides’ reported revenue. That’s fraud. Qwest is the canonical case: the SEC found that between 1999 and 2002 Qwest “fraudulently recognized over $3.8 billion in revenue,” using capacity swaps in which Qwest bought network capacity from other carriers in exchange for those carriers buying capacity from Qwest. $250 million penalty.

What’s happening now is not that. The AI deals are one-directional, disclosed in filings, and structured as equity. OpenAI genuinely wants the chips. Nobody is booking phantom revenue from a reciprocal handshake.

While we’re here, one correction to the popular narrative, because it’s the kind of thing that gets repeated in these debates: Global Crossing, the other name people invoke alongside Qwest, never produced a fraud judgment. It filed Chapter 11 in January 2002 with roughly $12.4 billion in debt, and the SEC’s action against three former officers ended in $100,000 settlements each for reporting and disclosure violations. Founder Gary Winnick was never charged, and DOJ closed its criminal probe. The “fraud” framing rests on a whistleblower account and congressional hearings, not a court finding. The record is looser than the story.

So what would actually tell us which way this goes?

Here’s what I’d watch, in rough order of signal quality.

1. Do the financing terms loosen? This is the single best tell, and it’s the one that preceded the telecom collapse. Vendor financing existing is not the warning sign. Vendor financing being extended to weaker counterparties on weaker security in order to keep them buying is the warning sign. So far, terms appear to be holding. If backstops start expanding down the credit curve, that’s the moment.

2. Does the capex-revenue gap narrow or widen? Allianz Research put the divergence at roughly 46% in March 2026 — and specifically noted this exceeds the 32% divergence during the 2001 telecom cycle. Sequoia’s David Cahn framed the same problem in “AI’s $600B Question” by taking Nvidia’s run-rate revenue and doubling it twice, once for total data center cost and once for end-user gross margin. Capex is currently accelerating faster than revenue. Moody’s and Goldman both project hyperscaler capex crossing $1 trillion in 2027.

3. Does equity financing drift toward debt? Equity absorbs losses quietly. Debt cascades. The reported movement toward guarantees, private credit, and GPU-collateralized structures is the thing to track, because it converts a capped loss into a contagion mechanism.

4. Does the depreciation argument resolve? Michael Burry’s thesis — that hyperscalers depreciating GPUs over five to six years are understating expenses by roughly $176 billion across 2026–2028 — is live and contested. Nvidia sent analysts a seven-page rebuttal; Burry responded: “I stand by my analysis... I am not claiming Nvidia is Enron. It is clearly Cisco.” Note what he’s actually claiming there. Cisco wasn’t a fraud. Cisco was a real company with a real product whose stock took fifteen years to recover. That’s a much more interesting accusation than the one people think he’s making.

Whether he’s right depends on empirical facts that aren’t in yet — specifically, whether older GPUs genuinely cascade down to inference workloads at economically meaningful utilization. We’ll know from secondary-market pricing before we know it from anyone’s argument.

5. Do the announced deals actually close? A striking amount of the headline arithmetic is conditional. The $100 billion Nvidia-OpenAI figure is a letter of intent, and Nvidia’s own risk-factor language says there’s “no assurance” a definitive agreement will be reached. The much-cited “$750 billion” figure circulating in the press this summer is an aggregation by reporters, not a filed transaction. Announced is not executed, and the gap between the two is where most of the confusion lives.

The honest ending

I don’t know how this resolves, and I’d be suspicious of anyone who tells you they do.

What I’m more confident about is that the question most people are arguing is malformed. “Is AI circular financing a bubble?” bundles together a question about infrastructure and a question about returns, and the historical record — fiber, railroads, solar — says those come apart routinely. The most likely outcome, if history is any guide at all, is not that one side of the current debate is vindicated. It’s that both are: the data centers get built and get used, and a lot of the capital that built them doesn’t come back.

If that sounds like a dodge, consider what it would have meant to be right about fiber in 1999. You could have been completely correct that the buildout was economically insane and the financing was reckless, shorted the whole sector, made a fortune — and been completely wrong about whether the fiber mattered. Both were true. They were always going to both be true.

The useful thing isn’t to pick a side. It’s to be clear about which question you’re answering, and to notice when the person you’re arguing with is answering the other one.

Corrections and sourcing quarrels welcome — I’d rather fix something than be quietly wrong. A few claims that circulate widely in this debate I couldn’t verify well enough to include, which is its own kind of data point.

Domo burned roughly seven dollars to buy one dollar of recurring revenue and then sold its business to Progress Software. Pluralsight went from a $3.5 billion unicorn to a lender takeover that wiped out Vista Equity Partners entirely — killed not by its product but by its capital stack, and specifically by a $1.175 billion recurring-revenue loan from a club of private credit funds.

Both stories have the same shape. Smart founders build something real in Utah. Institutional money arrives. The money reorganizes the company around the money. The company dies.

So here is a company that ran almost exactly the same play and didn’t die.

Instructure — the Salt Lake City company that makes Canvas, the learning management system sitting underneath roughly half of American higher education by enrollment — has been through four ownership changes in nine years. Public in 2015. Private in 2020. Public again in 2021. Private again in 2024. Each buyer paid more than the last. It currently carries about $2.275 billion in acquisition debt.

By the Pluralsight pattern, it should be dead. It isn’t. It’s the market leader by a wider margin than at any point in its history.

Understanding why turns out to be the most useful thing I’ve learned doing these. The difference isn’t that Instructure escaped private equity. It absolutely did not. The difference comes down to two things: what kind of loan sat on the bottom of the stack, and whether the business underneath could carry it.

Act One: The Class Assignment

In the spring of 2008, two graduate students at Brigham Young University sat in a software business course taught by a visiting adjunct named Josh Coates. Coates had made his money founding and selling the online backup company Mozy. The assignment he gave was blunt: find the worst software supporting online education.

Brian Whitmer and Devlin Daley picked the thing they used every day — Blackboard’s Vista learning management system.

Then they built the replacement. [1]

Coates funded them and, by 2010, was running the company. He’d stay CEO until 2018 and board chairman through 2020. Canvas launched in early 2011, and in February of that year Instructure did something that looked insane at the time: it released the core product as open source under an AGPL license. Coates described it as the fastest way to get the software into the market.

It worked. In December 2010 the Utah Education Network announced it would replace Blackboard across Utah institutions. By the end of 2013 Canvas was at 300-plus colleges, universities and districts with 9 million users. Venture money followed — Epic Ventures early, then OpenView, then a $30 million Series D led by Bessemer in 2013, then $40 million from Insight in 2015 that Coates openly called a pre-IPO round. About $90 million total before the company went public.

On November 13, 2015, Instructure IPO’d at $16 a share — the bottom of its $16-to-$18 range, raising roughly $81 million including the overallotment. [2] Revenue had gone from $8.8 million in 2012 to $44.4 million in 2014. Losses had gone from $18.5 million to $41.4 million over the same stretch. Classic land-grab software economics.

Whitmer had already stepped back from operations in 2014. Daley had left around 2013. By the time the interesting part started, neither founder was running the company.

Act Two: The Revolt

On December 4, 2019, Instructure announced it had agreed to be acquired by Thoma Bravo for $47.60 a share — about $2.0 billion.

The press release called it an 18% premium. It was an 18% premium to the three-month volume-weighted average price as of October 27. Measured against where the stock actually traded the morning of the announcement, it was roughly 10% below market.

Shareholders noticed immediately, and what followed was one of the loudest public revolts against a private equity take-private in recent memory.

Rivulet Capital (5.23%) went public on December 5, attacking a three-week strategic alternatives process run over the Thanksgiving holiday. Praesidium Investment Management (about 7.5%) wrote the board on December 12 that the process “was rushed, lacks transparency, and is potentially riddled with conflicts of interest.” Sachem Head, Lateef Investment Management, and Oberndorf Enterprises joined. Roughly a third of the shares were against.

Then, on January 27 and 29, 2020, both ISS and Glass Lewis recommended shareholders vote no. [3] Two proxy advisors rejecting the same PE take-private is rare. ISS pointed to the conduct of the process, an uncompelling valuation, and a strong standalone case.

Instructure fired back the same day, calling its process “exhaustive, conflict-free, and well-publicized” and noting that a post-signing go-shop had produced no serious interest above $47.60. [3] Lead Independent Director Lloyd “Buzz” Waterhouse wrote shareholders accusing an unnamed activist of perpetuating a false and misleading narrative.

Underneath the valuation fight was something uglier.

In March 2019, at an investor conference, CEO Dan Goldsmith had described what Instructure was sitting on: “the most comprehensive database on the educational experience in the globe.” He framed that data — every click, every submission, every grade, from millions of students — as central to the company’s multibillion-dollar value. Usable for training algorithms and building predictive models. [4]

When the Thoma Bravo deal was announced, more than fifty academics signed a public letter demanding a legally binding commitment to protect student data. [4] Their concern was simple: whatever privacy promises Instructure had made, a new owner wasn’t obviously bound by them.

(A later putative class action alleging Instructure had assembled intimate digital dossiers on children was dismissed; the judge called it an ambitious pleading that defeated its own claims.) [5]

And then there is the part I keep rereading.

Goldsmith stood to collect $473,000 in cash and roughly $22 million in equity that would vest instantly on a sale. In April 2019 — after the new incentive plan was in place and after the sale process had begun — he had hired his own sister, Jennifer Goldsmith, as Chief Strategy Officer.

Now watch what happened when the deal started to fail.

The February 13, 2020 special meeting was headed for defeat. Preliminary tallies showed the $47.60 deal losing. Instructure convened the meeting and immediately adjourned it to the next morning. [6] Thoma Bravo came back on February 12 with $48.50 in a two-step tender structure; the board — with Goldsmith and one other director recused — rejected it, worried about closing certainty given feedback from over 20 stockholders and the ISS and Glass Lewis reports. [7]

On February 14, Thoma Bravo raised to $49.00 and called it best and final. [8] To get there, it required Goldsmith to forfeit roughly half his equity package. He did.

Then on February 18, the structure changed. Instructure announced the special meeting scheduled for February 25 would no longer be held. Instead, Thoma Bravo would commence a tender offer under Section 251(h) of Delaware law. [9]

Read that again. The shareholder vote was going to fail, so the transaction was restructured into a form that didn’t require a shareholder vote — only that a majority of shares be tendered.

Dan Goldsmith announced his resignation the same day.

The tender expired March 20, 2020 with 24,828,913 shares tendered, about 64.4%. [10] The deal closed March 23. Dan Goldsmith, Jennifer Goldsmith, and an SVP were all out.

Act Three: The Round Trip

What Thoma Bravo did over the next fifteen months is the standard playbook, executed briskly.

January 2020: roughly 100 jobs cut, mostly from Bridge, the corporate learning product. May 2020: another 150-plus, about 12% of the workforce. Roughly a fifth of full-time staff gone in a year. Satellite offices in Chicago, Seattle, San Diego and the large London EMEA headquarters were closed, consolidating everything to Salt Lake City and remote.

Bridge — the entire corporate-learning half of the company — was exited. Instructure became a bet on Canvas alone.

In July 2020, Steve Daly became CEO. He had previously run LANDESK, which he sold to Thoma Bravo in 2010. The buyer brought in its own guy.

Then, in July 2021, Instructure went public again at $20 a share, selling 12.5 million shares for about $250 million. [11]

Here is where almost everyone gets the story wrong, so let me be precise: $20 in 2021 is not a loss against $49 in 2020. The company was recapitalized in between and the share count is entirely different. Thoma Bravo bought the equity at roughly $2.0 billion and refloated it at roughly $2.9 billion fifteen months later. That is a good trade, not a bad one.

Thoma Bravo also kept about 88% of the company at IPO, drifting toward 84% — meaning Instructure was a “controlled company” under NYSE rules. Public shareholders bought a minority sliver of something Thoma Bravo still ran.

And the money went where you’d expect. Thoma Bravo affiliates had bought $129.2 million of Instructure’s own term loans in the 2020 buyout financing, holding $131.6 million by March 2021. The IPO proceeds were used to repay debt. So roughly one-sixth of what public investors put in went straight back to Thoma Bravo as a lender.

The edtech analyst Phil Hill, who has tracked this company longer and more closely than anyone, told EdSurge at the time that he didn’t consider it as impactful as most IPOs — “it’s really financial management.” [12]

He’d later put it more sharply: “Think financial engineering more than strategy.” [13]

Act Four: The Ladder

In February 2024, Instructure closed its acquisition of Parchment, the largest academic credentialing platform, for about $835 million. [14] It was funded with cash and a large incremental draw on the existing credit facility.

The effect on the balance sheet was immediate. Total debt went from $491.3 million at the end of 2023 to $1.173 billion by March 31, 2024. Net leverage went from 0.7x to 4.7x in a single quarter.

The company was, at that moment, genuinely healthy. FY2023: $530.2 million in revenue, up 11.6%, with $214.2 million of adjusted EBITDA — a 40.4% margin. Through the first nine months of 2024, revenue was $499.1 million, up 26%.

Three months after Parchment closed, Reuters reported Thoma Bravo was exploring a sale. [15]

On July 25, 2024, Instructure announced it would be acquired by KKR, with Dragoneer participating, for $23.60 a share — an enterprise value of roughly $4.8 billion. [16] A 16% premium to the unaffected price of $20.27 on May 17. Thoma Bravo was cashed out completely, with no rollover. KKR had beaten Francisco Partners in the auction. Steve Daly stayed. The stated goal: $1 billion in revenue by 2028.

No shareholder vote was required. When one investor owns 84% of the company, the majority written consent is the vote.

The deal closed November 13, 2024, and INST stopped trading. [17]

Now — the part that matters more than everything above it.

How the $4.8 billion was financed.

Milbank, which represented the joint lead arrangers, describes it plainly: a $2.275 billion first- and second-lien credit facility package. [18] Broken out:

• $1.685 billion seven-year first-lien term loan B, led by Morgan Stanley, at roughly S+325 to 350

• $365 million eight-year second-lien term loan, led by KKR Capital Markets, at roughly S+525 to 550

• $225 million revolver due 2029, with a springing first-lien leverage covenant

Bookrunners included UBS, Jefferies, RBC, HSBC and Mizuho. The borrower entity is Icon Parent Inc. S&P rated it B- corporate, B- first lien, CCC second lien; Moody’s B3, B2, Caa2. Stable outlooks.

That is an ordinary, broadly syndicated leveraged buyout loan. It gets sold to CLOs and loan funds. Hundreds of holders. Liquid secondary market.

Pluralsight’s fatal loan was none of those things.

Pluralsight was financed with a $1.175 billion recurring revenue loan — a structure that lends against ARR rather than EBITDA — held by a concentrated private-credit club: Blue Owl, Ares, Golub, Oaktree, Benefit Street, Goldman Sachs and BlackRock. Seven parties. When the covenants tightened, those seven could sit in one room, coordinate, and take the company. Vista’s equity went to zero.

You cannot do that to a broadly syndicated term loan B. There is no room. There are hundreds of holders with different bases and different incentives, and no one of them has the standing to seize an asset.

Same ladder. Different rung material. Pluralsight’s last step was a trapdoor. Instructure’s is a staircase.

Act Five: Why It Actually Survived

The financing structure is half the answer. The other half is that Canvas is a genuinely dominant business, and that changes what leverage means.

In May 2025, Edutechnica — which has tracked US higher-ed LMS market share for twelve years — reported that “for the very first time,” Canvas held greater market share than its next three competitors combined. [19] Blackboard, Moodle and D2L Brightspace, added together, are smaller than Canvas.

Measured by enrollment rather than institution count, Phil Hill’s data puts Canvas at roughly 50% of North American higher-ed enrollment, against D2L at 20%, Blackboard at 12% and Moodle at 9%. [20] Every school in U.S. News’s 2026 top ten national universities runs Canvas.

Then in September 2025, Anthology — the owner of Blackboard — filed for Chapter 11. It emerged debt-free in February 2026, but the largest incumbent competitor spent the year in bankruptcy court while Canvas collected its refugees.

This is what a system of record looks like. Gross retention in the mid-to-high 90s. Switching costs measured in academic years, not quarters. Forty percent EBITDA margins.

And that is the entire difference. Domo had sub-100% net revenue retention and burned cash to replace churning customers. Pluralsight sold corporate skills content in a market that got crowded and price-competitive fast. Both were carrying debt structures that assumed the revenue would hold. It didn’t.

Instructure got loaded with the same kind of debt and the revenue held, because a university does not switch its gradebook the way a company switches a training vendor.

Private equity didn’t spare Instructure. It just picked a company that could survive it.

Because make no mistake — the playbook ran in full. Twenty percent of staff cut in 2020. Offices closed. An entire product line exited. Development work offshored, notably to Hungary. Another round of layoffs reported in September 2025 under KKR. The student-data monetization thesis that alarmed fifty academics. Four ownership changes in nine years, each one extracting a fee.

The company survived that. It doesn’t follow that the people inside it had a good decade.

Act Six: The Two Things That Could Still Break It

I want to be careful here, because “the reckoning hasn’t arrived” and “there will be no reckoning” are different claims and I can only defend the first.

The breach.

On April 29, 2026, Instructure detected unauthorized activity. It disclosed publicly on May 1. The extortion group ShinyHunters had exploited a vulnerability in the support-ticket path of Instructure’s free “Free-for-Teacher” signup environment and exfiltrated what it claims is 3.65 terabytes of data — roughly 275 million records across 8,809 institutions, including usernames, email addresses, course names, enrollment information and private messages. [21]

Instructure believed it was contained. It wasn’t. On May 7 — finals week — Canvas login portals at roughly 330 institutions, including Harvard, Penn, Duke and Wisconsin, were defaced with extortion messages and a May 12 deadline. [21]

On May 11, Inside Higher Ed reported that Instructure had negotiated with the attackers and obtained the return of the compromised data, under a settlement reportedly including a “shred logs” clause — the attackers agreeing to destroy what they took. [22] One secondary account puts the payment near $10 million; that figure is not confirmed, and Instructure has not published an independently verified victim count. TechCrunch, working from the group’s own leak-site claims, reported 231 million unique email addresses — a different unit of measure than “records,” which may explain part of the gap but not all of it. [23]

Instructure permanently shut down Free-for-Teacher accounts and announced an eighteen-month security remediation program.

It has been called the largest education data breach on record. Class actions have been filed. Nine thousand institutional contracts contain security provisions, and if even a fraction of those tenants invoke them, the cost dwarfs any ransom.

Here is why this is the thing to watch rather than a footnote: Instructure’s entire debt-survivability case rests on renewal. Mid-to-high-90s gross retention is the load-bearing assumption underneath $2.275 billion of leverage. A breach doesn’t move revenue this quarter. It moves the renewal conversation two and three years out, which is exactly when this debt matters.

The AI question.

On July 23, 2025, Instructure announced a global partnership with OpenAI and launched a platform called IgniteAI. [24] The flagship feature is the “LLM-Enabled Assignment” — a custom GPT-like experience built inside Canvas, where the instructor defines how the AI talks to students, sets the learning objectives, and specifies what evidence of learning to capture. That evidence flows automatically into the Gradebook. OpenAI is the launch partner; a bring-your-own-key model for other LLMs is planned. It’s free to Canvas customers, who supply their own API key.

There are two ways to read this.

The pessimistic read is that a learning management system is a thin layer — a place to submit files and see grades — and a sufficiently good AI tutor plus a sufficiently good AI grader disintermediates it. Google Classroom already sits on 170 million-plus Workspace users, Chromebooks and Gemini. OpenAI has ChatGPT Edu. On this reading, the OpenAI partnership is Instructure paying rent to its eventual replacement.

The optimistic read is that the LMS is the system of record — the FERPA layer, the roster layer, the transcript layer, the thing the registrar and general counsel both depend on — and no chatbot replaces that. On this reading, the AI cheating crisis actually strengthens Canvas, because the institution now urgently needs a trusted place where AI interaction is observed and recorded. Which is precisely the product Instructure shipped.

I don’t know which is right. Neither does KKR. But note that the optimistic case is the one the debt requires.

What I’d Watch

Instructure is private, so there are no more 10-Qs. The last audited numbers are the FY2023 10-K and the 2024 quarterlies through September. Everything since is estimate and inference, and I’d treat any post-2024 revenue or EBITDA figure you see — including in this piece — as unaudited.

But four things would still show up in public:

A downgrade. S&P or Moody’s taking Icon Parent Inc. negative would be the first hard signal. Instead, the opposite happened: in 2025, S&P affirmed its ratings alongside a roughly $1.3 billion first-lien repricing. A company repricing its debt into a friendlier market is not a company in trouble.

A dividend recapitalization. This is the tell. It’s the move where the sponsor borrows against the company to pay itself. If KKR does a divi recap on Instructure, the story changes character entirely, and you can read the last five years backwards.

Organic growth below high single digits. The $1 billion-by-2028 target implies roughly 11% a year off a ~$660 million base. Higher-ed sales cycles have lengthened. If organic ARR growth stalls, the debt paydown math stops working.

Migrations traceable to the breach. Not announcements — actual RFPs from institutions naming security. Edutechnica publishes share data every spring and fall. That’s where it would appear first.

The Thing I Actually Learned

I started this series believing the story was private equity kills companies. Three autopsies in, that’s not what the evidence says.

Domo died of unit economics — spending far more to acquire revenue than the revenue was worth, then never escaping the treadmill. Pluralsight died of loan structure — a recurring-revenue loan held by seven funds who could act as one. Instructure absorbed the same ownership churn, the same layoffs, the same offshoring and the same debt-funded acquisitions, and is standing, because Canvas is a near-monopoly on a system of record with mid-90s retention.

Leverage is not a verdict. It’s a multiplier. On a business with real switching costs and 40% margins, it’s a staircase. On a business with sub-100% net retention, it’s a trapdoor — and the specific structure of the loan determines how fast the floor opens.

Two of these companies are twenty minutes apart on I-15. The people who worked at them went to the same conferences and hired each other’s friends. From the outside, the same thing happened to all three.

It didn’t. And the difference wasn’t the state, or the founders, or even the buyer. It was whether the thing being borrowed against could still be there in five years.

Instructure’s answer, so far, is yes.

Ask me again after the renewals.

If you worked at Instructure through any of the four ownership changes, or you’re at an institution reassessing after the breach, I’d like to hear from you.

Sources

1. Instructure company history and founding — Brian Whitmer and Devlin Daley, BYU, 2008; Josh Coates’s software business course. Corroborated across Instructure corporate history and contemporaneous Utah tech coverage.

2. Instructure Inc., IPO prospectus (Form 424B4), November 2015 — SEC EDGAR, CIK 1355754: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=1355754

3. Instructure Inc., Form DEFA14A, January 27, 2020 — company statement responding to the ISS recommendation: https://www.sec.gov/Archives/edgar/data/1355754/000119312520015505/d837367ddefa14a.htm

4. Nadia Tamez-Robledo / Jeffrey R. Young, “As Instructure Changes Ownership, Academics Worry Whether Student Data Will Be Protected,” EdSurge, January 17, 2020: https://www.edsurge.com/news/2020-01-17-as-instructure-changes-ownership-academics-worry-whether-student-data-will-be-protected

5. Coverage of the Instructure student-data class action and its dismissal: https://injuryclaims.com/news/class-action-privacy-data-and-tcpa/instructure-canvas-student-data-tracking-class-action-lawsuit

6. Instructure Inc., Form 8-K Exhibit 99.1, February 13, 2020 — adjournment of the special meeting: https://www.sec.gov/Archives/edgar/data/1355754/000119312520034519/d878383dex991.htm

7. Instructure Inc., Form DEFA14A, February 2020 — board deliberations on the $48.50 proposal, recusals, and the ISS/Glass Lewis reports: https://www.sec.gov/Archives/edgar/data/1355754/000119312520034499/d886407ddefa14a.htm

8. “Instructure and Thoma Bravo Amend Definitive Merger Agreement to Increase Offer Price to $49.00 Per Share in Cash,” February 14, 2020 — SEC 8-K Exhibit 99.1: https://www.sec.gov/Archives/edgar/data/1355754/000119312520036801/d882741dex991.htm

9. “Instructure and Thoma Bravo Revise Definitive Agreement” (tender offer structure; February 25 meeting cancelled), February 18, 2020 — SEC filing: https://www.sec.gov/Archives/edgar/data/1355754/000119312520040650/d888837dex991.htm

10. Instructure Inc., Form 8-K, March 2020 — tender offer expiration and results (24,828,913 shares at $49.00): https://www.sec.gov/Archives/edgar/data/1355754/000119312520083235/d902911d8k.htm

11. “Instructure Announces Pricing of Initial Public Offering” — 12,500,000 shares at $20.00, July 2021: https://www.thomabravo.com/press-releases/instructure-announces-pricing-of-initial-public-offering

12. Jeffrey R. Young, “Instructure Is Back on the Stock Market, But Not Much Change Expected For Canvas Users,” EdSurge, July 22, 2021: https://www.edsurge.com/news/2021-07-22-instructure-is-back-on-the-stock-market-but-not-much-change-expected-for-canvas-users

13. Phil Hill, “About the KKR Instructure Acquisition Agreement,” On EdTech: https://onedtech.philhillaa.com/p/about-the-kkr-instructure-acquisition-agreement

14. Reuters, “Thoma Bravo explores sale of Instructure” — Parchment acquired for $835 million: https://finance.yahoo.com/news/thoma-bravo-exploring-sale-stake-042845325.html

15. Same as above — Reuters reporting the sale exploration, May 17, 2024.

16. “Instructure to be Acquired by KKR for $4.8 Billion,” July 25, 2024: https://www.instructure.com/press-release/instructure-to-be-acquired-by-KKR

17. “KKR and Dragoneer Complete Acquisition of Instructure,” November 13, 2024 — SEC 8-K Exhibit 99.1: https://www.sec.gov/Archives/edgar/data/1841804/000119312524257230/d878225dex991.htm

18. “Milbank Advises on $2.275B Financing for Acquisition of Instructure Holdings, Inc. by KKR,” Milbank LLP, November 13, 2024: https://www.milbank.com/en/news/milbank-advises-on-dollar2275b-financing-for-acquisition-of-instructure-holdings-inc-by-kkr.html

19. “LMS Data – Spring 2025 Updates,” Edutechnica, May 26, 2025: https://edutechnica.com/2025/05/26/lms-data-spring-2025-updates/

20. Phil Hill, “State of Higher Ed LMS Market for US and Canada: Year-End 2024 Edition,” On EdTech: https://onedtech.philhillaa.com/p/state-of-higher-ed-lms-market-for-us-and-canada-year-end-2024-edition

21. “Instructure Reaches Ransom Agreement with ShinyHunters to Stop 3.65TB Canvas Leak,” The Hacker News, May 12, 2026: https://thehackernews.com/2026/05/instructure-reaches-ransom-agreement.html

22. “Canvas/Instructure cyberattack – Key developments and action items for higher education institutions,” Reed Smith LLP, May 2026: https://www.reedsmith.com/articles/canvasinstructure-cyberattack-key-developments-and-action-items-for-higher-education-institutions/

23. Same as [22] — Reed Smith summarizing TechCrunch’s May 5, 2026 report of 231 million unique email addresses.

24. “Instructure and OpenAI Announce Global Partnership to Embed AI Learning Experiences within Canvas,” July 23, 2025: https://www.instructure.com/press-release/instructure-and-openai-announce-global-partnership-embed-ai-learning-experiences

Additional reference: “Instructure to be acquired by KKR for $4.8B,” Higher Ed Dive, July 2024: https://www.highereddive.com/news/instructure-kkr-acquisition-ed-tech/723082/ · “KKR and Dragoneer Complete Acquisition of Instructure,” KKR: https://media.kkr.com/news-details?news_id=04bc08e0-3954-48cc-b85a-9ba5ae6e6c84 · InstructureCon 2025 product announcements: https://www.instructure.com/resources/blog/instructurecon-2025-partner-product-announcements

Previously in this series: “How Domo Burned a Fortune, Missed Its Moment, and What Happens Next” · “Josh James Just Ran the Same Play Twice” · “What Happened to Pluralsight, Utah’s $3.5 Billion Unicorn?”

In December 2021, Airtable raised at an $11.7 billion post-money valuation.

That’s an 81% haircut on the equity value, and a 89% haircut if you measure the operating business against its peak. It is one of the largest destructions of paper value in the 2021 SaaS cohort, and it happened quietly, in August, in an all-cash deal that most people will read as a headline number without understanding what it actually means.

I want to walk through what it means. Specifically: the waterfall math, the strange carve-out buried in the SEC filing, and whether you should be worried about the product you’ve built your business on.

Part 1: The deal, from the filing

The terms come from Bending Spoons’ Form 6-K filed with the SEC on August 4. The structure:

Bending Spoons US Inc. is acquiring 100% of the shares of Formagrid Inc. — that’s Airtable’s legal name — from Formagrid Holdings LLC. All cash. Both boards approved unanimously. The deal is subject to regulatory clearance with an outside date of February 4, 2027, and the company expects to close before the end of this year.

Advisors: Willkie Farr & Gallagher and EY for Bending Spoons; Latham & Watkins and AXOM Partners for Airtable. Goldman Sachs Bank Europe and J.P. Morgan are co-financial advisors on the buy side.

Bending Spoons CEO Luca Ferrari described Airtable as “a pioneering brand reshaping how teams organize data,” and disclosed that ARR is growing over 20% year-over-year to roughly $480 million as of June 2026. Airtable serves more than 500,000 organizations, including 80% of the Fortune 100.

So: a $1.285 billion enterprise value on ~$480 million of ARR. About 2.7× revenue.

That’s the number to sit with. Airtable is not a dying business. It’s growing 20% a year with ~90% gross margins and 170% net dollar retention in enterprise — retention that beats Asana and Monday.com as public comps. And it sold for less than 3× revenue.

Where the money came from

A week before signing, Bending Spoons announced a €500 million SACE-backed term loan from HSBC, Intesa Sanpaolo, and BPER, maturing March 2031. That was part of roughly €1.49 billion in new and expanded facilities, earmarked in the company’s own language for “general corporate purposes and acquisitions.”

No filing explicitly ties that money to Airtable. But the timing is not subtle.

Part 2: The waterfall — who actually gets paid

This is the part that matters if you hold equity, and it’s the part almost every news writeup skipped.

Airtable raised more than $1.4 billion across nine rounds. Here’s the stack:

Seed — February 2015. About $3 million, led by Freestyle Capital and Caffeinated Capital. No disclosed valuation.

Series A — June 2015. About $7.6 million, led by CRV. No disclosed valuation.

Series B — March 2018. About $52 million, co-led by CRV and Caffeinated Capital. No disclosed valuation.

Series C — November 2018. $100 million at roughly $1.1 billion post-money, with CRV, Coatue, and Benchmark.

Series D — September 2020. $185 million at $2.5 billion, led by Thrive Capital.

Series E — March 2021. $270 million at $5.77 billion, led by Greenoaks. Six months after the Series D, at more than double the price.

Series F — December 2021. $735 million at $11.7 billion post-money, or roughly $187.28 per share, led by XN, with Franklin Templeton, Salesforce Ventures, MSD Capital, Silver Lake, T. Rowe Price, and J.P. Morgan joining the existing syndicate. Nine months after the Series E, at another 2× step-up.

Note the compression at the end. The first $160 million took six years to raise across four rounds. The last $1 billion took nine months across two.

The Series F is the one that hurts, and the $187.28 per share is the number to hold onto.

Reporting on the cap table indicates the preferred is 1x non-participating — the founder-friendly standard. In a great exit, preferred converts to common and everyone shares the upside. In a mediocre exit, preferred takes its money back off the top and common eats the difference.

At ~$2.25 billion of equity value against a ~$1.4 billion preference stack, the preferences clear. Which means:

• Late investors (Series F) get roughly their money back. Not a loss on paper capital, but a catastrophic outcome measured against a $11.7B mark held for four and a half years. Zero return on a five-year hold in a period when the S&P did what it did.

• Early investors do fine. CRV, Freestyle, Caffeinated bought in at valuations under $1.1 billion. They’re up meaningfully.

• Common — founders and employees — split what’s left. Evan Armstrong’s analysis at The Leverage and waterfall modeling circulated by Hustle Fund’s Angel Squad put the residual to common at roughly $850 million. That’s real money. It’s also about a 10× haircut from where common was marked at peak.

The illustrative per-share number in the Angel Squad model came out around $25.53 for common. Compare that to $187.28 for the Series F, and to secondary trades that were printing in the low $40s earlier this year.

Employee equity is the real casualty. Options and RSUs granted between 2021 and 2024 were struck at 409A values reflecting an $11.7 billion — or later, a ~$4 billion — company. Anyone who joined during the peak and exercised options is likely deep underwater. Anyone holding unexercised options above the clearing price holds nothing.

The 6-K does not disclose per-share common consideration, RSU treatment, option treatment, any tender offer, or 409A figures. Everything in this section beyond the $2.25 billion headline is estimated or reported, not confirmed. Treat it accordingly.

The $965 million nobody’s talking about

Here’s the arithmetic nobody highlighted: enterprise value $1.285 billion, equity value $2.25 billion. The gap is net cash of roughly $965 million, as Tech Startups worked out.

That’s derived, not disclosed. But it tracks with Howie Liu’s own comments earlier this year that roughly half the venture capital Airtable raised was still in the bank.

Read that again. Nearly half of what shareholders are receiving is money they already put in. Bending Spoons is paying $1.285 billion for the operating business and handing back the unspent war chest to the people who funded it.

Part 3: The carve-out

Buried in the filing is the sentence that reframes the entire deal:

“assets and liabilities relating to the ‘Hyperagent’ business line were transferred by the Company to Hyperagent Inc.”

Before signing, Airtable reorganized to spin Hyperagent out into a separate entity. Hyperagent Inc. is a named party to the purchase agreement.

Hyperagent was Airtable’s agentic AI product, launched February 2026 — the successor to Superagent (January 2026), which succeeded Omni (June 2025), which was the centerpiece of what Liu called Airtable’s “refounding” as an AI-native company.

So Bending Spoons is buying the database, the 500,000 organizations, the Fortune 100 logos, the $480 million of recurring revenue, and the ~$965 million of cash. It is not buying the frontier AI bet.

The most plausible read: Liu doesn’t want to be the person who lays off his own company, and he does want to keep chasing the agentic-AI frontier with a clean balance sheet and a fresh cap table. Selling the mature business to an operator who specializes in extracting cash from mature businesses, while retaining the option value, is a rational move for a founder who’s watched his valuation fall 88%.

It is a considerably worse deal for whoever holds Airtable common.

Part 4: How Airtable got here

Founded 2012–13 by Howie Liu, Andrew Ofstad, and Emmett Nicholas. Liu had already sold a company — Etacts, a YC-backed CRM, acquired by Salesforce in December 2010 — and spent time at Salesforce before starting Airtable. The founders built for three years before a 2014 invite-only beta and a March 2015 public launch.

The product was genuinely novel: a spreadsheet you could think in, backed by a real relational database, without needing to know what a relational database was. It grew bottom-up, then moved upmarket.

Then the positioning started churning. Interface Designer (2021). “Connected Apps Platform” (2022). “Next-generation app platform” with AI workflows (2023). Cobuilder, prompt-to-app, its fastest-adopted feature ever (2024). Omni and the AI-native refounding (2025). Superagent, then Hyperagent (2026).

Six repositionings in five years is not a company that found its second act. It’s a company that kept looking for one.

Meanwhile the headcount came down hard. Roughly 254 people (~20%) cut in December 2022, with the CRO, CPO, and Chief People Officer departing. Then 237 more (~27%) on September 14, 2023, as Liu refocused on million-dollar-plus enterprise accounts.

The refocus worked, financially. ARR went from ~$375M (2023) to ~$478M (2024) to ~$480M (mid-2026). The company reportedly crossed $100 million in free cash flow after the AI restructuring.

But growth decelerated from 27% to 20%, and the competitive picture got worse from both directions — Notion, Monday.com, Smartsheet, ClickUp on one side; Replit, Lovable, Bolt, Cursor and the frontier chat models on the other.

Armstrong’s line is the best summary of the strategic problem I’ve read: “The company’s original interface once eliminated the need to understand databases. Coding agents now eliminate the need to understand Airtable.”

Airtable’s moat was abstraction. AI commoditized abstraction.

Part 5: So who is Bending Spoons, and is this just PE?

Bending Spoons was founded in 2013 in Copenhagen by Luca Ferrari, Matteo Danieli, Francesco Patarnello, and Luca Querella, then moved to Milan. It grew out of a failed diary app called Evertale. The founders’ takeaway was that finding product-market fit is mostly luck, but operating a product that already has it is a learnable skill. Their first acquisition cost $10,000.

The numbers now are not small. Revenue went from $387M (2023) to $671M (2024) to $1.31 billion (2025) — an ~84% CAGR — with $601 million in Q1 2026 alone, per company filings. Over 500 million monthly active users and 9 million paying customers. It IPO’d July 1, 2026 at $29/share and closed at $40.50 that day, a 39.7% pop, raising $1.68 billion at roughly a $25.7 billion market cap.

The acquisition list is long: Evernote (2023), FiLMiC (2022), Meetup (2024), Mosaic’s app portfolio from IAC (2024), WeTransfer (2024), Issuu (2024), StreamYard/Hopin assets (2024), Brightcove (~$233M, 2025), Komoot (~€300M, 2025), Harvest (2025), Vimeo ($1.38B, 2025), AOL from Yahoo (~$1.4B, closed January 2026), Eventbrite (~$500M, closed March 2026), Tractive (2026). And now Airtable.

The company says it has never sold a material business and has identified more than 1,000 potential future targets.

Is it private equity? Ferrari’s own answer, quoted in Colin Keeley’s operating-manual writeup: “a bit of a mix of a private equity, maybe 25%, and 75% a technology company.”

I think the honest classification is: a leveraged roll-up operator. Closer to Constellation Software than to Thoma Bravo, but running the same financial machinery. Bloomberg has called it “private equity for apps.” It genuinely does rewrite codebases, re-platform infrastructure, and hold indefinitely rather than flip on a five-year fund clock — that’s a real difference from a buyout shop.

But the debt is real and compounding. Total debt sat around $4.36 billion against ~$741 million cash at March 31, 2026. Interest expense was $143 million in 2025 and $93 million in Q1 2026 alone — roughly 78% of that quarter’s operating income. The ratings agencies have it at B+/B2.

That is the number I’d watch as a BSP shareholder. Not revenue growth, which is acquisitional and therefore purchasable. Interest as a share of operating income.

Part 6: Should Airtable users expect the product to get worse?

Let’s be empirical about this. There are eight-plus prior acquisitions with observable outcomes.

The consistent pattern is deep layoffs and price increases. Evernote cut 129 people in February 2023, then most of its US and Chile staff by mid-2023, moving operations to Europe — Financial Times reporting put the headcount decline at roughly 82%, from 341 to 60 by end-2024. The free tier was cut to 50 notes total. The personal plan reportedly went from about $69.99 to $129.99.

WeTransfer lost roughly 75% of staff within weeks of the 2024 close, capped the free plan at 10 transfers per month, and in July 2025 pushed a terms-of-service change granting AI-training rights that triggered enough backlash it was reversed within days. Brightcove let go over 85% of its ~200 employees. Meetup paywalled aggressively enough to generate sustained “enshittification” complaints. Vimeo laid off nearly its entire global workforce, including the video team, by January 2026.

But the pattern is not uniformly bad for product. Komoot lost ~75% of staff and both founders after the 2025 acquisition — and subsequently shipped a well-received Garmin integration, with monthly actives reportedly at all-time highs. Evernote, after the rebuild, is credited by some longtime users with better stability and faster shipping than it had in its final independent years. Bending Spoons says retention across acquisitions stays “remarkably stable.”

Both things are true. Prices go up, staff goes down, and most users don’t leave, because switching costs on a database you’ve built your operations inside are enormous.

Which is exactly why Airtable is such a good acquisition. 80% of the Fortune 100 use it. Those are precisely the customers who can least afford to migrate and can most afford to pay more.

What I’d actually do if I were running on Airtable

Not panic. But not wait, either.

1. Export everything now. Full base exports plus documented schemas. Do it before close, not after.

2. Inventory your API and automation dependencies. This is the real risk surface — not the UI. If your business logic runs through Airtable’s API and third-party integrations, map it and build a fallback path.

3. Do not sign a long prepaid multi-year deal without price protection. If your renewal is coming up, negotiate a hard cap on increases and explicit data-portability and SLA terms before the deal closes. Your leverage is highest right now.

4. Price an alternative even if you don’t move. Notion, Smartsheet, Monday.com, or a Postgres-plus-app-builder stack. Knowing your migration cost is what makes a renewal negotiation real.

5. Watch for three specific tells post-close: a renewal quote up more than 20–30%, any cut to free or entry-tier limits, or any API deprecation notice. Any one of those means the standard playbook is running and you should be actively planning.

And the counter-signal: if Bending Spoons publicly commits to price and API stability, retains the core product team, and keeps investing — the Komoot outcome rather than the Vimeo one — the downside case genuinely softens. Absence of any such commitment within six months of close should be read as confirmation.

The bigger point

Airtable was a good business that raised too much money at too high a price, and the gap between those two facts destroyed roughly $9.5 billion of paper value and most of what its employees thought their equity was worth.

The company grew 20% a year at 90% gross margins with 170% enterprise net retention and generated over $100 million in free cash flow. In almost any other financing environment, that’s a company that stays independent, or IPOs at a reasonable multiple, and everyone does fine.

Instead it raised $735 million at 75× revenue in December 2021, and the preference stack that money created became the thing that ate its employees’ outcome.

The venture math only works on the way up.

Sourcing note: Deal terms, financing, and Bending Spoons financials are from SEC filings and company releases and are confirmed. The waterfall analysis, per-share common estimates, and the $965M net-cash figure are derived or reported from secondary sources and are explicitly not confirmed by the filing — the 6-K discloses no per-share common consideration, RSU/option treatment, or 409A values. Post-acquisition layoff percentages and pricing changes for prior deals come from a mix of company statements, press reporting, and user forums; Bending Spoons declines to confirm several of the headcount figures. Airtable’s founding year is variously reported as 2012 or 2013; public launch was March 2015.

It is a Florida not-for-profit corporation. You can read its articles of incorporation on Sunbiz, filed May 19, 2025, document number N25000006299. Journalist Dave Levinthal obtained and reported on the seven-page filing the day it landed, noting it was sent to Florida regulators by fax.

Its stated purpose, per that document, is to steward, preserve, and celebrate the legacy and historical record of Trump and his presidency.

Eric Trump is its president — he is quoted as such in the foundation’s own announcement, carried by CBS Miami, saying the completed library will be a landmark on Miami’s skyline. Madison Faupel is treasurer, which is how congressional investigators addressed her in July 2026. The foundation’s principal address is 115 Eagle Tree Terrace, Jupiter, Florida.

The detail most coverage skips: this is the second entity.

A predecessor called the Donald J. Trump Presidential Library Fund, Inc. was incorporated December 20, 2024 — after the election, before the inauguration. Florida administratively dissolved it on September 26, 2025 for failure to file a mandatory annual report, and articles of dissolution followed on December 29, 2025.

No public explanation has ever been given for why the first entity was created, allowed to lapse, and replaced by a nearly identical one. As the April 2026 congressional letter puts it, no Fund board members were ever appointed, and the only person to sign any of its public documents minimized his role, saying his job was simply to be the registrant. Dave Levinthal’s OpenSecrets investigation broke that story in October 2025.

Where is the money coming from? The settlement list

Trump sued Meta, Twitter/X, and Google/YouTube in July 2021 over his post–January 6 account suspensions. He separately sued ABC over defamation and CBS/Paramount over an edited “60 Minutes” interview. After the 2024 election, the defendants settled in sequence.

Company Date Amount Destination ABC / Disney Dec. 14, 2024 $15M + $1M fees Library, as a “charitable contribution” Meta Jan. 29, 2025 ~$25M ~$22M to the library X (Twitter) Feb. 13, 2025 ~$10M Library Paramount / CBS July 2, 2025 $16M Library and plaintiffs’ fees Google / YouTube Sept. 29, 2025 $24.5M $22M to the White House ballroom, not the library

The ABC settlement is worth reading closely. Per the congressional summary of the agreement, the payment was to be made as a charitable contribution “to a Presidential foundation and museum to be established by or for” Trump — meaning the recipient entity did not yet exist when the money was promised.

Do not misquote the Google settlement in either direction. Per CBS News, Alphabet agreed to pay $22 million directed into the Trust for the National Mall to support construction of the White House State Ballroom, with $2.5 million going to other plaintiffs including the American Conservative Union. The bulk of that one never touched the library.

How much has the library foundation actually raised?

This is where it gets genuinely murky, and where I’d urge caution before repeating any single number.

Trump’s own federal financial disclosure — the OGE Form 278e received June 29, 2026, at page 869 — reports that $56.5 million from Meta, ABC, and Paramount was paid to the Foundation, with an additional $8 million from X that the disclosure does not clearly allocate. Total: $64.5 million.

The foundation’s own Florida filing — the Solicitation of Contributions Annual Financial Reporting Form filed with FDACS on October 10, 2025 — reports $50 million in revenue under “all other contributions, gifts, grants & similar amounts.”

The individual figures don’t reconcile either. The July 2026 congressional letter lays the discrepancies out item by item: ABC reported at $15M but disclosed at $16M; Meta reported at $22M but disclosed at $24.5M; X reported at $10M but disclosed at $8M.

The spending side of that Florida form shows $400,000 for officer compensation, $490,000 in legal fees, $130,000 for conferences and meetings, $15,000 for travel, $50,000 for insurance, and roughly $6 million under “other (list item)” with no further explanation. The officer-compensation line is odd because the foundation’s own small charitable organization registration states that none of the listed personnel, including Eric Trump, are compensated. The form doesn’t even reconcile internally — the letter’s footnote flags that $6.180 million is listed under other “program services” while the total of other expenditures is listed as $5.795 million.

Lawmakers’ bottom line: at least $13.5 million and up to $21.5 million is missing or has been spent with no clear explanation.

Here is the caveat you should know before you repeat that number. NBC’s reporting notes that the $50 million on the Florida form is described as a proposed budget of expected fundraising rather than cash in the bank. If that reading is correct, comparing it against $64.5 million in settlements is not an apples-to-apples subtraction, and the “missing money” framing weakens considerably. The unexplained $6 million in expenditures is a separate and sturdier question.

Also relevant: the same NBC piece reports that ABC’s payment sat in an escrow account set up by Trump’s lawyers until at least March 2026. The companies themselves told Congress they couldn’t say where the money went — Meta confirmed paying $22 million but declined further “confidential” details, and X confirmed $10 million but likewise declined to comment on confidentiality grounds.

The foundation has not yet filed a public IRS Form 990. When it does, that document resolves most of this.

Who else has donated?

The largest publicly known single gift is not a settlement. Politico reported in May 2026 that SoftBank Group donated $50 million to the library and foundation, intended to support a library focus on the U.S.-Japan alliance. SoftBank pledged in 2024 to invest $100 billion in American companies during Trump’s term and is one of the largest AI investors in the world, and it lobbies on AI policy. Congressional Democrats opened an inquiry into that donation in July 2026.

One detail from that reporting is directly relevant to the precedent question: per Politico’s sourcing, SoftBank had previously donated to the Reagan and George W. Bush presidential libraries, but those contributions came after the libraries were already completed.

The broader fundraising context comes from a Wall Street Journal investigation summarized here, which found that Trump has taken personal control of the fundraising and that no sitting American president has ever raised these amounts for his pet projects, with Apple, Microsoft, and Amazon writing large checks toward the ballroom project.

Where will it be built, and who paid for the land?

A 2.63-acre former parking lot on Biscayne Boulevard in downtown Miami, next to the historic Freedom Tower, from Miami Dade College’s Wolfson Campus.

The price was $10, twice.

The Miami Herald’s investigation, “Push to build Trump library on Miami Dade College land shrouded in secrecy,” documented that the college’s board agreed to transfer the land in a special meeting with no debate, no public comment, and no discussion of the fact that it was intended for a presidential library — after which the college’s own president and vice chairman said they weren’t entirely sure what the state wanted to do with the property.

On valuation: the county property appraiser valued the plot at $67 million, but developer Gregg Covin said the commercial value is significantly higher — he estimated between $200 and $300 million in cash.

The sequence took longer than most people realize. Activist and historian Marvin Dunn sued, and a judge temporarily blocked the transfer, with the judge stating the case was not, for that court, rooted in politics. The college held a second public vote on December 2, 2025 after three and a half hours of public comment, and approved it unanimously.

The deeds finally recorded in early 2026. Per county property records reported by NBC Miami, the college transferred the property to a state board overseen by DeSantis for $10 in January, and a month later the state board transferred it to the foundation for $10 via quitclaim deed, with a requirement that construction start within five years.

The foundation’s tax filings indicate it plans to raise almost $1 billion over the next three years, while Trump is still in office.

What about the Qatari plane?

Trump said he would accept a Boeing 747-8 from Qatar’s royal family, valued around $400 million, for use as Air Force One before transfer to the library foundation. The Defense Department officially confirmed acceptance on May 21, 2025.

Attorney General Pam Bondi signed a memo calling it lawful. Per CNN, the Office of Legal Counsel approved the memo before Bondi signed it, and career ethics officials determined there was no conflict requiring her recusal — Bondi previously lobbied on behalf of the Qatari government. Sen. Dick Durbin pressed her on exactly that, arguing the gift violates the Constitution and the statutory regime Congress enacted to govern such gifts.

The memo itself is still not public. American Oversight sued after DOJ granted expedited processing on a FOIA request but estimated fulfillment would take more than 600 days. FactCheck.org’s full write-up is the best neutral summary.

Didn’t Clinton and Bush do the exact same thing?

This is where the online argument usually collapses, because both sides are half right. Here is the honest version, with receipts.

Fundraising while in office is not new. Reagan broke ground on his library on November 21, 1988 — two months before leaving office, with a speech he delivered as sitting president. The National Archives’ own account covers the ceremony.

Foreign government money is not new. The Washington Post’s December 2007 investigation, “Clinton Library Got Funds From Abroad,” found that Clinton’s library raised more than 10 percent of its $165 million cost from foreign sources, and that the Saudi royal family gave about $10 million to the Clinton facility, roughly the same as it gave toward George H.W. Bush’s library. The Post also documented that Kuwait, Qatar, the UAE, Taiwan, and Brunei each gave at least $1 million.

Donor secrecy is not new, and it isn’t a one-party habit. The Post found the Clinton library steadfastly declined to reveal donors, saying they were promised confidentiality. The Chronicle of Philanthropy’s summary adds the useful contrast: the George H.W. Bush and Carter libraries made most foreign donations public, while Reagan’s did not.

Pay-to-play scandals are not new. In July 2008 the Sunday Times of London ran a video sting on lobbyist Stephen Payne. ProPublica’s contemporaneous account reports that Payne told a potential foreign client seeking access to senior administration officials that he ought to donate a couple hundred thousand dollars to Bush’s presidential library to show he was serious — out of a package priced between $600,000 and $750,000. ProPublica’s framing is the key line: the report shined a light on fundraising for presidential libraries, which is unregulated. Payne denied wrongdoing, saying the Times took him out of context and that he had counseled the client that explicit exchange would be bribery.

A president operating outside the NARA system is not new. The Nixon library operated privately from its July 19, 1990 dedication until July 11, 2007, when NARA took it over. History.com’s account is blunt about what that meant: unburdened by NARA’s curatorial standards, the museum included a Watergate exhibit claiming Nixon was the victim of a coup by his political enemies, and featured a heavily edited version of the 1972 “smoking gun” tape. It was overhauled after NARA took over.

For the system itself — the Presidential Libraries Act, the Presidential Records Act, and the escalating private endowment requirements — the Congressional Research Service’s Presidential Libraries and Museums is the standard reference.

So if the argument is “private money, big money, secret donors, foreign donors, or fundraising while in office” — the defenders have a real point. That has been the bipartisan norm, and it has been ugly for a long time.

So what’s actually different this time?

Four things, and they’re specific.

1. The funding source is the president’s own lawsuits against companies he regulates. No prior library was capitalized primarily by litigation proceeds extracted from parties with mergers and regulatory matters pending before that president’s administration. Donations from favor-seekers have deep precedent. This mechanism does not.

2. Corporations may be writing it off. Congressional investigators flagged that if these contributions were all structured as “charitable contributions” as ABC’s was, the companies may have been able to deduct them as a tax write-off. That is a wrinkle no previous library controversy had.

3. The land transaction is larger and stranger than the norm. Donated land is common — Reagan’s 29-acre site was donated by a development partnership. What is unusual is the hotel. Benjamin Hufbauer, a University of Louisville professor who studies presidential libraries, told the Miami Herald that while transferring land in a prime spot isn’t unusual, making part of it a commercial hotel is, and he had never heard of anything like it connected to any other presidential library. He also called the speed of the state transfer unusually fast.

4. The books don’t reconcile in public documents. Reported settlements, the president’s federal disclosure, and the foundation’s state charitable filing produce three different totals — plus a dissolved predecessor entity that nobody has explained, and roughly $6 million in expenditures filed under “other.”

Is any of this illegal?

Separate the appearance from the provable crime, because the answers differ.

Criminal bribery requires evidence of a specific quid pro quo. Suspicious timing, even very suspicious timing, is circumstantial. Notably, Payne himself said the explicit version would be bribery in this country — the legal line has always been the exchange, not the donation.

Disclosure law doesn’t currently reach any of this. Rep. Henry Waxman’s disclosure bill passed the House 390–34 in March 2007 and stalled in the Senate. Then-Senator Barack Obama introduced similar legislation, saying he wanted to avoid even the appearance of impropriety. Neither became law. The current attempt is S.2300, the Presidential Library Anti-Corruption Act, introduced July 2025, which would cap donations and require disclosure while a president is in office.

The pattern is old and bipartisan in its failure: the House votes for transparency, the Senate lets it die, and the next president inherits the loophole. That loophole is not Trump’s creation. What is at issue is the scale at which it’s now being used.

The short version

Private library fundraising is normal. Fundraising while in office is not new. Foreign donors are not new. Secret donors are not new. A president operating outside the NARA system is not new.

What is new is a president suing companies he regulates and directing the settlements into a nonprofit run by his own son — while in office, while those same companies had business before his government — receiving a parcel appraised at $67 million for ten dollars, and filing state and federal documents that produce three different totals.

You can call that a scandal or hardball inside the rules. But “everybody does this” and “unprecedented corruption” are both, on the evidence, about half true. Knowing which half you’re arguing is the whole game.

Primary documents

Read them yourself rather than taking my word for it:

• Foundation articles of incorporation, May 19, 2025 (Florida Sunbiz)

• Fund articles of incorporation, December 20, 2024

• Fund administrative dissolution, September 26, 2025

• Fund articles of dissolution, December 29, 2025

• Trump OGE Form 278e, received June 29, 2026 — settlement income at p. 869

• FDACS Annual Financial Reporting Form, October 10, 2025

• FDACS Small Charitable Organizations Registration Application

• Congressional letter to Trump, April 19, 2026

• Congressional letter to the Foundation treasurer, July 15, 2026

• Letters to ABC, Meta, X, and Paramount, March 9, 2026

• S.2300, Presidential Library Anti-Corruption Act of 2025

• CRS, Presidential Libraries and Museums (IF12781)

• Washington Post, “Clinton Library Got Funds From Abroad,” Dec. 15, 2007

• ProPublica on the Payne affair and unregulated library fundraising, 2008

Corrections welcome and will be posted with attribution. If you obtain the foundation’s IRS Form 990 when it lands, send it — that document answers most of the open questions here.

There’s a version of the Pluralsight story where the company is the hero of Silicon Slopes.

It’s not hard to imagine, because for most of two decades it was true. Four developers put in $5,000 each in 2004 and built a training business with no outside money for nine years. They pivoted to online video before “the creator economy” was a phrase. They took venture capital late, and only after they were already winning. They raised what was then the largest venture round in Utah history, became a unicorn, then an IPO that popped 33% on day one. If you were making the case in 2018 that Utah companies could be built patiently and still win big, Pluralsight was your Exhibit A.

Six years later, Vista Equity Partners — arguably the most sophisticated enterprise software investor on earth — wrote its Pluralsight equity down to zero and handed the keys to a group of lenders led by Blue Owl and Ares. Total loss for Vista and its co-investors: roughly $4 billion. The class-action lawsuit settled for $20 million. The Draper headquarters closed, and the company is now run from Westlake, Texas.

Domo, the subject of my last two pieces, died slowly of its own spending. Pluralsight is a different autopsy entirely. The product worked. The revenue was real. What killed it was the capital stack — three successive owners, each paying for a version of the company that didn’t exist yet, until the debt finally did what debt does.

It’s also, quietly, one of the most important deals in American finance this decade. But we’ll get there.

Act One: Nine years of no (2004–2012)

Pluralsight was founded in 2004 by Aaron Skonnard, Keith Brown, Fritz Onion, and Bill Williams as an in-person classroom training company. Instructors flew to offices and taught developers how to use Microsoft tools. It was profitable, unglamorous, and entirely self-funded — each founder contributed $5,000, and for nine years that was the only equity capital in the business.

In 2007 they made the pivot that mattered: ditch the classroom, put the courses online, charge a subscription. This sounds obvious now. In 2007, years before anyone said “MOOC,” it wasn’t. The pivot turned a services business with linear economics into a library with software economics. One course, recorded once, sold thousands of times.

By the time Insight Venture Partners showed up with a $27.5 million Series A in December 2012, Pluralsight wasn’t a bet. It was a working machine that had never needed anyone’s money.

Remember that detail. Everything that follows is the story of a company that learned to need money.

Act Two: The treadmill starts (2013–2018)

What does a bootstrapped company do with venture capital? In Pluralsight’s case, it went shopping. TrainSignal for $23.6 million. Digital-Tutors for $45 million. Smarterer for $75 million. Code School for $36 million. Between 2013 and 2016 the company bought its way from a Microsoft-centric course library into a broad technology skills platform.

The capital markets rewarded the ambition. In August 2014, Insight, ICONIQ, and Sorenson led a $135 million round — reported at the time as the largest venture round ever raised by a Utah company — that took Pluralsight from a sub-$100 million valuation in 2012 to roughly $1 billion in under two years. Utah had unicorns before, but none that had bootstrapped for nine years first.

The IPO came on May 17, 2018. Priced at $15, closed its first day at $20, up 33%. Skonnard rang the bell on a company valued north of $2 billion that he’d started with $20,000 of founder money. As of that morning, the Pluralsight story was perfect.

It had about fourteen months left.

Act Three: Fifty missing salespeople (2019–2021)

Growth-stage SaaS companies live and die by a simple machine: hire quota-carrying sales reps, ramp them for six months, collect the billings. In January 2019, CFO James Budge told investors Pluralsight had “about 250” quota-bearing reps. The actual number, he would later concede, was closer to 200.

Fifty missing salespeople doesn’t sound like a scandal. But billings are a leading indicator, and sales headcount is the leading indicator of billings. In March 2019, with the stock riding high, the company raised $456 million in a secondary offering. During the same general stretch, per the securities litigation, Skonnard sold $22.2 million in stock, Budge sold $15.1 million, and chief revenue officer Joe DiBartolomeo sold $9.7 million.

On August 1, 2019, the machine’s output became visible: billings growth had decelerated from over 40% to 23%. The stock fell $12.13 — nearly 40% — in a single day, closing at $18.56. “Sales execution challenges” entered the Pluralsight vocabulary and never really left. The shareholder lawsuits were filed within weeks. (The Tenth Circuit would later revive the case after an initial dismissal, and a federal judge approved a $20 million settlement in February 2025 — a melancholy postscript that outlived the public company itself by four years.)

The stock never fully recovered. So in December 2020, with shares languishing, Vista Equity Partners arrived with an offer: $20.26 per share, roughly $3.5 billion. Eminence Capital, one of its largest outside shareholders, called the price opportunistic and publicly campaigned against the deal — and won, sort of. Vista raised its bid 11% to $22.50 in March 2021 and closed the acquisition that April.

Here is the detail that decided everything, and almost nobody outside credit markets noticed it at the time: how Vista paid for it.

Act Four: The loan that ate the company (2021–2024)

Vista funded the buyout in part with a $1.175 billion “recurring revenue loan,” plus a $100 million revolver, from a consortium of private credit funds — Blue Owl, Ares, Golub, Oaktree, Benefit Street, Goldman Sachs, and BlackRock among them.

A recurring revenue loan is exactly what it sounds like, and it is stranger than it sounds. The loan isn’t underwritten against profits or cash flow — Pluralsight didn’t reliably have those. It’s underwritten against ARR: the assumption that subscription revenue is so durable, and growing so fast, that it can service debt all by itself. Regulated banks largely can’t write this kind of loan. Private credit funds, flush with institutional money and unburdened by bank capital rules, wrote it eagerly. At the time, Pluralsight’s was one of the largest recurring revenue loans ever made — a flagship deal for an asset class that was busy convincing the world that lending against growth was safe.

Then Vista doubled down, buying A Cloud Guru — the Australian cloud-training darling — in mid-2021, in a deal the Australian Financial Review pegged among the largest tech exits in that country’s history, with comparisons running from $1.6 billion to $2.2 billion.

Count the assumptions stacked on top of each other in mid-2021: that pandemic-era demand for online tech training was the new baseline, not a spike. That tech hiring would keep growing, so companies would keep buying seats. That interest rates would stay near zero, keeping the floating-rate debt cheap. That ARR only goes up.

Every one of those assumptions broke, more or less simultaneously. Tech hiring froze in 2022 and reversed in 2023. Training budgets are the first line item cut in a downturn. Rates went from zero to five percent, repricing the debt in the worst possible direction. And generative AI arrived to whisper an existential question into the ear of every “learn to code” business: what, exactly, are we training people for?

Revenue flattened. Interest expense didn’t. By May 2024, Vista had written its entire equity position — the full $3.5 billion of value, plus follow-on money — down to zero. The company cut 17% of its workforce that July.

Then came the maneuver that made Pluralsight famous in credit circles. In early summer 2024, the company moved certain intellectual property into a new subsidiary — partially outside the lenders’ collateral net — and Vista injected $50 million against those assets, with the proceeds funding an interest payment to the very lenders whose collateral had just been diluted. Lenders were, to put it mildly, displeased. The move ignited an industry-wide debate about “liability management exercises” migrating from junk-bond land into the supposedly genteel world of private credit.

It bought a few months. It didn’t change the arithmetic.

Act Five: Repossession (2024–today)

In August 2024, the lender group took 100% ownership of Pluralsight. About $1.3 billion of debt was converted to equity, the lenders put in $250 million of new capital, and Vista walked away from a software company with — this almost never happens — nothing. Law firm Goodwin called the deal “the tip of the private credit iceberg,” and it remains the marquee case study of what happens when an ARR loan meets ARR that stops growing.

The aftermath has been a quiet, methodical shrinking of everything the old Pluralsight stood for. Aaron Skonnard, CEO for twenty years, was replaced by Chris Walters in April 2024. In August 2025, the company announced it was relocating its headquarters from Draper to Westlake, Texas, closing the Utah office entirely; the remaining Utah employees work remote. The flagship campus of Silicon Slopes’ most bootstrapped success story now belongs to another state’s press releases.

The business itself, it should be said, is alive. Pluralsight still has thousands of enterprise customers, a 6,500-course library, and a product strategy now aggressively rebuilt around AI skills — betting that the same technology that threatened it will fuel the next training cycle. Its owners are lenders who never wanted to own it, which tells you what happens next: stabilize, restore some growth story, and sell — to a strategic, or to the next private equity firm with a thesis. Credit funds are not natural long-term operators of software companies. The clock is already running.

The lesson

Domo’s failure was a spending problem: a company that paid $7 to acquire $1 of recurring revenue and could never step off the treadmill. Pluralsight’s failure is more unsettling, because Pluralsight mostly did the things you’re supposed to do. Bootstrap. Reach profitability. Raise late. Build real revenue.

What it couldn’t survive was three rounds of owners each underwriting the previous owner’s story instead of the business in front of them. Public investors in 2018 paid for the bootstrapped legend. Vista in 2021 paid for the pandemic growth curve. The lenders underwrote the idea that recurring revenue recurs forever. Each was buying the chart, extrapolated. The company underneath was a good, cyclical, competitive business — which is a fine thing to be, unless someone has borrowed a billion dollars against the assumption that you’re not cyclical.

The most Utah part of the story might be this: the equity that built Pluralsight — four checks of $5,000 — produced a multibillion-dollar outcome. The capital that came later did progressively worse the more sophisticated it got. The venture investors did fine. Public shareholders were taken out at a premium. Vista, the most disciplined buyer in enterprise software, lost everything. And the lenders — the capital engineered to be safest in the stack — ended up as involuntary owners of the company itself.

And one more thing worth sitting with. When the write-off happened, ratings agencies rushed to reassure everyone that losses were “widely dispersed” across dozens of funds — no single lender badly hurt, no ratings actions needed, nothing to see here. That’s either the system working exactly as designed, or exactly how every bubble describes itself right up until it isn’t. Private credit is now an asset class measured in the trillions, built substantially on loans like Pluralsight’s. The iceberg metaphor wasn’t mine. It was the lawyers’.

Sources: Pluralsight corporate history and funding timeline via Wikipedia and Utah Business; Forbes and CNBC coverage of the May 2018 IPO; Cohen Milstein case files on the securities litigation (billings, rep counts, insider sales, $20M settlement, Feb. 2025); KSL on the settlement; Pluralsight and StockTitan press materials on the Vista agreement ($20.26) and amended agreement ($22.50, March 2021); Business Wire statements from Eminence Capital (Jan.–Mar. 2021); TechCrunch on the $3.5B Vista deal; Goodwin Law client alert on the change-of-control transaction ($1.175B recurring revenue loan, $100M revolver, lender group, IP transfer, $1.3B debt-for-equity, $250M new money); Axios and Slashdot on Vista’s write-off (May 2024); Transacted on the lender takeover; Reuters via Yahoo Finance on creditor negotiations; SmartCompany/AFR on the A Cloud Guru acquisition; KBRA via Business Wire on dispersed lender impact; Pluralsight press releases on the Chris Walters appointment (April 2024) and the Westlake, Texas headquarters relocation (August 2025).

In May, I wrote about how Domo burned a fortune and missed its moment. The short version: $715 million of venture capital raised, three straight years of revenue flat around $318 million, an accumulated deficit of roughly $1.55 billion, and — as of the June 10-Q — $39.1 million of cash against $137.1 million of debt reclassified as current, a breached minimum-ARR covenant, a going-concern warning, and a forbearance agreement requiring a signed purchase agreement by July 31, 2026. CFO Tod Crane addressed it directly on the Q1 call: the lender agreed to hold off on remedies only while Domo completed a strategic transaction.

On July 22 — nine days before that deadline — Domo signed one. But the structure is the story.

The deal that isn’t an acquisition

Progress Software (NASDAQ: PRGS) is paying $400 million in cash for — in the words of Domo’s own press release — “substantially all of the assets and employees, excluding the Company’s net operating loss (’NOL’) carryforwards.” The platform, the IP, the customer contracts, employees, vendor relationships, and foreign subsidiaries — everything that makes Domo Domo, including its 2,400+ business customers, goes to Burlington, Massachusetts, with closing expected within Progress’s fiscal year ending November 30, 2026.

But Progress is not buying Domo, Inc. The Delaware corporation that trades on Nasdaq survives. Per the announcement, it will remain a publicly-listed corporation under a new name and ticker, pay off its credit facility, and hold approximately $246 million in net cash — about $4.84 per share, an 81% premium to the 30-day volume-weighted average price — plus the one asset Progress explicitly did not buy: more than $900 million of net operating loss carryforwards.

Josh James remains CEO of the surviving entity. And here’s the sentence in the release worth reading twice: “Mr. James, as the Company’s controlling shareholder, executed an irrevocable consent providing shareholder approval of the transaction.” Public shareholders will not vote on this deal. Because James’s Class A shares carry 40 votes per share against one vote for public Class B — a structure in place since the IPO — he approved the entire transaction himself, by written consent. Shareholders will receive an information statement telling them it already happened.

If your first reaction is “wait, can they do that?” — yes. Delaware law permits action by written consent of a majority holder, and Domo has been a controlled company since 2018. Your second reaction should be the more interesting question: why structure it this way at all?

What an NOL actually is (and why you can’t sell one)

A net operating loss carryforward is exactly what it sounds like: when a corporation loses money, the tax code lets it bank those losses and deduct them against future profits. Lose $900 million over sixteen years, and you’ve built a $900 million deduction waiting for income to offset — at a 21% federal corporate rate, theoretically worth up to ~$189 million in avoided future taxes.

Domo, having lost money every year of its existence, is sitting on one of the larger NOL piles in Utah corporate history. An accumulated deficit north of $1.5 billion produces a lot of tax attributes.

Here’s the catch that shapes this entire deal: NOLs are not a transferable asset. They belong to the corporate entity that generated them. You can’t put them on a truck. And Congress, having watched decades of profitable companies buying dead shells purely to harvest their losses, wrote Section 382 of the Internal Revenue Code specifically to kill the trade.

Section 382 works like this: if a “loss corporation” undergoes an ownership change — its 5%+ shareholders collectively increasing their stake by more than 50 percentage points over a rolling three-year window — annual use of pre-change NOLs gets capped at the company’s equity value multiplied by the IRS long-term tax-exempt rate, currently in the neighborhood of 3.5%. For a small company, that math is brutal. A $250 million shell that trips an ownership change can use roughly $9 million of NOLs per year. The other $891 million becomes wallpaper.

This is why the deal looks the way it does:

1. If Progress bought Domo whole, the acquisition itself would be an ownership change. The $900 million in losses would be capped into near-worthlessness — Progress would be paying for an asset the purchase destroys.

2. So Progress buys the assets instead. The transaction is structured as an asset purchase; the business moves, and the entity — with its tax attributes — stays behind.

3. The asset sale itself generates a taxable gain at the corporate level, which is normally the drawback of asset deals. But Domo’s losses absorb the gain, which is why nearly all of the $400 million (less the debt payoff) lands in the shell as untaxed cash.

4. The shell adopts a “tax benefits preservation plan” — announced in the same press release — an NOL poison pill designed to stop anyone from accumulating a stake large enough to inch toward the 50-point ownership change. These plans, used by companies like INNOVATE Corp to protect tax assets, typically trigger at 4.9% ownership: cross the line without board blessing and your position gets diluted into oblivion.

Every piece of the structure is downstream of one fact: the losses die if the entity changes hands.

Why you almost never see this

Asset sales happen constantly in small-cap M&A. What’s rare is the second half: the seller staying public afterward as a deliberate cash-and-NOL vehicle. Most acquirers want the whole company — cleaner, no rump entity, no ongoing public-company costs. Most sellers don’t have a tax asset big enough to justify the hassle, or a controlling shareholder positioned (and motivated) to keep running a shell. When the stars do align, the track record is a genuinely mixed bag:

The wins. WMIH Corp — the post-bankruptcy husk of Washington Mutual, sitting on NOLs its own filings pegged in the billions — was steered by KKR into a $3.8 billion merger with Nationstar Mortgage in 2018, creating Mr. Cooper Group, a profitable mortgage servicer that could actually use the shelter. Sam Zell’s Equity Group took bankrupt Delta Petroleum’s ~$1.3 billion in NOLs and rebuilt the entity into Par Pacific, now a multi-billion-dollar NYSE refiner. And the granddaddy of the genre: Berkshire Hathaway was a dying textile mill Warren Buffett repurposed into a holding company.

The near-perfect template for Domo. In February 2024, ContextLogic — the corpse of e-commerce flameout Wish — sold its operating business to Qoo10 for approximately $173 million while explicitly preserving approximately $2.7 billion of NOL carryforwards. Read that 8-K next to Domo’s release and the language is nearly word-for-word: limited operating expenses, debt-free balance sheet, board intends to monetize NOLs, will consider returning capital. Two years later, backed by David Abrams’ Abrams Capital and BC Partners, ContextLogic acquired US Salt in a $907.5 million transaction and completed the deal in 2026, rebranding as a “business ownership platform” — an explicit mini-Berkshire. Shell holders have done well. But note the sobering detail: independent analysis pegs the realistically usable value of that $2.7 billion at roughly $1–2 per share — a small fraction of face value — because even a healthy operating business only generates so much taxable income per year to shelter.

The failures. Support.com, a cash-rich shell whose tax assets drew a bitcoin miner as its suitor, merged into Greenidge Generation in September 2021; Support.com holders received just 0.115 Greenidge shares per share, and GREE subsequently collapsed with the crypto-mining bust — the shell’s value was vaporized inside someone else’s story. Enzon Pharmaceuticals wound down its drug business and spent a decade as a self-described public-company acquisition vehicle organized around its NOLs, with Carl Icahn in the cap table — and never landed the transformational deal. Even Icahn can sit on a pile of losses forever.

The base rate is honest coin-flip territory, heavily dependent on the discipline of whoever controls the shell. Which brings us to Utah’s most interesting capital allocator.

The Josh James pattern

Here’s where this stops being a tax-law explainer and starts being a sequel.

Go back to 2018. Domo had raised $715 million of venture capital, essentially all preferred stock with liquidation preferences — the standard VC term that says we get our money back first in any sale. By spring 2018, per its own IPO filings, Domo had $72 million in cash, was burning roughly $37 million a quarter, and admitted that without financing by August 2018, management would “begin to implement plans to significantly reduce operating expenses.”

A private sale at that moment — at any realistic price — would have gone almost entirely to the preference stack. Had Domo sold for $500 million, the VCs’ $715 million of preferences would have consumed every dollar. Common stockholders, the founder included, would have received approximately nothing.

But Domo didn’t sell. It went public — at roughly a $511 million valuation, a 77% haircut from its $2.28 billion private mark. Why take the humiliation? Because an IPO automatically converts preferred stock to common. Domo’s own prospectus spells it out: conversion of preferred is automatic upon a firm-commitment underwritten IPO raising at least $50 million. The moment the offering priced, $715 million of liquidation preferences evaporated. Everyone became a common shareholder — and Josh James’s stake, worthless in a private sale, became worth real money at any valuation, with 40-vote-per-share Class A stock ensuring he’d control the public company indefinitely. I walked through this mechanism in detail in May; it’s the structural trick most 2018 coverage glossed over.

Now look at 2026. Domo is again out of road — covenant breach, forbearance deadline, going-concern language. The conventional endgame is a whole-company sale: buyer pays a premium, shareholders get cashed out, and Josh James’s eighteen-year run as a public-company CEO ends with him walking away holding his pro-rata share and no platform.

Instead: the assets sell, the entity survives, and James emerges as controlling shareholder and CEO of a debt-free public vehicle holding a quarter-billion in cash and $900 million in tax attributes — protected by a poison pill that structurally prevents any outsider from accumulating enough stock to challenge him, with a stated mandate to pursue transactions “where the company can employ its expertise in AI and automation to enhance profitability”.

Twice now, at the exact moment the standard transaction structure would have left Josh James with little or nothing, the transaction took the unusual form that left him with something — a public currency in 2018, a public acquisition vehicle in 2026. That is either a remarkable coincidence or a genuinely consistent skill: James is very, very good at corporate structure when it counts. It’s worth remembering how he got his first fortune: Omniture, the Orem-founded analytics company he took public in 2006, was sold to Adobe in a $1.8 billion all-cash tender offer at $21.50 per share — a clean, conventional exit where every shareholder got the same deal at the same time. He knows what the normal version looks like. He’s chosen the abnormal version twice.

Is that bad for shareholders? Honestly — it’s complicated

The cynical reading writes itself, and parts of it are earned. Minority holders got no vote. The press release says the board will consider “options to return capital to shareholders” — which is a commitment to nothing. Compare ContextLogic, whose announcement explicitly promised that absent opportunities it “intends to promptly return all capital to shareholders”. Domo made no such promise. The written-consent process plus dual-class stock plus the NOL pill means public shareholders have essentially zero recourse if the shell drifts. And James’s history includes the related-party arrangements — the jet lease, the family-linked vendors — that his own IPO-era filings had to disclose and unwind, plus a 2022 board shakeup that briefly moved him out of the CEO chair before he returned.

But the fair reading deserves the column inches too. Against a company that traded near a $156 million market cap the week of the announcement — with $39 million in the bank and a lender holding an acceleration right on $137 million — $4.84 per share in net cash, an 81% premium to the 30-day VWAP, is a better outcome than the realistic alternatives, which included distressed scenarios far uglier than this. A whole-company sale to Progress would have killed the NOLs outright; this structure preserves the option value, and minority shareholders keep it pro-rata alongside James. The ContextLogic playbook shows the option can pay.

The honest frame: Domo shareholders traded a failing software company for a cash box with a lottery ticket stapled to it — and the only person who decides whether the ticket gets scratched, or whether the cash box ever opens, is the same person who has now twice demonstrated he optimizes deal structure for the long game. Whether you find that reassuring or alarming probably depends on whether you think you’re in the boat with him or standing on the dock.

What I’ll be watching: the Schedule 14C information statement when it’s filed (it will detail the shell’s post-close cost structure); whether the board ever sets a deadline for returning capital absent a deal, the way ContextLogic did; and what the first acquisition target looks like. The Section 382 math says even $900 million of losses shelters only modest income per year, and the ContextLogic comps suggest realistic NOL value is a small fraction of the headline. The shell is worth its cash plus a James-shaped call option.

Silicon Slopes’ first great exit was an Orem company selling to Adobe for $1.8 billion. Its strangest second act might be Domo’s afterlife as a Utah County Berkshire wannabe. Either way, eighteen years in, Josh James still hasn’t given up the keys — and the structure guarantees nobody can take them.

Primary sources: Domo press release, July 22, 2026 · Progress Software announcement · Domo Q1 FY2027 10-Q coverage (forbearance, going concern) · Domo 424B4 IPO prospectus (SEC, 2018) · ContextLogic 8-K (SEC, 2024) · Adobe/Omniture tender offer materials (SEC, 2009). Nothing here is investment advice; it’s an autopsy with a sequel hook.

• On July 10, 2026, President Trump announced he “will not sign” the bipartisan 21st Century ROAD to Housing Act — but he also declined to veto it, so the bill became law automatically at 12:01 a.m. ET Saturday, July 11, under the Constitution’s 10-day rule. It passed 85–5 in the Senate and 358–32 in the House, with every “no” vote coming from Republicans.

• Refusing to sign, threatening vetoes, and vetoing broadly popular or bipartisan bills is a recurring feature of American history: FDR was overridden on the 1936 Bonus Bill, Truman on Taft-Hartley (1947), Nixon on the War Powers Resolution (1973), Reagan on the Anti-Apartheid Act (1986), Obama on JASTA (2016), and Trump on the 2021 NDAA — his only first-term override.

• The key civics facts: a president can sign a bill, veto it, or do nothing. If he does nothing for 10 days (Sundays excepted) while Congress is in session, it becomes law without his signature. If Congress has adjourned, inaction kills the bill — a “pocket veto.” Congress can override a regular veto only with a two-thirds vote in both chambers; pocket vetoes cannot be overridden.

Key Findings

Part 1 — The 2026 housing bill. Trump used a rarely-seen maneuver: rather than sign or veto, he publicly refused to sign the 21st Century ROAD to Housing Act “in PROTEST” that the Senate had not passed his priority voter-ID bill, the SAVE America Act. Because the housing bill passed with veto-proof, overwhelmingly bipartisan majorities and Trump did not formally veto it, it became law without his signature. The episode is politically striking because the bill targets housing affordability — a leading pocketbook issue heading into the midterms — and Trump’s own party had wanted a signing ceremony.

Part 2 — Historical precedent. Presidents refusing to sign or vetoing popular/bipartisan bills is common. The strongest override cases (Taft-Hartley, War Powers, Anti-Apartheid, JASTA, 2021 NDAA) all involved genuinely bipartisan supermajorities overriding the president. Vetoes and overrides have grown rarer in recent decades. FDR holds the all-time veto record (635); Andrew Johnson holds the record for overrides (15).

Details

PART 1: Trump and the 21st Century ROAD to Housing Act (July 2026)

The bill and its substance. The 21st Century ROAD to Housing Act is described by the Bipartisan Policy Center and multiple outlets as the most significant federal housing legislation since the Cranston-Gonzalez National Affordable Housing Act of 1990 (The New York Times noted it is the most significant home-construction legislation since 1990; Newsweek called it the “most significant federal housing legislation since the 1990s”). It is a supply-side package of 56 provisions (per the Niskanen Center, “a big bill, comprising 56 provisions designed to tackle housing barriers”), incorporating provisions from over 60 separate pieces of legislation — 36 of which were introduced with bipartisan sponsors, per the Bipartisan Policy Center. It aims primarily at boosting housing supply and lowering costs rather than creating a large new spending program. Key provisions include:

• Streamlining/expanding NEPA categorical exclusions and delegating environmental review to states, tribes, and localities to speed homebuilding;

• Making it easier and cheaper to build manufactured housing by removing the outdated permanent-chassis requirement — a change Sen. Elizabeth Warren said “brings down the cost of a new unit by up to $10,000” (the Niskanen Center estimated savings of up to 9% of the average price);

• A “Build Now”-style incentive steering more federal funds to localities that build more housing;

• Restricting large institutional investors that own at least 350 single-family homes from buying more (with rental-market exemptions), with civil penalties of $1,000,000 per violation or three times the purchase price of the property involved — whichever is greater (per Greenberg Traurig analysis). The provision grew out of Executive Order 14376, “Stopping Wall Street from Competing with Main Street Homebuyers,” which Trump signed in January 2026;

• Raising banks’ Public Welfare Investment cap from 15% to 20%; allowing CDBG funds for new affordable-housing construction; small-dollar mortgage support; and veterans’ housing measures;

• A conservative-sought provision barring the Federal Reserve from creating a central bank digital currency through 2030.

The bill was led in the Senate by Republican Tim Scott and Democrat Elizabeth Warren, and in the House by Republican French Hill and Democrat Maxine Waters.

The votes and how bipartisan they were. The Senate passed the final bill 85–5 on June 22, 2026; the House passed it 358–32 on June 23, 2026. All opposition came from Republicans: five GOP senators (Ron Johnson of Wisconsin, Mike Lee of Utah, Rand Paul of Kentucky, Rick Scott of Florida, and Tommy Tuberville of Alabama) and 32 House Republicans. Every Democrat in both chambers voted yes. Both margins far exceed the two-thirds needed to override a veto. (Note: one outlet, Spectrum News, reported the Senate vote as 89–10; the large majority of sources — CNBC, PBS, ABC, Time, Newsweek — report 85–5, with 10 senators not voting, which appears to be the correct final tally.)

Republican opponents objected on ideological grounds: that the bill expanded HUD, intervened in the free market by restricting investor home purchases, only temporarily (rather than permanently) banned a CBDC, and increased spending. Several also tied their “no” votes to Trump’s demand that the SAVE America Act pass first; Rep. Ralph Norman said he “joined 31 of my colleagues to STAND FIRM with President Trump in refusing to pass ANY SENATE BILLS until the SAVE America Act is passed.”

Trump’s stated reasons and the sequence of events. Republicans initially planned a high-profile Capitol signing ceremony for June 24, 2026, and the White House (via Press Secretary Karoline Leavitt) called it “one of the most significant pieces of housing affordability legislation in American history.” Less than two hours before the planned signing, Trump canceled it, posting that the ceremony was “canceled until such time as we pass the desperately needed SAVE AMERICA ACT, which I consider to be a National Emergency.” He dismissed the housing bill as “a yawn” and “so unimportant” relative to the elections bill.

The housing bill was formally presented to Trump on June 29, 2026, starting the 10-day clock (which began June 30). On July 10, 2026, Trump posted on Truth Social: “I will not sign the Housing Bill, which has been fully approved by Congress and sent to the White House, in PROTEST over the fact that the United States Senate is not capable of passing THE SAVE AMERICA ACT.” He also urged the Senate to “TERMINATE THE FILIBUSTER.”

The SAVE America Act (Safeguard American Voter Eligibility Act) would require documentary proof of citizenship to register and photo ID to vote in federal elections. It has passed the House multiple times but repeatedly failed in the Senate, where it needs 60 votes and Republicans hold 53 seats. (It is already illegal for noncitizens to vote in federal elections, and research finds it is extraordinarily rare.) That the housing bill was a top pocketbook priority is reflected in polling: Gallup’s May 2026 “Most Important Problem” survey found 15% of Americans named the high cost of living the nation’s biggest problem and 13% cited the economy, versus only about 2% naming elections/election reform (per Newsweek).

Procedural status and override. This was not a veto and not a classic pocket veto. Trump let the bill become law without his signature under Article I, Section 7: because Congress was in session and he neither signed nor vetoed within 10 days (Sundays excepted), the bill became law automatically at 12:01 a.m. ET Saturday, July 11, 2026. NBC News reported the White House confirmed he would not veto it. Because the bill became law, no override was necessary; had Trump vetoed it, the 85–5 and 358–32 margins were well above the two-thirds threshold, so an override was likely feasible (though override votes often lose some support).

Reactions split along predictable lines. Democrats seized on it as evidence Trump doesn’t prioritize affordability: Senate Democratic leader Chuck Schumer said Trump “called their crisis ‘a big yawn’ — then refused to sign the most significant bipartisan housing bill in decades,” adding “His priorities couldn’t be clearer: higher costs for families and more power for himself.” House Speaker Mike Johnson, who had urged Trump to sign with “the fattest black marker you have,” said “If he doesn’t, it’s still law. We’ll still celebrate it.”

Historical housing parallel. The closest historical housing-policy parallel is not a veto but Nixon’s January 1973 housing moratorium: Nixon (through HUD Secretary George Romney) imposed an 18-month freeze/impoundment on most federally subsidized housing programs (Sections 235 and 236, urban renewal, Model Cities), declaring the government had become the “biggest slumlord in history.” Congress was infuriated at the impoundment of appropriated funds; the moratorium was rescinded in summer 1974, and the Housing and Community Development Act of 1974 followed under President Ford.

PART 2: A Century of Vetoes and Refusals (1926–2026)

How the mechanics work. Article I, Section 7 gives the president 10 days (excluding Sundays) to act on a bill. He can (1) sign it; (2) return it unsigned with objections — a “regular veto,” which Congress can override with two-thirds of both chambers; (3) do nothing while Congress is in session, in which case it becomes law without his signature after 10 days; or (4) do nothing when Congress has adjourned and thereby prevented the bill’s return — a “pocket veto,” which cannot be overridden. As of the Senate’s official count, there have been about 2,599 vetoes in U.S. history, and Congress has overridden fewer than 5%.

Aggregate statistics (from Senate.gov and CRS). Veto totals per president:

• Franklin D. Roosevelt: 635 total (372 regular, 263 pocket), 9 overridden — the all-time record, aided by 12+ years in office.

• Grover Cleveland: 584 (mostly private pension bills), 7 overridden.

• Harry Truman: 250 (180 regular, 70 pocket), 12 overridden.

• Dwight Eisenhower: 181 (73 regular, 108 pocket), 2 overridden.

• Gerald Ford: 66 (48 regular, 18 pocket), 12 overridden — extraordinary given ~2.5 years in office.

• Ronald Reagan: 78, 9 overridden.

• George H.W. Bush: 44, 1 overridden.

• Bill Clinton: 37, 2 overridden.

• George W. Bush: 12, 4 overridden.

• Barack Obama: 12, 1 overridden.

• Donald Trump (first term): 10, 1 overridden.

• Joe Biden: 13, 0 overridden.

Andrew Johnson holds the record for overrides (15 of 29 vetoes). Truman and Ford are tied second at 12 each. Vetoes and overrides have become markedly rarer since the 1950s. Overrides are far more likely under divided government (one analysis found a 16.4% override rate when the opposition controlled both chambers vs. 0.5% under unified government).

FDR and the Bonus Bill (1935–1936). FDR twice opposed early cash payment of WWI veterans’ bonuses. In 1935 he delivered his veto of the Patman Greenback Bonus Bill in person to a joint session; the House overrode 322–98 but the Senate sustained the veto 54–40. In January 1936, Congress passed the Adjusted Compensation Payment Act; FDR vetoed it again, but this time both chambers overrode — the House 322–98 within an hour, and the Senate on January 27, 1936 — enacting roughly $1.5 billion (about 2% of GNP) in bonds to WWI veterans over his objection. This was a case of a broadly popular bill (with strong bipartisan and grassroots veterans’ support) enacted over a president’s fiscal-conservative veto. FDR’s veto message argued that “wealth is not created, nor is it more equitably distributed by this method.”

Truman and Taft-Hartley (1947). The Labor Management Relations (Taft-Hartley) Act restricted union power — banning closed shops, permitting state right-to-work laws, requiring anti-communist affidavits from union officers. Truman vetoed it on June 20, 1947, calling it “a dangerous intrusion on free speech” and “bad for labor, bad for management, and bad for the country.” The Republican-controlled 80th Congress overrode him with heavy Democratic support: the House 331–83 and the Senate 68–25. This is a canonical example of a bipartisan override — 106 of 177 House Democrats and 20 of 42 Senate Democrats joined Republicans.

Nixon: War Powers Resolution (1973) and the Child Development Act (1971).

• War Powers Resolution: Passed with bipartisan support (sponsored by Democrat Clement Zablocki, backed by Republican Jacob Javits), it limited the president’s ability to commit troops without congressional authorization. Nixon vetoed it October 24, 1973, calling it “both unconstitutional and dangerous to the best interests of our nation.” Congress overrode November 7, 1973 — the House 284–135 (four votes over two-thirds) and the Senate 75–18. It was the first override of the 93rd Congress.

• Comprehensive Child Development Act (1971): Co-sponsored by Democrat Walter Mondale and Democrat John Brademas, it would have created a national network of child-care centers and passed the Senate 63–17 and House 211–187. Nixon vetoed it December 9, 1971, with a message (drafted partly by Pat Buchanan) warning it would “commit the vast moral authority of the National Government to the side of communal approaches to child rearing over against the family-centered approach.” The veto was not overridden and effectively ended the last major push for universal childcare.

Ford’s divided-government vetoes (1974–1977). Facing a heavily Democratic Congress, Ford used the veto aggressively (66 vetoes) and was overridden 12 times — tied with Truman for the most in modern history despite only ~2.5 years in office. Key bipartisan overrides:

• Freedom of Information Act Amendments of 1974 (H.R. 12471): Strengthened FOIA by allowing judicial review of classified documents and setting agency response deadlines. Ford vetoed it October 17, 1974, writing that he remained “concerned that our military or intelligence secrets and diplomatic relations could be adversely affected by this bill” and that courts “should not be forced to make what amounts to the initial classification decision in sensitive and complex areas where they have no particular expertise.” (Declassified records show aides Donald Rumsfeld and Dick Cheney, and DOJ’s Antonin Scalia, urged the veto.) Congress overrode November 21, 1974: House 371–31, Senate 65–27, enacting Public Law 93-502.

• Second Labor-HEW Appropriations Act (H.R. 8069, FY1976): Ford vetoed it on fiscal-restraint grounds; Congress overrode in late January 1976 — House 310–113, Senate 70–24 (Public Law 94-206).

• Education Division appropriations (H.R. 5901, 1975): Ford vetoed a $7.9 billion education bill as $1.5 billion over his request, arguing “we must stop doing what need not be done.” The House overrode 379–41, enacting Public Law 94-94.

• Labor-HEW Appropriations Act, 1977 (H.R. 14232): Ford vetoed a ~$56.6 billion bill (~$4 billion over his request) as “the triumph of election-year politics over fiscal restraint,” even though he “agree[d] with the restriction on the use of Federal funds for abortion.” A Democratic Congress overrode it on September 30, 1976, weeks before the election, enacting Public Law 94-439.

Reagan and the Comprehensive Anti-Apartheid Act (1986). The Act imposed sanctions on South Africa. The Senate passed a compromise version 84–14; Reagan vetoed it September 26, 1986, calling it “economic warfare” that would mostly hurt South Africa’s impoverished Black majority. In a striking bipartisan rebuke, Congress overrode: the House 313–83 (with 81 Republicans joining 232 Democrats) and the Senate 78–21. Republican Senator Richard Lugar, chair of the Foreign Relations Committee, led the override effort against his own party’s president. It was the first override of a major foreign-policy veto since the War Powers Resolution.

George H.W. Bush and the Civil Rights Act of 1990. Bush vetoed the bill (which would have made it easier to win employment-discrimination suits), calling it a measure that “employs a maze of highly legalistic language to introduce the destructive force of quotas into our national employment system.” The bill had bipartisan support (GOP cosponsors including Danforth, Specter, and Jeffords). The Senate override attempt failed by a single vote — 66 senators, including 11 Republicans, voted to override, one short of two-thirds. A year later, Bush signed the very similar Civil Rights Act of 1991.

Clinton and welfare reform (1995–1996). A reverse example of using rejection as leverage: Clinton vetoed two Republican welfare-reform bills he considered too harsh before signing the third, the Personal Responsibility and Work Opportunity Reconciliation Act, in August 1996. He said the earlier bills “failed to protect our children,” and that the final version “had broad bipartisan support, and is much, much better on both counts.” Clinton’s veto record (37 total) is notable for having only one pocket veto.

George W. Bush: SCHIP and stem cells (2007). Bush vetoed a bipartisan expansion of the State Children’s Health Insurance Program (H.R. 976) in October 2007, arguing it would shift families from private to government coverage and cover children in families earning up to ~$83,000. The Senate had passed it with a veto-proof 67–29, but the House override attempt failed 273–156 (13 short). Bush also twice vetoed the Stem Cell Research Enhancement Act, saying it would “compel American taxpayers... to support the deliberate destruction of human embryos.” Both stem-cell vetoes stood.

Obama and JASTA (2016). The Justice Against Sponsors of Terrorism Act let 9/11 families sue foreign governments (i.e., Saudi Arabia). It passed both chambers without recorded opposition (unanimous consent in the Senate, voice vote in the House). Obama vetoed it September 23, 2016, warning it would erode sovereign immunity and expose U.S. personnel abroad to reciprocal lawsuits. Congress delivered the only override of Obama’s presidency by huge bipartisan margins: Senate 97–1 (only Harry Reid sustained; Kaine and Sanders did not vote) and House 348–77.

Trump’s first-term NDAA veto (2020–2021). Trump vetoed the FY2021 National Defense Authorization Act, objecting to provisions renaming Confederate-named bases and to the absence of a Section 230 repeal. Congress overrode him — the House 322–87 (Dec. 28, 2020) and the Senate 81–13 (Jan. 1, 2021) — the first and only override of Trump’s first term. Only seven GOP senators voted to sustain (Braun, Cotton, Cruz, Hawley, Kennedy, Lee, and Paul).

Biden’s vetoes (2021–2025). Biden issued 13 vetoes, none overridden. His first (March 2023) rejected a bipartisan resolution blocking a Labor Department ESG-investing rule; the House override attempt failed 219–200. Notably, Biden issued no vetoes while Democrats controlled Congress, and 12 of his 13 vetoes drew an override attempt (only one went unchallenged).

Pocket vetoes as a distinct tool. A pocket veto occurs when the president takes no action on a bill and Congress’s adjournment prevents its return; the bill dies and cannot be overridden. James Madison was the first to use it (1812). By the Senate’s official count, FDR used it most — 263 pocket vetoes (some tallies of his inaction cite higher figures, but 263 is the Senate’s number). Eisenhower had 108, Cleveland 238. Pocket vetoes have become contested because presidents and Congress dispute what “adjournment” means — courts have generally held a pocket veto is valid only when Congress has adjourned sine die (ended a session), not during short recesses. Clinton (twice) and George W. Bush attempted intra-/inter-session withholdings that Congress disputed and that are not counted as pocket vetoes. The 2026 Trump housing episode is NOT a pocket veto — Congress was in session, so the bill became law rather than dying.

Caveats

• The housing-bill vote count has a minor discrepancy: one outlet (Spectrum News) reported the Senate vote as 89–10, while the strong majority of reputable sources (CNBC, PBS, ABC News, Time, Newsweek, Bipartisan Policy Center) report 85–5 with 10 senators not voting. Use 85–5.

• This is a fast-moving story dated July 10–11, 2026; some framing (e.g., NBC’s “he does not plan to veto”) reflects White House statements at the deadline rather than a formal document. There was no veto message because Trump did not veto.

• A few historical vote tallies vary slightly by source (e.g., the Taft-Hartley Senate override appears as both 68–25 and 68–32; FDR’s 1936 Bonus override figures vary between accounts). Where sources conflict, cite the range or defer to the Senate’s official record.

• Some retrieved sources are lower-tier (e.g., Grokipedia, Quora, Brainly, partisan blogs); their specific facts were cross-checked against primary sources (Senate.gov, the American Presidency Project, House.gov, Congress.gov) or reputable outlets wherever used. Aggregate veto counts rely on the Senate Historical Office and CRS.

• Individual party-by-party breakdowns for some overrides (e.g., Ford’s) were not available at the member level in the sources gathered; every override necessarily drew bipartisan support because it cleared the two-thirds threshold in both chambers.

Within hours, the reaction on the right made it sound as though the Court had just legalized an invasion. The Department of Justice posted about cracking down on “birth tourism schemes.” Members of Congress amplified claims about a multi-million-dollar birth tourism industry. Cable segments and social feeds filled with the specter of pregnant foreigners flying in to mint American babies [3].

Here is the number that should sit next to every one of those segments: in 2024, the CDC recorded 9,576 births in the United States to mothers who listed a foreign residence. There were roughly 3.6 million total births that year. That’s about a quarter of one percent — and even that figure includes women who gave birth here for reasons that have nothing to do with acquiring citizenship for their child [3, 4].

This is not a one-off. It’s a pattern, and once you see it, you can’t unsee it: American conservatism has developed a remarkable capacity for taking phenomena that are statistically tiny — sometimes literally nonexistent — and converting them into dominant national narratives, legislative waves, and nine-figure ad campaigns. This piece is an attempt to document that pattern with numbers, and to ask honestly whether it is actually a conservative-specific phenomenon or just politics being politics.

I think the evidence says it’s asymmetric. Not because conservatives are uniquely credulous — they aren’t — but because the right has built political and media machinery that is unusually good at manufacturing bigness out of smallness. Let me show my work.

Case one: the trans athlete panic, by the numbers

Start with the anchor example, because it’s the one where the gap between prevalence and attention is most precisely documented.

The prevalence. In December 2024, NCAA President Charlie Baker — a former Republican governor, not a progressive activist — testified before the Senate Judiciary Committee. Asked how many transgender athletes were competing in college sports, he said he was aware of fewer than 10. Out of approximately 510,000 NCAA athletes. That is roughly 0.002 percent of college athletes [5].

The story is the same at the state level, and here Utah readers will recognize it. In March 2022, Governor Spencer Cox vetoed HB11, Utah’s transgender sports ban, and in his veto letter he did something politicians almost never do: he published the denominator. Of the roughly 75,000 kids playing high school sports in Utah, four were transgender. One was a transgender girl playing girls’ sports. “Rarely has so much fear and anger been directed at so few,” Cox wrote [6]. The legislature overrode his veto four days later — a statewide ban, aimed at one child.

West Virginia passed its ban in 2021 while the only publicly known transgender athlete in the state’s schools was a middle-schooler who wanted to run cross country [7]. Multiple states passed bans without being able to identify a single affected athlete in the state at all.

The attention. Now the other side of the ledger. The ACLU and the independent Trans Legislation Tracker counted over 500 anti-LGBTQ bills introduced in state legislatures in 2023, more than 600 anti-trans bills in 2024, and the count climbed again in 2025 — the largest coordinated legislative campaigns aimed at any minority group in modern state-house history [8].

Then came the money. In the closing months of the 2024 presidential campaign, Republicans spent an estimated $215 million on network television ads attacking Democrats on transgender issues, according to AdImpact data reported by the New York Times — headlined by the Trump campaign’s ubiquitous “Kamala is for they/them, President Trump is for you” spot, which aired relentlessly during NFL and college football broadcasts [9]. Nearly a quarter of a billion dollars, in one election cycle, on an issue that — in its most-litigated form, sports — involves a population the NCAA’s own president counts on two hands.

The perception gap. Does the machine work? Ask Americans how many of their fellow citizens are transgender. In a 2022 YouGov survey on population estimates, the average American guessed that about 21 percent of Americans are transgender. The actual figure, per UCLA’s Williams Institute, is roughly 0.5 percent of adults [10]. The public isn’t off by a rounding error. It’s off by a factor of forty — which is what you’d expect when attention is allocated by rhetorical utility rather than by prevalence.

Case two: birth tourism, by the numbers

Back to this week’s panic. “Birth tourism” — traveling to the U.S. while pregnant so your child is born a citizen — is real. It exists. Federal prosecutors have broken up commercial operations catering to Chinese and Russian clients, and nobody serious disputes that. The question, as always, is: how big is it?

The most aggressive estimate comes from the Center for Immigration Studies, a think tank that advocates for lower immigration. After the Niskanen Center’s Jeremy Neufeld identified an error in its original 33,000-per-year figure — an error CIS, to its credit, acknowledged — CIS revised its estimate to 20,000–26,000 possible birth-tourism births per year [11, 12]. Niskanen’s own analysis of the same data, after excluding long-term U.S. residents whom the CIS method sweeps in, lands at roughly 2,000–2,700 [12]. A 2026 Penn State research brief using CDC natality data from 2014–2024 estimates 5,000 to just under 10,000 births per year to non-resident mothers — about 0.3 percent of annual U.S. births [3].

So the honest range, spanning restrictionist and pro-immigration analysts alike, runs from about 0.06 percent of U.S. births to about 0.7 percent at the absolute ceiling — using the estimate produced by the advocacy group most motivated to find a big number. And this is the phenomenon that has been offered, this week, as a reason to unsettle a constitutional rule that has stood since Reconstruction, one the Supreme Court first affirmed 128 years ago [1, 2].

Note the structure of the argument, because it recurs: a real-but-marginal phenomenon (birth tourism) is used as the emotional justification for a policy whose actual scope is vastly larger (the executive order would have denied citizenship to an estimated 255,000 children born each year to noncitizen parents, per the Migration Policy Institute — twenty-five to fifty times the size of the “problem” cited to justify it) [13].

The pattern, cataloged

Two examples make a coincidence. Here are six more, each compressed to its essential ratio.

Critical race theory (2021–22). Christopher Rufo, the activist who engineered the panic, described the strategy openly on Twitter in March 2021: freeze the brand, then fold every unpopular cultural grievance into it so the public reflexively attributes anything “crazy” it encounters to CRT [14]. It worked. Media Matters counted nearly 1,900 mentions of “critical race theory” on Fox News in a single 3.5-month window in 2021 [15]. State legislatures introduced hundreds of restriction bills, tracked by UCLA Law’s CRT Forward project [16]. What was the underlying prevalence? CRT is a graduate-level legal framework; surveys of K-12 teachers found vanishingly few claiming to teach it, and panic-era school board meetings routinely failed to produce curricular examples when asked.

Litter boxes in classrooms (2022). The claim that schools were installing litter boxes for students who identify as cats was investigated by NBC News, which found at least 20 conservative candidates and officials — including gubernatorial nominees and a state party chair — repeating it publicly during the 2022 cycle. The number of verified cases was zero. It is a hoax with a prevalence of literally nothing, repeated by people running for the office of governor [17].

Sharia law bans (2010–2017). Between 2010 and 2016, state legislators introduced more than 200 anti-Sharia or “foreign law” bills across 43 states, with over a dozen enacted, per the Haas Institute’s Legalizing Othering report [18]. Muslims are about 1 percent of the U.S. population, and the number of instances in which Sharia has overridden U.S. constitutional law is zero — it can’t; the Supremacy Clause exists. Several of the states that passed bans have Muslim populations measured in fractions of a percent. Meanwhile, in the Ipsos “Perils of Perception” surveys, Americans guessed Muslims make up roughly 17 percent of the U.S. population — an overestimate of similar magnitude to the trans one [19].

Voter fraud. The Heritage Foundation maintains the most-cited database of proven voter fraud, assembled specifically to demonstrate the problem is real. Its total: roughly 1,500 proven instances — spanning decades and all fifty states, across billions of ballots cast. The Brennan Center’s studies put documented in-person impersonation fraud — the specific kind that voter ID laws address — at between 0.0003 and 0.0025 percent of votes [20]. On the strength of this, “election integrity” became the organizing grievance of an entire political era, culminating in a majority of Republican voters believing a presidential election was stolen despite over sixty failed court challenges.

The migrant caravan (2018). In the weeks before the 2018 midterms, a group of a few thousand Central American migrants — walking, on foot, over a thousand miles away — became the dominant story on conservative media, complete with troop deployments to the border. Media Matters’ segment counts showed the caravan mentioned dozens of times per day on Fox News in late October. Immediately after Election Day, coverage collapsed to near zero [21]. The caravan hadn’t arrived, disbanded, or changed. Only its electoral usefulness had expired — which is about as clean a natural experiment on motive as you will ever get.

Springfield, Ohio (2024). Roughly 12,000–15,000 Haitian immigrants — overwhelmingly in the country legally under Temporary Protected Status — had settled in Springfield to work. The claim that they were eating residents’ pets was investigated and rejected by Springfield’s own city officials, who reported no credible cases. It was repeated anyway by the Republican vice-presidential nominee and then by the presidential nominee on the debate stage in front of 67 million people. Bomb threats closed Springfield’s schools and municipal buildings in the days that followed [22]. Prevalence of verified pet-eating: zero. Reach: the largest broadcast audience of the election.

Six cases, plus the two anchors. In every one, the ratio of attention to prevalence isn’t 2-to-1 or 10-to-1. It’s three, four, five orders of magnitude — or division by zero.

Why the right? The machinery, not the psychology

Here’s where I want to be careful, because there’s a lazy version of this argument and a rigorous one.

The lazy version says conservatives are just wired for fear. There is a psychology literature associating conservatism with heightened threat sensitivity — John Jost’s influential “motivated social cognition” work is the touchstone [23]. But you should know that the most vivid strand of that literature, the physiological studies claiming conservatives startle harder at scary images, failed high-profile replication attempts in 2020 [24]. Karen Stenner’s work suggests something subtler: an “authoritarian dynamic” in which a subset of people across societies respond to perceived normative threat with demands for enforcement and conformity — a disposition that is activated by threat narratives rather than a constant trait [25]. That reframing matters, because it relocates the explanation from brains to supply: someone has to manufacture and distribute the threat narratives that flip the switch.

And that’s where the asymmetry actually lives. Two structural facts, each documented at book length:

First, the parties are different kinds of organizations. Matt Grossmann and David Hopkins’ Asymmetric Politics marshals decades of survey and organizational data showing the Democratic Party operates as a coalition of discrete interest groups making concrete policy demands, while the Republican Party operates as the vehicle of an ideological movement [26]. Movements run on symbolic threat. A coalition of teachers’ unions and construction trades needs deliverables; a movement needs enemies, and it needs them vivid more than it needs them large. A phenomenon involving ten athletes or 9,500 births is a poor input for interest-group bargaining but a perfect input for movement politics — small enough to be unfalsifiable in any voter’s direct experience, symbolic enough to stand in for civilizational decline.

Second, the media ecosystems are different shapes. This is the core finding of Yochai Benkler, Robert Faris, and Hal Roberts’ Network Propaganda, the largest empirical mapping of American political media ever done — millions of stories, link structures, and sharing patterns around the 2016 election. Their conclusion: the American media landscape isn’t two mirror-image bubbles. The center-left ecosystem is anchored to institutions that, whatever their biases, are disciplined by professional fact-checking norms and pay reputational costs for retracted stories. The right’s ecosystem is a closed, insulated loop anchored by Fox and Breitbart in which content is disciplined by audience demand rather than accuracy, and correction signals from outside the loop are treated as attacks that increase in-group trust [27]. In one ecosystem, a litter-box story eventually hits a fact-check firewall. In the other, the fact-check is itself proof of the conspiracy.

Add the demonstrated agenda-setting power of a single dominant outlet — economists Stefano DellaVigna and Ethan Kaplan, and later Gregory Martin and Ali Yurukoglu, have measured Fox News’ causal effect on Republican vote share using cable-channel-position natural experiments [28] — and you have the full machine: a movement party that runs on symbolic threat, fed by a closed media loop with no correction mechanism, amplified by an outlet with proven power to move millions of votes. Smallness in, bigness out.

The honest counterargument

If this thesis is right, it should survive contact with its best objections. There are three worth taking seriously.

“The left does it too.” Sometimes, yes. The Satanic ritual abuse panic of the 1980s ran through daytime television, feminist activists, and prosecutors of both parties, and put innocent people in prison over crimes that never occurred. GMO health fears and some strands of anti-nuclear sentiment are progressive-coded panics with prevalence-of-harm problems. Portions of the post-2016 “Russian bots” discourse attributed to foreign influence operations a causal power the empirical literature never supported. And political scientist Joseph Uscinski’s research finds conspiratorial thinking as a disposition roughly symmetric across the partisan spectrum [29]. Anyone who tells you liberals are immune to moral panic hasn’t read a history of the 1980s.

“Small numbers can still be legitimate issues.” Also true in principle. We regulate plenty of rare things — plane crashes, meltdowns, contaminated Tylenol — because severity, precedent, and principle matter alongside frequency. A conservative can argue in good faith that fairness in one girls’ race, or the integrity of one election, or the meaning of citizenship, is worth contesting regardless of headcount. Prevalence isn’t the whole of moral seriousness.

“Salience always outruns statistics — that’s just media.” Shark attacks, stranger abductions, crime coverage generally: attention has never tracked base rates, for anyone.

Here’s why I don’t think these objections defeat the thesis. The question was never whether the left is capable of panic or whether rare things can matter. The question is which side has built durable institutional machinery that converts micro-phenomena into legislative programs, nine-figure ad buys, and executive orders — repeatedly, on schedule, as its primary mode of politics. The Satanic panic had no party apparatus writing model bills in fifty statehouses. There is no left-wing equivalent of 200 anti-Sharia bills targeting a legal impossibility, or $215 million in ads about ten athletes, or a hoax about litter boxes repeated by gubernatorial nominees, or a caravan that dominated a national news network until the polls closed and then ceased to exist. And the “severity, not frequency” defense has to explain why the remedy so consistently dwarfs the stated problem — a 255,000-child citizenship order justified by, at most, 26,000 births.

The asymmetry isn’t in human nature. It’s in the plumbing.

The denominator test

I write a lot around here about a habit I’d call denominator journalism: never report a numerator without its denominator. Five hundred of something sounds like a lot until you learn the base is 510,000, or 3.6 million, or 158 million ballots.

So here’s a portable test for the next panic, whichever direction it comes from. When a political movement tells you something is an emergency, ask three questions. What’s the count? What’s it out of? And who benefits from you not asking the first two?

By that test, this week’s birth-tourism eruption is running true to form: a real phenomenon, a fraction of a percent, a reaction sized for an invasion. The Supreme Court just spent 128 years of precedent reaffirming that a child born here is one of us. The people angriest about that ruling would like you to believe the maternity wards are full of tourists. The CDC counted 9,576 — out of 3.6 million.

Keep the denominator. It’s the one thing the smallness machine can’t survive.

Sources

1. Supreme Court of the United States, Trump v. Barbara, No. 25-365 (decided June 30, 2026). https://www.supremecourt.gov/opinions/25pdf/25-365_4hdj.pdf

2. SCOTUSblog, “Supreme Court strikes down Trump’s order ending birthright citizenship” (June 30, 2026). https://www.scotusblog.com/2026/06/supreme-court-strikes-down-trumps-order-ending-birthright-citizenship/ ; see also NPR’s decision coverage (June 30, 2026). https://www.npr.org/2026/06/30/nx-s1-5839358/birthright-citizenship-decision-scotus-trump

3. PolitiFact, “Trump allies blame birthright citizenship for ‘birth tourism.’ How often does it happen?” (July 1, 2026) — includes the Penn State research brief (5,000–10,000 births/yr, ~0.3%) and DOJ/congressional reaction posts. https://www.politifact.com/article/2026/jul/01/birth-tourism-birthright-citizenship-Supreme-Court/

4. FactCheck.org, “What Do We Know About ‘Birth Tourism’?” (April 16, 2026) — CDC 2024 figure of 9,576 births to foreign residents; ~3.6M total U.S. births. https://www.factcheck.org/2026/04/what-do-we-know-about-birth-tourism/

5. Testimony of NCAA President Charlie Baker, U.S. Senate Judiciary Committee hearing on safety in women’s sports, December 2024 (widely covered; see AP and Reuters reports of the “less than 10” figure against ~510,000 NCAA athletes).

6. Gov. Spencer Cox, HB11 veto letter to Utah legislative leadership, March 22, 2022 (published at governor.utah.gov; the 75,000 / 4 / 1 figures and the “so much fear and anger” line appear in the letter’s text).

7. Coverage of B.P.J. v. West Virginia State Board of Education (filed 2021), the ACLU challenge brought on behalf of the middle-school runner who was the state’s only known affected athlete when HB 3293 passed.

8. ACLU, “Mapping Attacks on LGBTQ Rights in U.S. State Legislatures” (annual tracker); Trans Legislation Tracker (translegislation.com) — 500+ bills in 2023, 600+ in 2024. Verify current-year counts before publication; they update continuously.

9. New York Times, reporting on AdImpact data on 2024 Republican anti-trans advertising (~$215 million on network TV), November 2024; the Trump campaign’s “Kamala is for they/them” ad.

10. YouGov, “From millionaires to Muslims, small subgroups of the population seem much larger to many Americans” (2022) — average estimate that 21% of Americans are transgender; Williams Institute, UCLA School of Law, “How Many Adults and Youth Identify as Transgender in the United States?” (2022) — ~0.5% of adults, ~1.4% of teens.

11. Center for Immigration Studies, Steven Camarota, “A Revised Estimate of Birth Tourism” (March 13, 2020) — 20,000–26,000/yr, with acknowledgment of the error Niskanen identified. https://cis.org/Camarota/Revised-Estimate-Birth-Tourism

12. Niskanen Center, Jeremy Neufeld, “The Birth Tourism Bogeyman” and “A Look at CIS’s Revised Birth Tourism Estimate” (2020) — long-term-resident correction yielding ~2,000–2,700. https://www.niskanencenter.org/the-birth-tourism-bogeyman/ ; https://www.niskanencenter.org/birth-tourism-revisited/

13. Migration Policy Institute estimate of ~255,000 children per year affected by the executive order, as cited in ABC News’ pre-decision explainer (June 2026). https://abcnews.com/Politics/faq-birthright-citizenship-ahead-supreme-courts-ruling/story?id=134215675

14. Christopher Rufo, Twitter/X thread of March 15, 2021, describing the strategy of freezing and expanding the “critical race theory” brand (screenshotted and reported widely, incl. The New Yorker’s June 2021 profile by Benjamin Wallace-Wells).

15. Media Matters for America, counts of “critical race theory” mentions on Fox News (≈1,900 mentions in a 3.5-month window, reported June 2021).

16. CRT Forward Tracking Project, UCLA School of Law — database of anti-CRT measures introduced at state and local levels.

17. NBC News (Tyler Kingkade et al.), “How an urban myth about litter boxes in schools became a GOP talking point” (October 2022) — 20+ conservative candidates/officials repeating the claim; zero verified cases.

18. Haas Institute (now Othering & Belonging Institute), UC Berkeley, “Legalizing Othering: The United States of Islamophobia” (2017) — 200+ anti-Sharia bills in 43 states, 2010–2016; Pew Research Center estimates Muslims at ~1.1% of the U.S. population.

19. Ipsos, “Perils of Perception” survey series — U.S. respondents’ average estimate of the Muslim population share (~17%) vs. actual (~1%).

20. Heritage Foundation Election Fraud Database (≈1,500 proven cases across decades — check the live count; it grows slowly); Brennan Center for Justice, “The Truth About Voter Fraud” (Justin Levitt, 2007, and subsequent updates) — incident rates of 0.0003%–0.0025%.

21. Media Matters for America, Fox News caravan coverage tracking, October–November 2018 — pre-election saturation followed by a post-election collapse in mentions.

22. Reporting from Springfield, Ohio, September 2024: city officials’ statements of no credible reports (AP, Reuters), the September 10 debate audience (~67 million, Nielsen), and subsequent bomb threats closing schools and city buildings.

23. Jost, Glaser, Kruglanski & Sulloway, “Political Conservatism as Motivated Social Cognition,” Psychological Bulletin (2003).

24. Bakker, Schumacher, Gothreau & Arceneaux, “Conservatives and liberals have similar physiological responses to threats,” Nature Human Behaviour (2020) — failed replication of the Oxley et al. (2008) startle-response findings.

25. Karen Stenner, The Authoritarian Dynamic (Cambridge University Press, 2005).

26. Matt Grossmann & David A. Hopkins, Asymmetric Politics: Ideological Republicans and Group Interest Democrats (Oxford University Press, 2016).

27. Yochai Benkler, Robert Faris & Hal Roberts, Network Propaganda: Manipulation, Disinformation, and Radicalization in American Politics (Oxford University Press, 2018).

28. DellaVigna & Kaplan, “The Fox News Effect: Media Bias and Voting,” Quarterly Journal of Economics (2007); Martin & Yurukoglu, “Bias in Cable News: Persuasion and Polarization,” American Economic Review (2017).

29. Uscinski & Parent, American Conspiracy Theories (Oxford University Press, 2014), and subsequent Uscinski et al. survey work on the partisan symmetry of conspiratorial predispositions.

A note on sourcing: items 1–4 and 11–13 link to primary documents I pulled directly this week. Items 5–10 and 14–29 are cited by author/outlet/date from the research file; I’d verify the exact live URLs and current tracker counts (items 8 and 20 especially, since they’re running databases) before hitting publish.

• The June 2026 US-Iran “Islamabad Memorandum” is not a peace treaty but a 14-point Memorandum of Understanding — a non-binding political instrument that merely extends a 60-day ceasefire, reopens the Strait of Hormuz, and defers every hard question (above all Iran’s nuclear program) to talks that have already been postponed.

• It is the latest example of a consistent pattern: of the wars Trump claims to have “ended” — a number that has crept from “six” to “seven” to “eight” and beyond — almost none rest on a ratified treaty, and several conflicts (DRC-Rwanda, Thailand-Cambodia, Gaza, India-Pakistan) kept fighting or remained disputed after the handshake.

• Independent experts and the parties themselves repeatedly describe these arrangements as fragile pauses, premature declarations, or status-quo deals; the Iran MOU fits the mold so precisely that its own text commits both sides only to “maintain the current status quo.”

Key Findings

Donald Trump has spent his second term branding himself the “President of PEACE,” and the centerpiece of that brand is a running tally of wars he says he has personally ended. The number is not fixed. In an August 5, 2025 fact-check, PolitiFact examined his claim of stopping “six wars” and rated it “Mostly False,” concluding: “Trump had a hand in temporary ceasefires in a few conflicts. But there’s little evidence he permanently resolved them and, in some cases, little evidence of U.S. intervention.” By his September 23, 2025 address to the UN General Assembly it was “seven unendable wars.” By October 9, 2025, with the Gaza ceasefire, he told reporters Gaza was “number eight”: “Nobody in history has solved eight wars in a period of nine months. And I’ve stopped eight wars. That’s never happened before,” adding he hoped Ukraine-Russia would be “number nine” (Axios, October 10, 2025). By a January 2026 text message to Norway’s prime minister he claimed “8 Wars PLUS,” and told Fox’s Sean Hannity he should count Thailand-Cambodia twice (”eight and a quarter”). The State Department’s official account simply posted: “8 wars ended in 8 months.” PolitiFact + 2

The conflicts the White House cites are consistent: Israel-Iran, Rwanda-DRC, Armenia-Azerbaijan, Thailand-Cambodia, India-Pakistan, Egypt-Ethiopia, Serbia-Kosovo, and (latterly) Israel-Hamas. What none of them produced is a ratified peace treaty. The instruments range from ceasefires and joint declarations to a memorandum of understanding and, in one case, a deal experts say addressed a war that wasn’t happening.

Details

The treaty/MOU distinction matters. Under the law of treaties, a treaty creates binding obligations enforceable under international law and typically requires ratification; a memorandum of understanding embodies a political commitment that is, in the words of one government explainer, “only politically or morally binding.” The US State Department’s own guidance notes that negotiators of non-binding documents should “avoid using the terms ‘treaty’ or ‘agreement.’” A treaty is the durable instrument; the MOU is the handshake written down. Trump’s signature achievements are overwhelmingly the latter. Bureau of International Information Programs

Iran: the newest “deal,” the oldest pattern. What Trump loosely brands a peace with Iran is the “Islamabad Memorandum,” a 14-point MOU signed June 17, 2026 — Trump signing on paper at a Versailles dinner with Macron after the G7, Iranian President Masoud Pezeshkian signing remotely in Tehran. It followed the 2026 Iran war (Operation Epic Fury), which began February 28, 2026 with US-Israeli strikes and cost the lives of 13 US service members per CENTCOM. The MOU does not resolve the nuclear question. Its Article 9 commits Iran only to “maintain the current status quo of its nuclear program” during a 60-day negotiating window, while the US pledges no new sanctions. It calls for reopening the Strait of Hormuz, lifting the US naval blockade, sanctions relief, and a $300 billion regional reconstruction fund for Iran — though Trump later disputed the $300 billion figure as “fake news.” WikipediaABC News

Crucially, it is explicitly an interim framework, not a settlement. The Council on Foreign Relations called it a memorandum that “leaves many challenging issues unresolved,” with CFR senior fellow Steven Cook warning on June 17, 2026: “We have been here before only to discover the parties cannot bridge the remaining gaps. Negotiations on the outstanding issues, especially on Iran’s nuclear program, will be long and difficult.” Carnegie’s James Acton wrote that “by burying disagreements in imprecision, the new deal risks same fate as its predecessors,” noting Iran’s pledge not to build a weapon is “the same promise it has made for years,” including in the 2015 JCPOA. That contrast is stark: the JCPOA ran to 159 pages of verification detail negotiated by six governments and nuclear physicists; Trump’s framework was negotiated bilaterally by Steve Witkoff and Jared Kushner and runs to one page. Ali Vaez, director of the International Crisis Group’s Iran project, captured the logic in Newsweek (June 15, 2026): “The immediate objective for both Washington and Tehran is to stop the bleeding without healing the wound… The MoU can help them stabilize the situation and buy time,” adding that “what gives this moment momentum is not trust, but fear.” Carnegie Endowment for International Peace

Iran, tellingly, has signaled it does not consider itself fully bound. Tehran released a statement (June 14, 2026, per CFR) saying “final negotiations will be postponed until after the implementation of the other party’s commitments under the memorandum.” Its Foreign Ministry went further, warning that “if part of the other side’s commitments is not implemented, the entire memorandum of understanding will be jeopardized” (CBS News, June 20, 2026). Within days, the formal Geneva signing ceremony was postponed amid renewed fighting in Lebanon, and CENTCOM reported a Kuwaiti interception of an Iranian missile it called an “egregious ceasefire violation.” Even the marquee win — reopening Hormuz — is, as several analysts noted, simply a return to the pre-war status quo for a waterway that was open before the war began. Antony Blinken put it sharply: “The only ‘achievement’ of the ceasefire is the likely reopening of the Strait of Hormuz — which was open before the war started.” ABC News

This is the same shape as the June 2025 Israel-Iran ceasefire that ended the original Twelve-Day War — a US-brokered truce Trump announced on Truth Social, immediately violated by both sides (”I’m not happy,” he said on live television), that held only until war resumed in February 2026. A ceasefire that collapses into a larger war eight months later is precisely the kind of impermanence the treaty/MOU distinction is meant to flag.

The other “ended” wars. The pattern repeats across the catalog:

• India-Pakistan: A May 10, 2025 ceasefire after a four-day conflict. But India categorically denies US mediation. Foreign Secretary Vikram Misri said on June 18, 2025: “PM Modi told President Trump clearly that during this period, there was no talk at any stage on subjects like India-U.S. trade deal or US mediation between India and Pakistan… Talks for ceasing military action happened directly between India and Pakistan through existing military channels.” CFR’s Josh Kurlantzick said flatly, “There is no real ending to this conflict.”

• Rwanda-DRC: A June 2025 framework, formalized in the December 4, 2025 “Washington Peace Accords.” Fighting flared within hours of the ceremony; the Rwanda-backed M23 — which is not party to the deal — kept advancing. Amnesty International’s regional director for East and Southern Africa, Tigere Chagutah, said: “Months of discussions and the signing of multiple agreements in Washington and Doha have had no tangible impact on the lives of Congolese civilians.” Brookings’ Michael O’Hanlon called it “a premature declaration of success, when in fact we are just getting to the starting line.” WikipediaCBS News

• Thailand-Cambodia: A July 2025 ceasefire, formalized in October in Kuala Lumpur, that broke down into renewed fighting in December 2025 (Thai F-16 airstrikes, dozens dead). By February 2026 Cambodia’s PM said Thai forces still “occupy deep into Cambodian territory.” PBSAl Jazeera

• Armenia-Azerbaijan: The August 8, 2025 White House event produced a joint declaration and an initialed — not signed or ratified — 17-point “Agreement on the Establishment of Peace and Interstate Relations,” plus the “Trump Route for International Peace and Prosperity” (TRIPP), a US-developed transit corridor over which Washington holds a 99-year mandate. As CBS noted, “the deal is not a formal peace agreement and requires ratification.” The Dutch think tank Clingendael headlined its analysis “Peace or Pause?”, warning that “the durability of US engagement is the only — but highly uncertain — factor in ensuring implementation.” WikipediaClingendael

• Serbia-Kosovo: PolitiFact found “there is little evidence a potential war was brewing.” Serbian President Vučić denied any war plans; the two have not been in active conflict since 1999. Anchorage Daily News

• Egypt-Ethiopia: No agreement exists at all; the Grand Ethiopian Renaissance Dam dispute remains unresolved, and Ethiopia accused Trump of escalating it.

• Gaza: The October 2025 ceasefire (Trump’s “number eight”) achieved a real and significant hostage-prisoner exchange and aid surge — but by June 2026 the Gaza Health Ministry reported over 1,000 Palestinians killed during the “ceasefire,” Hamas remained armed and undisarmed, Israel controlled well over half the territory, and the UN’s own monitor, Nikolay Mladenov, warned the stalemate risked “a dangerous status quo.” NPR

The counterpoint, fairly stated. Trump does deserve genuine credit in several cases, and a fair accounting must say so. Brookings’ Michael O’Hanlon and others acknowledge the administration pushed the Armenia-Azerbaijan deal forward by hosting both leaders; experts credit US pressure (via tariff threats) for the initial Thailand-Cambodia and India-Pakistan de-escalations and the Israel-Iran ceasefires. Stanford political scientist Ken Schultz offered the balanced verdict (PolitiFact, September 23, 2025): “Brokering an agreement is a first and important step in ending wars, but it is also just the start of a process that needs follow-through.” The Gaza hostage release was a real humanitarian achievement. The honest critique is not that Trump achieved nothing — it is that stopping the shooting is not the same as ending a war, and a ceasefire or MOU is not a treaty. PolitiFact

The Nobel subtext. The peace-deal tally is inseparable from Trump’s open campaign for the Nobel Peace Prize. He was passed over in October 2025 for Venezuela’s María Corina Machado; in a January 18, 2026 text he warned Norway’s PM Jonas Gahr Støre that, having been denied the prize “for having stopped 8 Wars PLUS,” he “no longer feel[s] an obligation to think purely of Peace.” Machado then handed him her medal in the Oval Office — a gesture the Nobel Foundation said could not transfer the honor itself, since a prize “can therefore not, even symbolically, be passed on.” The incentive structure rewards announcements, not durable outcomes: a signing ceremony delivers a headline; ratification and implementation take years. CNN

Recommendations

For readers and analysts assessing these claims, a staged framework:

1. Ask what was signed. A ratified treaty registered with the UN is durable; a “declaration,” “framework,” or “MOU” is a political commitment that can evaporate. The Iran deal is explicitly an MOU — treat it as a pause, not a peace.

2. Ask whether the fighting stopped — and stayed stopped. Apply a 6–12 month test. DRC-Rwanda, Thailand-Cambodia, and the 2025 Israel-Iran ceasefire all failed it. Watch whether the Iran ceasefire survives the 60-day window and the Lebanon flashpoint.

3. Ask whether the parties agree on what happened. When one side (India) denies US involvement, or the document defers every hard issue, the “deal” is thinner than the announcement.

4. Benchmarks that would change the assessment: Iran verifiably down-blending its ~440 kg of 60%-enriched uranium under IAEA supervision; a final, ratified US-Iran agreement within the 60-day window; Hormuz reopening permanently and toll-free beyond the initial period; the Armenia-Azerbaijan text being signed and ratified rather than merely initialed. Absent these, the skeptical read holds.

Caveats

This analysis relies on reporting and primary documents through late June 2026; the Iran situation is fluid and the 60-day negotiating window had not closed at the time of writing. Some claims rest on contested casualty figures (Gaza Health Ministry data) or anonymous officials describing an unreleased MOU text. Trump’s defenders make a legitimate point that de-escalation has real value even without a treaty, and that several conflicts are genuinely quieter than before. The argument here is narrower and, on the evidence, sound: the Iran arrangement is not a peace treaty, it does not resolve the underlying conflict, and it is “more of the same” — another impermanent, non-binding, status-quo deal being marketed as a war-ending triumph.

• In early June 2026, the Pentagon’s newly slimmed-down list of “religious affiliation codes” — cut from 211 categories to 31 — labeled 21 denominations as “Christian” but listed The Church of Jesus Christ of Latter-day Saints separately, outside that heading; after a furious, largely Republican backlash led by Utah Sens. Mike Lee and John Curtis, the Department of Defense revised the list on Monday, June 8, removing the “Christian” label from every group so no denomination is singled out.

• The codes are an administrative tool — used to help chaplains plan religious support, and historically tied to dog tags, casualty care, and burial rites — not a theological ruling; the Pentagon itself said its “job is not to adjudicate theological debates,” and at least one LDS veteran argued a distinct LDS code actually helps Latter-day Saints get LDS-specific chaplains and care.

• The flap sits atop a genuine, centuries-old dispute: Latter-day Saints insist they are unequivocally Christian (the faith is literally named for Jesus Christ), while many creedal, evangelical, and mainline churches classify the LDS Church as a separate, “restorationist” tradition outside Nicene Christianity — a distinction reflected for decades in how Pew, ARDA, and other classifiers handle Mormonism.

Key Findings

What happened. On May 20, 2026, Under Secretary of Defense for Personnel and Readiness Anthony Tata signed a memorandum, at the direction of Defense Secretary Pete Hegseth, slashing the military’s list of recognized “religious affiliation codes” from 211 to 31. Military.com first reported the memo on June 4. When the 31-code list circulated, it grouped 21 denominations under “Christian” labels (Baptist, Catholic, Methodist, Lutheran, even Jehovah’s Witnesses), but assigned The Church of Jesus Christ of Latter-day Saints its own standalone code — “CJ” — outside the Christian grouping. Utah’s congressional delegation erupted over the weekend of June 6–7. By Monday, June 8, the Pentagon posted a revised list that stripped the “Christian” descriptor from all groups, listing denominations individually by name, and called the original a “mistake” that “has been fixed.”

It was administrative, not doctrinal — but the optics were terrible. Pentagon spokesman Sean Parnell insisted the consolidation was “not designed to make any claims on the legitimacy of any faith or religious belief, nor is it intended to provide a list of ‘officially approved’ religions.” But because the leaked list explicitly tagged a couple dozen groups “Christian” and conspicuously left the Church of Jesus Christ off that list, the framing looked like a government verdict on a theological question — exactly the kind of judgment the First Amendment is supposed to keep the state out of.

The codes do real-world work. Religious-preference data feeds chaplain planning, appears (optionally) on dog tags, and informs last rites, casualty notification, and burial honors. But chaplains serve everyone regardless of code, and the data is voluntary.

Mormons mostly don’t want to be lumped under generic “Christian” anyway. The deeper irony: a distinct LDS code is arguably better for Latter-day Saint service members than being folded into a generic Protestant/Christian bucket, because distinctiveness is what justifies LDS-specific chaplains, worship groups, and correct rites.

Details

1. The news: a weekend firestorm over a leaked list

The trigger was a Hegseth-driven overhaul of the military chaplaincy. In March 2026, Hegseth announced he would shrink the faith-code system, calling the old list “impractical and unusable.” In his words: “The previous system had ballooned to well over 200 faith codes. … It was impractical and unusable, and many codes were never used at all” — he said roughly 82% of religiously affiliated service members use only six of the codes, and that some codes were used by no one. The reduction was formalized in Tata’s May 20 memo, which said the change would “streamline the DoW collection of religious preferences collection for service members to enhance the delivery of targeted religious support from the Chaplaincy” and gave the department 60 days to implement.

Military.com broke the story on June 4. The published 31-code list included: Agnostic (AN), Baha’i (BH), Buddhism (BU), a long run of “Christian –” entries (Assemblies of God, Baptist, Brethren, Catholic, Church of Christ, Church of God, Church of the Nazarene, Episcopal/Anglican, Evangelical, Jehovah’s Witnesses, Lutheran, Methodist, Non-Denominational, Orthodox, Other, Pentecostal, Presbyterian, Quaker, Reformed, Scientist, Seventh Day Adventist), then “Church of Jesus Christ of Latter Day Saints (CJ)” as its own line, followed by Hindu, Islam, Judaism, No Religion, Other Religions, and Sikh. Roughly 180 belief groups were dropped entirely — including atheists, humanists, Wiccans, pagans, Druids, deists, and Unitarian Universalists.

The Latter-day Saint reaction was immediate. Sen. Mike Lee (R-Utah) posted a screenshot asking, “Can anyone tell me why The Church of Jesus Christ of Latter-day Saints was left out of the list of Christian churches?” and added, “If only we, as Latter-day Saints, belonged to a church that had ‘Jesus Christ’ in its name and His image in its logo … Oh wait.” In a video he called the move “repugnant,” saying the government “needs to not weigh in on doctrinal disputes between various religious denominations,” and declared, “Secretary Hegseth—tear down that wall!” Sen. John Curtis (R-Utah) called it “unacceptable for a government entity to characterize a faith in a manner that contradicts the religion’s own foundational tenets.” Rep. Mike Kennedy (R-Utah) said the decision “is wrong and needs to be corrected.” Democratic Rep. Greg Stanton said LDS service members “deserve to know why Hegseth excluded their faith.”

The resolution came fast. Lee said he spoke directly with President Trump on Monday: “I just got off the phone with President Trump. We discussed the Pentagon’s ‘Christian list.’… We’re most fortunate that President Trump (1) loves Latter-day Saints, and (2) is our commander in chief.” Hours later, a Department of Defense account posted a revised list and a statement: “In order to clarify the work of chaplains, and simplify the work of commanders, the Pentagon has consolidated and simplified the list to roughly thirty codes — using the previously used labels for faiths. The Pentagon’s job is not to adjudicate theological debates, but to ensure sincerely-held faith is respected and encouraged in our ranks.” The Pentagon distanced itself from the May 20 document, calling it “leaked.” Lee responded: “I agree with this statement, and am grateful to [Secretary] Hegseth for correcting the error.”

A note on tone and context that any fair account should include: critics see the episode as of a piece with Hegseth’s broader, overtly Christian reshaping of the Pentagon. He has hosted Christian prayer services at the Pentagon, welcomed Christian-nationalist pastor Doug Wilson (who has flatly said “Mormonism is not Christian”), fired Army Chief of Chaplains Maj. Gen. William Green Jr. in April 2026, and ordered chaplains to wear religious rather than rank insignia. The Military Religious Freedom Foundation’s Mikey Weinstein called the 31-code list a “middle finger to the United States Constitution’s separation of church and state.” Supporters frame it as a long-overdue cleanup of a bloated system. Both things can be true: the underlying consolidation may be defensible, while the specific decision to tag some groups “Christian” and not others was a self-inflicted wound.

2. What the codes actually are

The formal object at the center of this story is the DoD’s Faith and Belief Code system (the 2026 memo also calls them “religious affiliation codes”). These are short alphanumeric codes attached to a service member’s personnel record and propagated through the Defense Enrollment Eligibility Reporting System (DEERS) and the Defense Manpower Data Center (DMDC), the Pentagon’s central demographics warehouse. The data also appears in casualty-processing systems and personnel-reporting manuals.

The system has a long lineage:

• WWI–WWII: Religious preference on ID tags was reduced to single letters — P (Protestant), C (Catholic), H (Hebrew/Jewish) — with no preference left blank.

• Cold War expansion: Army Regulation 606-5 added “X” (a religion outside the big three, e.g., Islam or Eastern Orthodoxy) and “Y” (no religion or no listing).

• Vietnam era (c. November 1962): The military moved from single letters to spelling out preferences, and began recording specific denominations like “Methodist” or “Baptist” — the point at which a distinct, spelled-out Latter-day Saint preference became possible. Before this, LDS members were generally classified under “Protestant.”

• March 27, 2017: The list was formally expanded to 211 faith and belief codes. The expansion was recommended by the Armed Forces Chaplains Board and issued via a memorandum signed by Lernes J. Hebert in his capacity as Acting Deputy Assistant Secretary of Defense for Military Personnel Policy (a point worth getting right: some accounts have mislabeled him “Secretary of Defense”). That memo renamed the “Faith Group Code” the “Faith and Belief Code,” and said the expansion would provide “more accurate demographic data for religious groups” and better planning for chaplain support. It was the 2017 list that added Wiccans, pagans, Druids, and dozens of other small groups.

• May 20, 2026: Tata’s memo cut 211 codes down to 31.

(One sourcing caveat: a small number of secondary sources put the 2017 count at 221 rather than 211, but 211 is by far the better-supported figure, cited consistently by Military.com, Task & Purpose, Religion News Service, and others.)

Crucially, the religion marked on a service member’s dog tags is not limited to the code list. Tata’s memo explicitly said members “will not be limited to the list of ‘religious affiliation codes’ when selecting information for their dog tags.”

3. What the codes are for — and the real-world stakes

There are two basic purposes. First, last rites and burial in the event of death: the preference printed on an ID tag (or held in the record) signals what religious rites a fallen service member should receive, and informs casualty notification and funeral honors. Army doctrine (ATP 1-05.02) directs that “any requirement for specific rites and sacraments are guided by the deceased Soldier’s faith group.” Second, demographic planning for chaplain support: knowing how many adherents of a given faith are in a unit theoretically helps the military allocate chaplains, facilities, and worship materials.

Both uses are voluntary, and several points blunt the claim that being mis-coded causes concrete harm to the living:

• Chaplains serve everyone. The bedrock principle of the chaplaincy is “perform or provide” — a chaplain either personally performs the religious support a member needs or arranges for someone who can. As one Army command chaplain put it in the casualty-notification context, “Chaplains are not there to exercise their faith tradition.”

• Accommodations don’t run through the code. Religious accommodation — dietary needs, grooming/uniform waivers (beards, head coverings), Sabbath observance, etc. — is governed by DoD Instruction 1300.17, “Religious Liberty in the Military Services” (effective September 1, 2020), which implements the Religious Freedom Restoration Act and the First Amendment’s Free Exercise Clause. That 2020 instruction was a paradigm shift: its predecessor was titled “Accommodation of Religious Practices,” and the rename to “Religious Liberty” signaled a move from grudging accommodation to a default-yes posture. It states DoD “will normally accommodate practices of a Service member based on a sincerely held religious belief,” and protects members from adverse action for expressing belief. The key legal test is “sincerely held belief” — not which box a member checked.

• Headstone emblems are a separate system. Burial markers in national cemeteries draw on the VA’s much longer list of “emblems of belief” — more than 80 approved symbols — not the DoD code list. (Weinstein’s pointed quip: “if you’re dead, you’ll get your emblem. But if you’re alive, you can’t even get it on your dog tags.”)

So the practical harm of being coded one way versus another is, for most purposes, modest — which is part of why the controversy is better understood as symbolic than operational. But symbols matter, especially when a government appears to be ratifying a contested theological judgment.

4. How military chaplaincy actually works

The U.S. military chaplaincy is built on a church-state bridge called ecclesiastical endorsement. A chaplain is a commissioned officer, but to serve — and to keep serving — they must be endorsed by a recognized religious organization that vouches they are qualified to represent that faith. Lose the endorsement, and separation from the military begins almost immediately. The requirement for such an endorsement has existed since 1901. Endorsement, governed by DoD Instruction 1304.28 (”The Appointment and Service of Chaplains”), keeps the government out of the business of deciding who is a bona fide minister.

The Armed Forces Chaplains Board (AFCB) — composed of the chiefs and deputy chiefs of chaplains of the Army, Navy, and Air Force — advises the Under Secretary of Defense for Personnel and Readiness and maintains the official list of endorsing agencies. The military recognizes a large and diverse set of endorsers: estimates run to over 200 groups granted endorsing status and well over 100 active endorsing agencies, representing the full range of American religion. There are more than 3,000 chaplains serving across the U.S. armed forces. Critically, every endorsed chaplain must agree to serve a religiously diverse force and coordinate with chaplains of other traditions — a Catholic priest, a rabbi, an imam, and an LDS chaplain are all expected to “provide for” members outside their own faith.

The LDS Church is its own endorser, and has been deeply embedded in the chaplaincy for over a century. The Church’s Military Relations and Chaplain Services Division (part of its Priesthood and Family Department) provides centralized endorsement for Latter-day Saint chaplains. Per Church News, “the Church has endorsed more than 300 military and civilian chaplains as of 2022 and wants to expand opportunities for more to be endorsed.” The LDS chaplaincy lineage is old and storied: Elias S. Kimball was the first commissioned LDS chaplain, during the Spanish-American War (1898); three LDS chaplains — Calvin S. Smith, B.H. Roberts, and Herbert B. Maw — served in World War I; and 45 Latter-day Saint chaplains served in World War II, at battles from Iwo Jima to the Bulge, two earning Silver Stars. WWI was the first time the U.S. military allowed the Church to directly select active-duty LDS chaplains, a practice that continues. In July 2022, Jenna Carson became the first female chaplain endorsed by the Church, commissioned in the Air Force. (Notably, where the LDS Church has too few chaplains to cover its scattered members, it relies on “Distinctive Religious Group Leaders” / lay leaders — enlisted members authorized to conduct LDS sacrament services where no chaplain is available.)

The relationship between the code a member selects and the chaplain who serves them is loose by design: a member coded “CJ” is most likely to be connected to an LDS chaplain or lay leader where one exists, but any chaplain is obligated to facilitate their free exercise. The code helps the system find and resource members of a faith; it does not gate access to care.

5. The “are Mormons Christian?” question, fairly stated

This is the genuinely contested core, and it predates the Pentagon by two centuries.

The LDS position: unequivocally yes. The Church’s own FAQ states: “The Church of Jesus Christ of Latter-day Saints is a Christian church but is neither Catholic nor Protestant. Rather, it is a restoration of the Church of Jesus Christ as originally established by the Savior in the New Testament of the Bible.” Members affirm faith in Jesus Christ as the literal Son of God and Savior; the faith is named for Him. Pew Research’s “Mormons in America” study (surveyed Oct. 25–Nov. 16, 2011; n=1,019) found Mormons “nearly unanimous in describing Mormonism as a Christian religion, with 97% expressing this point of view,” and “fully 98% say they believe in the resurrection of Jesus.” This self-understanding has been sharpened in recent years: in his October 2018 General Conference address “The Correct Name of the Church,” President Russell M. Nelson directed members and media to drop the nicknames “Mormon” and “LDS Church” in favor of the full name, saying: “When it comes to nicknames of the Church, such as the ‘LDS Church,’ the ‘Mormon Church’ … the most important thing in those names is the absence of the Savior’s name. To remove the Lord’s name from the Lord’s Church is a major victory for Satan.” Historian Benjamin Park reads Nelson’s campaign as reflecting “this long, this deep-seated anxiety that Mormons be accepted as Christians.” The June 2026 reaction from Lee, Curtis, and Kennedy flows directly from that emphasis.

The creedal/evangelical position: not in the historic sense. Many traditional Christian bodies classify the LDS Church as outside historic, creedal Christianity — chiefly because Latter-day Saint theology rejects the Nicene Creed and its formulation of the Trinity (one God in three co-equal, co-eternal, consubstantial persons). LDS doctrine teaches that the Father, Son, and Holy Ghost are three distinct “personages,” that humans can progress toward godhood, and that a “Great Apostasy” corrupted the early church — viewing the creeds themselves as a symptom of that apostasy. Joseph Smith’s First Vision account has God describe the creeds as “an abomination.” From the creedal side, that places the LDS Church in a different category. The Presbyterian Church (USA) describes it as “a new and emerging religious tradition distinct from the historic apostolic tradition”; the United Methodists wrote in 2000 that the Church “by self-definition, does not fit within the bounds of the historic, apostolic tradition of Christian faith.” A LifeWay Research survey of 1,000 Protestant pastors (conducted Oct. 7–14, 2010) found 75% disagreed that they personally consider Mormons to be Christians (60% strongly, 15% somewhat); notably, two-thirds (67%) of evangelical pastors strongly disagreed that Mormons are Christians, compared with 48% of mainline pastors. Catholic, Orthodox, and most evangelical bodies do not recognize LDS baptism. (Generous voices exist on this side too — Fuller Seminary’s Richard Mouw has written that Latter-day Saints exhibit “qualities that are worthy of the Christian label.”)

It is also worth noting the public is divided but leans toward inclusion: in a November 2011 Pew survey, 32% of non-Mormon U.S. adults said Mormonism is not a Christian religion, with another 17% unsure — meaning roughly half did regard it as Christian.

How classification systems handle it. Scholarly and statistical taxonomies have long treated Mormonism as its own category — not as an insult, but as taxonomy. Pew’s Religious Landscape Study sorts “restorationist” denominations into evangelical and mainline traditions but handles Mormon denominations separately. Reference works routinely describe the LDS Church as “Restorationist” and “Nontrinitarian” — the same bucket that often holds Jehovah’s Witnesses. This is the crucial context for the Pentagon flap: listing the Church separately is, in much of the academic and demographic world, an ordinary classificatory choice. What made the June 2026 list inflammatory was not separateness per se but the contrast — explicitly labeling 21 other groups “Christian” while pointedly omitting a church named for Jesus Christ created the appearance of a government ruling that LDS is not Christian.

6. Timeline of LDS recognition in the military religious apparatus

• 1898: Elias S. Kimball becomes the first commissioned LDS military chaplain (Spanish-American War); first LDS servicemen’s worship group organized.

• 1917–18 (WWI): Three LDS chaplains (Calvin S. Smith, B.H. Roberts, Herbert B. Maw) serve; first time the U.S. military lets the Church directly select active-duty chaplains. LDS members coded under “Protestant.”

• 1941–45 (WWII): 45 LDS chaplains serve in all theaters; the General Servicemen’s Committee (1941, chaired by Elder Harold B. Lee) builds the lay group-leader system; dog tags still limited to P/C/H.

• c. November 1962: Military shifts to spelling out denominations, making a distinct Latter-day Saint preference recordable.

• 2017 (March 27): Faith and belief code list expanded to 211, with LDS as a distinct entry among many.

• July 2022: Jenna Carson becomes the first LDS-endorsed female military chaplain.

• December 2024: The UK Ministry of Defence approves the LDS Church as an endorsing authority for the British Armed Forces.

• May 20, 2026: Tata memo cuts codes to 31; LDS (”CJ”) listed outside the “Christian” grouping.

• June 4, 2026: Military.com reports the memo.

• June 6–7, 2026: Utah delegation revolts.

• June 8, 2026: Pentagon revises the list, removing “Christian” labels from all groups.

7. Impact statement: grievance, misunderstanding, or both?

The most clarifying fact in this whole episode came not from a senator but from a Latter-day Saint veteran. State Sen. Dan McCay amplified commentary from Dan Shaha, who argued that being listed separately may actually benefit LDS service members: when the Church is folded into a broad Protestant/Christian category, “an LDS Chaplain and a Protestant Chaplain become interchangeable” for manpower planning — meaning the Pentagon could decide Protestant chaplains are sufficient to meet LDS members’ needs. By listing the Church on its own, the DoD could be obligated to ensure a minimum number of LDS chaplains, “potentially creating more opportunities for LDS chaplains and increasing their presence.” Jenna Carson, who served as an active-duty LDS Air Force chaplain until 2025, told the Salt Lake Tribune she had never previously had any issues regarding her Christianity in the military.

That points to the central tension. On the merits of how a religious-classification system should work, a distinct code for the Church of Jesus Christ of Latter-day Saints is not a slight — it is exactly what guarantees that LDS members get LDS chaplains, LDS sacrament services, and correct rites. Most Latter-day Saints, following President Nelson, fiercely guard their distinctiveness; they would not actually want to be dissolved into a generic “Christian” or “Protestant” label. In that sense, the grievance is partly a misunderstanding of what a taxonomy is for.

And yet the grievance is also real, because the government didn’t merely list the Church separately — it published a list that affirmatively certified other groups as “Christian” and left this one off. That is the move a secular state should never make. The instinct of Lee, Curtis, and Kennedy — that “the government should not be in the business of deciding who counts as Christian,” in Kennedy’s words — is the correct constitutional principle, and it cuts against Hegseth’s critics and defenders alike. The fix the Pentagon ultimately adopted — listing every denomination by its own name, under no theological umbrella — is the right answer precisely because it lets the state resource religious practice without grading anyone’s orthodoxy.

The honest verdict: the controversy was a genuine grievance triggered by a clumsy, arguably ideologically tinged taxonomy, resolved by retreating to the neutral ground the government should have occupied from the start. The distinct LDS code itself was never the problem — and may even be an asset. The problem was the “Christian” label stamped on everyone else.

Recommendations

For a reader or writer trying to assess this story and what to watch next:

1. Treat the resolved list as the substantive outcome, not the rhetoric. The operative fact is that the final 30-code list groups no one under “Christian” and names denominations individually. If you’re evaluating whether real harm occurred, the answer is: a symbolic injury that was corrected within days, with no documented loss of accommodations, chaplain access, or burial rites in the interim.

2. Watch whether the consolidation itself gets litigated or reversed. The sharper unresolved issue is the dropping of ~180 faith groups (atheists, pagans, Wiccans, Unitarian Universalists, humanists, Druids), which affects minority-faith members and potentially the endorsements of chaplains serving them. Retired Air Force Chief of Chaplains Steve Schaick warned that if a chaplain’s endorsing body loses recognition, “the chaplain no longer has legitimacy in the military.” That is the place where concrete harm — not just symbolism — could materialize. Benchmarks to watch: whether any endorsing agency loses AFCB recognition, whether any chaplain faces separation, and whether MRFF or others file suit.

3. Keep the theological and the administrative questions separate. The defensible journalistic and civic position is: (a) whether the LDS Church is “Christian” is a contested theological question on which the U.S. government has no business ruling; (b) whether the DoD should track Latter-day Saints under a distinct code is an administrative question whose answer is “yes, and that’s good for LDS members.” Conflating the two is what produced the firestorm.

4. If you cite figures, use the well-supported ones: 211 codes (2017) reduced to 31 (2026), later trimmed to ~30; ~180 groups dropped; 21 groups originally tagged “Christian”; more than 3,000 military chaplains; 300+ LDS-endorsed chaplains (as of 2022); first LDS chaplain 1898; 45 LDS chaplains in WWII. Flag the 211-vs-221 discrepancy if precision matters.

Caveats

• The May 20 memo’s status is contested. The Pentagon initially defended it through spokesman Parnell, then on June 8 called it a “leaked” “proposed” list containing a “mistake.” Whether the “Christian”-labeling was a deliberate policy choice or a drafting artifact is genuinely unclear from the public record; this piece treats the labeling as real (it was published and defended) while noting the Pentagon’s later disavowal.

• Motive is interpretive. Critics tie the list to Hegseth’s broader Christian-nationalist-tinged agenda; the Pentagon insists it was pure administrative streamlining. Both readings are represented above; readers should weigh them. Statements about intent (e.g., that the change targets minority faiths) are characterizations by named critics, not established fact.

• The exact origin date of a distinct LDS DoD code is not documented in available primary sources. The progression (Protestant → spelled-out denomination post-1962 → discrete 2017 code) is well-supported; a specific first-assignment date or regulation number for an LDS-specific code could not be verified and should not be asserted.

• The 211 vs. 221 count for the 2017 expansion has a minor discrepancy across sources; 211 is the dominant, better-corroborated figure.

• Numbers like “more than 3,000 chaplains” and “over 200 endorsing groups” are drawn from standard reference and DoD-adjacent sources and fluctuate year to year; treat them as approximate orders of magnitude.

Sources / Bibliography

The June 2026 news and primary statements

• Military.com, “DOD Officially Drops 180 Faiths From Military’s Recognized Religion List” (June 4, 2026) — first report, full 31-code list, Tata May 20 memo.

• Military.com, “DOD’s New Official Recognized Religions List Draws Strong LDS Rebuke” — Parnell statement, June 8 revision, “CJ” code.

• Washington Post, “After Mormon lawmakers object, Pentagon revises Christian religious categories” (June 8, 2026).

• CBS News / WTOP, “Pentagon updates religious codes after criticism from Mormons.”

• NOTUS, “The Pentagon Changed Its Religious Classifications After Outrage From LDS Lawmakers” — “mistake … fixed,” Lee–Trump call, Tata background.

• Deseret News, “Utah senators question Pentagon designation for Latter-day Saints” (June 6, 2026) — Lee/Curtis posts, LDS FAQ text.

• KSL.com / KSLTV.com, “Lee, Curtis defend Church of Jesus Christ as ‘Christian’…” and “Pentagon updates religious categories…” — Kennedy statement, Parnell quotes.

• KUTV, “Pentagon revises religious classification list after Utah lawmakers push for change” — Dan Shaha / chaplain-manning argument.

• The Hill, “Pentagon, Pete Hegseth rework ‘offensive’ policy affecting LDS…”; Newsweek; Salon; ABC4; FOX 13 (Salt Lake); NewsNation.

• Religion News Service, “Defense Department to drop atheists, pagans, 175 others…” (June 4, 2026) — Schaick and Mehta quotes.

• Task & Purpose, “Pentagon cuts 180 faiths from recognized religion list”; The Spokesman-Review (USA Today wire), “Hegseth directs DOD to drop hundreds of faiths…” — MRFF/Weinstein, 2017 AFCB context.

• Yahoo/The Independent, “LDS groups and Utah Republicans rage…” — Jenna Carson and Eric Biggart comments.

The codes, chaplaincy, and DoD policy (primary/government)

• DoD Instruction 1300.17, “Religious Liberty in the Military Services” (eff. Sept. 1, 2020), esd.whs.mil.

• DoD Instruction 1304.28, “The Appointment and Service of Chaplains,” esd.whs.mil; DD Form 2088 (”Statement of Ecclesiastical Endorsement”).

• March 27, 2017 memorandum, “Faith and Belief Codes for Reporting Personnel Data of Service Members,” signed Lernes J. Hebert, Acting DASD (Military Personnel Policy) (hosted at militaryatheists.org/regs).

• U.S. Army, “What’s on your dog tag?” (army.mil); Wikipedia, “Dog tag” and “Religious symbolism in the United States military” (WWII letters, AR 606-5 X/Y, Nov. 1962 shift).

• Army ATP 1-05.02, “Religious Support to Funerals and Memorial Ceremonies”; army.mil, “Casualty notification process emphasizes dignity, respect.”

• Armed Forces Chaplains Board endorser materials (prhome.defense.gov / war.gov); army.mil, “Armed Forces Chaplains Board Military Endorser Conference”; tjaglcs.army.mil endorser list; CPSP, “The Challenges of Endorsers and Endorsements” (150+ federally recognized endorsing bodies); Wikipedia, “Military chaplain” / “United States military chaplains” (3,000+ chaplains, endorsement since 1901).

LDS Church, chaplaincy history, and “are Mormons Christian”

• ChurchofJesusChrist.org, “Latter-day Saint Chaplains” and “Resources & Information for Military Members”; Newsroom, “Church of Jesus Christ Supports Military Families.”

• Church News / Deseret News / LDS Living, “Making history: Church endorses first female military chaplain” (Jenna Carson, July 7, 2022); Church News, “How the Church’s Military Advisory Committee…” (300+ chaplains as of 2022); MormonWiki, “Jenna Carson.”

• Newsroom, “How a Diverse Trio of Latter-Day Saint Chaplains From World War I Remain Relevant”; ChurchofJesusChrist.org, “Servicemember Branches”; Encyclopedia of Mormonism, “Military and the Church”; Wikipedia, “LDS Church members in 20th-century warfare”; Richard Maher, “For God and Country: Mormon Chaplains During World War II” (BYU, 1975).

• Thechurchnews.com (Dec. 2024), UK Ministry of Defence approves LDS endorsing authority.

• Russell M. Nelson, “The Correct Name of the Church” (Oct. 2018 General Conference); Wikipedia, “Name of the Church of Jesus Christ of Latter-day Saints”; The Conversation / RNS, Konden Smith Hansen on Nelson and “Mormon”; KUER, Benjamin Park comments.

• Pew Research Center, “Mormons in America” (Jan. 2012; surveyed Oct.–Nov. 2011); LifeWay Research (Oct. 2010 pastors survey, via Christian Century / Mormonism and Nicene Christianity coverage); Wikipedia, “Mormonism and Nicene Christianity” and “Restorationism”; FAIR (fairlatterdaysaints.org); Presbyterian Church (USA) and United Methodist (2000) statements as quoted in Christian Century, “Are Mormons Christian? It’s complicated”; Pew Religious Landscape Study restorationist categories.

As of late May 2026, Iran’s elected head of government is President Masoud Pezeshkian, a reformist heart surgeon in office since July 2024.[^5] But real power has shifted sharply toward the Islamic Revolutionary Guard Corps (IRGC) and its hardline commander Ahmad Vahidi, in the aftermath of the February 28, 2026 assassination of Supreme Leader Ali Khamenei in a joint US-Israeli air campaign,[^39][^41] and the wartime elevation of Khamenei’s son Mojtaba Khamenei as the new Supreme Leader.[^10][^13]

The viral headline — “Iran’s president offers resignation, citing total takeover by IRGC commanders” — traces to a single, anonymously sourced Iran International report dated May 31, 2026,[^6] which the Iranian government denied within hours, including via IRGC-affiliated media.[^7]

The fault line on war versus peace is unusually clear: Pezeshkian and the reformist camp have pushed for de-escalation;[^55] the IRGC hardliners around Vahidi favor continued confrontation;[^31] and the Supreme Leader, who in theory must approve any deal, is reportedly wounded and in seclusion.[^12][^20]

1. The President: Masoud Pezeshkian

Who he is. Masoud Pezeshkian (b. September 29, 1954), a former heart surgeon and health minister, has been Iran’s ninth president since July 28, 2024.[^5] He is a reformist who won a snap election held after President Ebrahim Raisi died in a May 2024 helicopter crash, defeating hardliner Saeed Jalili in the runoff.[^5] He campaigned on reviving the nuclear deal, easing sanctions, and modest social reform.

Head of state or head of government? In Iran’s system the president is the head of government — the chief executive who runs the cabinet and day-to-day administration — while the Supreme Leader is the head of state and paramount authority.[^1][^3] The Council on Foreign Relations notes Iran’s constitution designates the Supreme Leader’s office as head of state, with authority grounded in the idea that political authority flows from religious authority.[^3] The president, in this structure, is second-in-command of the executive branch and, as one widely used explainer puts it, answers to the Supreme Leader and carries out his decrees.[^1]

Constitutional powers and limits. Under the 1979 constitution, the president is elected by direct popular vote to a four-year term and is eligible for a second consecutive term (Articles 114–115).[^2] Article 113 makes the president responsible for implementing the constitution and heading the executive, except in matters directly tied to the office of the Leadership.[^2] In practice this means the president:

• Proposes the budget, runs national administration, and nominates cabinet ministers (subject to parliamentary confirmation).[^2][^3]

• Signs treaties and legislation and conducts diplomacy — but within the policy framework the Supreme Leader sets.[^1][^4]

What the president cannot do is decisive: he is not commander of the armed forces, does not control the IRGC, and cannot set foreign, defense, or nuclear policy independently. The Supreme Leader sets national policy and commands the military; through the Guardian Council he vets who may even run for office; and he has reportedly influenced the selection of the defense, intelligence, and foreign ministers.[^3] Every presidential candidate must first be approved by the Guardian Council, half of whose members the Supreme Leader appoints.[^1][^3]

His relationship to the Supreme Leader. The president sits at the top of the executive branch but below the Supreme Leader, who can confirm and — after impeachment or a court ruling — effectively remove him.[^3] After Ali Khamenei’s death, Pezeshkian briefly served on a three-member Interim Leadership Council (with the chief justice and a Guardian Council cleric) that executed the Supreme Leader’s duties until Mojtaba Khamenei was selected.[^5][^13]

The resignation headline — verified, but contested

The story behind the headline is real and accurately summarized. On Sunday, May 31, 2026, the London-based, anti-regime outlet Iran International reported — citing “a source familiar with the matter” — that Pezeshkian had submitted a letter of resignation to the Office of the Supreme Leader.[^6] Per the report, the letter said the president and government had been effectively excluded from major decision-making, that the resulting vacuum let hardline IRGC factions take control, and that he could therefore no longer carry out his legal responsibilities.[^6] Iran International framed it against months of friction between the government and the security establishment, and said it was not yet clear whether Mojtaba Khamenei would accept the resignation.[^6]

Two important caveats:

1. It rests on a single anonymous source.[^6] The Times of Israel and others relayed it explicitly as an unconfirmed report attributed to that outlet.[^8][^9]

2. The Iranian government denied it within hours. A presidential communications deputy, Mehdi Tabatabaei, called the report false and a “media game,” and — notably — the IRGC-affiliated Tasnim News Agency also denied it, saying Pezeshkian had not resigned and was still carrying out his duties.[^7] As of this writing, no outlet has reported the resignation being formally accepted, rejected, or withdrawn; the official position is that no resignation occurred.[^7] Analysts have pointed out that IRGC-aligned media denying the story complicates the “the Guards forced him out” reading.[^7]

The episode is, however, consistent with a documented pattern of public friction between Pezeshkian and the IRGC over the conduct of the war and its economic fallout.[^6]

2. The Supreme Leader: Mojtaba Khamenei (since March 2026)

The succession. The longtime Supreme Leader, Ayatollah Ali Khamenei (in power since 1989), was killed on February 28, 2026 in the opening salvo of the US-Israeli air campaign.[^39][^41] His son Mojtaba Khamenei (b. September 8, 1969), a 56-year-old mid-ranking cleric with deep IRGC ties, was selected by the Assembly of Experts and announced as the third Supreme Leader on March 8–9, 2026 — the post’s first father-to-son succession.[^10][^13][^14] Britannica notes that although Mojtaba had made almost no public appearances beforehand, he was well positioned through his network inside his father’s office to consolidate control quickly during wartime.[^10]

How a Supreme Leader is chosen. Under the constitution (Articles 107–111), the Assembly of Experts — an 88-member body of clerics, themselves popularly elected but Guardian-Council-vetted — selects, supervises, and can in theory dismiss the Supreme Leader.[^4][^11] Mojtaba’s selection was extraordinary: it happened during active war, the threshold for victory was a two-thirds majority (59 of 88),[^13] and three Iranian officials told The New York Times (via Middle East Monitor and AP) that the Revolutionary Guards had pushed hard for his appointment.[^15][^14] AP described him as a secretive cleric with close ties to the paramilitary Guard.[^14]

Powers of the Supreme Leader (Article 110). The Supreme Leader is the highest political and religious authority. Drawing on Articles 57 and 110, he:[^3][^4][^11][^18]

• Sets the general domestic and foreign policies of the state and supervises all three branches.[^1][^3]

• Is commander-in-chief; declares war and peace; appoints and dismisses the senior military commanders and the heads of the IRGC and the regular army.[^3][^18]

• Appoints the head of the judiciary, half the Guardian Council, and the head of state broadcasting (IRIB).[^3][^18]

• Confirms (and can move to remove) the elected president, and ratifies decisions of the Supreme National Security Council.[^3]

Why he outranks the president. The Supreme Leader controls the levers that bound the elected branches: the Guardian Council vets all candidates, the judiciary he appoints can prosecute opponents, and the armed forces answer to him rather than to the president.[^1][^3] The president administers; the Leader sets the limits.

Velayat-e Faqih. The system rests on the doctrine of velayat-e faqih (”Guardianship of the Islamic Jurist”), developed by Ayatollah Ruhollah Khomeini and built into the 1979 constitution. Under Twelver Shiism, it holds that in the absence of the hidden Twelfth Imam, a qualified jurist should govern.[^1][^17] After Khomeini’s 1989 death, the constitution was amended to an “absolute” version of the doctrine, expanding the Leader’s authority across the state and dropping the requirement that he be a grand ayatollah — which allowed the then mid-ranking Khamenei to take the post.[^1][^17] Analysts describe the effect as placing the Supreme Leader, in practice, above the law and all other institutions.[^17][^48-note]

A figurehead in hiding. Mojtaba Khamenei was reportedly wounded in the February 28 strike that killed his father, and he has been largely absent from public view since his selection, fueling uncertainty over his fate and capacity to govern.[^12][^20][^21] US and Israeli officials have asserted he was seriously injured; the Jerusalem Post reported that, given his lack of public appearances, doubts have been raised about his ability to rule.[^45] These intelligence-sourced claims cannot be independently verified.[^21] He is also widely described as more hardline than his father.[^12][^10]

3. The IRGC (Sepah / Pasdaran)

What it is. The Islamic Revolutionary Guard Corps was founded by Khomeini after the 1979 revolution to protect the revolution from internal and external enemies — a regime-protection force rather than a conventional territorial military.[^22][^24] It owes loyalty directly to the Supreme Leader, not the elected government, and has grown into a vast military-industrial complex with its own ground, naval, and aerospace forces (the latter running Iran’s missile program), plus intelligence and cyber units and the Basij militia.[^24][^22]

IRGC vs. Artesh. The regular army (Artesh) is the pre-revolutionary, professional, comparatively apolitical military, with little role in internal politics and far fewer economic assets.[^24][^23] The two forces are deliberately kept separate and even rivalrous — a coup-proofing design reflected in unequal access to funding, recruitment, and the political leadership.[^23] The IRGC has historically received significantly more resources than the Artesh.[^28]

The Quds Force. The Quds Force is the IRGC’s extraterritorial arm, supporting allied militias across the region (the “Axis of Resistance”) and reporting up through the IRGC to the Supreme Leader.[^24][^22] It was long led by Qasem Soleimani until his 2020 killing in a US drone strike.

Command and reporting. The IRGC commander-in-chief is appointed by, and reports to, the Supreme Leader.[^3] Since March 1, 2026, that is Ahmad Vahidi — a hardliner and former Quds Force commander, under US and EU sanctions and subject to an Interpol notice over the 1994 AMIA bombing in Buenos Aires.[^31] He rose after a brutal churn at the top: commander Hossein Salami was killed in June 2025, and his successor Mohammad Pakpour was killed on February 28, 2026.[^32]

Economic empire. The IRGC controls a sprawling business empire — anchored by Khatam al-Anbiya Construction Headquarters (a.k.a. Ghorb), which builds refineries, dams, pipelines, and railways and runs Tehran’s main international airport — plus banks and foundations (bonyads).[^25][^26][^27] The US Treasury has long sanctioned Khatam al-Anbiya as IRGC-controlled and warned the Guard uses such firms to mask its economic reach.[^27] Estimates of the IRGC’s share of Iran’s economy vary widely — commonly cited at roughly one-third to two-thirds of GDP — precisely because the holdings are so opaque; one analysis pegs just Khatam al-Anbiya’s contract value at around 14% of GDP.[^29]

Political ascendancy after the strikes. This is the heart of the present crisis. With Ali Khamenei dead, much of the senior military leadership killed, and Mojtaba wounded and hidden, IRGC commander Vahidi has emerged as one of Iran’s most powerful figures.[^31] Fox News, citing analysts, reported that the key question is no longer whether Iran wants a ceasefire but whether Vahidi — described as a radical even within the regime’s hardline elite — judges that continued confrontation better serves his interests.[^31] That is the institutional backdrop to the May 31 resignation report.[^6]

4. The US strikes: a factual timeline

Background. After the US withdrew from the 2015 nuclear deal in 2018 and reimposed “maximum-pressure” sanctions, tensions escalated through the 2023–25 regional crisis. There were two distinct waves of US strikes:

Wave one — Operation Midnight Hammer (June 21–22, 2025). During the Israel-Iran “Twelve-Day War,” the US struck three Iranian nuclear sites. Per the Pentagon briefing by Joint Chiefs Chairman Gen. Dan Caine: seven B-2 stealth bombers flew from Missouri and dropped 14 GBU-57 30,000-pound “bunker-buster” bombs on Fordow and Natanz — the weapon’s first operational use — while a guided-missile submarine launched more than two dozen Tomahawk cruise missiles at Isfahan.[^33][^35] Roughly 75 precision-guided munitions and about 125 aircraft were involved.[^34][^37] A ceasefire followed days later. The Iranian parliament voted to close the Strait of Hormuz, leaving the final decision to the Supreme National Security Council.[^68-note]

Wave two — Operation Epic Fury (began February 28, 2026). After indirect US-Iran nuclear talks stalled and were cancelled amid Israeli strikes,[^38] the US (code name “Epic Fury”) and Israel (”Roaring Lion”) launched a massive joint surprise campaign — by Britannica’s account, nearly 900 strikes in 12 hours targeting Iranian missiles, air defenses, military infrastructure, and leadership.[^39] The opening wave killed Supreme Leader Ali Khamenei and dozens of senior officials,[^39][^41] including IRGC chief Pakpour and the defense minister.[^42] A US House Republican Policy Committee memo confirms the operation’s name and that the strikes killed Khamenei along with scores of senior officials.[^43] Trump described it publicly as major combat operations aimed at regime change.[^42]

Iran’s response and escalation:

• Iran launched waves of ballistic missiles and drones at Israel, US bases, and Gulf states, and closed the Strait of Hormuz.[^40] Iranian forces declared the strait “closed” from early March, attacking or threatening transiting ships.[^51]

• The closure triggered a global energy shock. Per the US EIA, the strait carried about 20 million barrels per day in early 2025 — roughly one-fifth of global oil consumption.[^47] The EIA later estimated Gulf producers shut in 7.5 million b/d of crude in March 2026, rising to 9.1 million b/d in April; Brent crude averaged $103/barrel in March and was forecast to peak around $115.[^46] The Dallas Fed and the UK House of Commons Library both treat the closure as a major supply disruption.[^50][^48]

• On March 17, 2026, Israel assassinated Ali Larijani, secretary of the Supreme National Security Council and — per multiple reports — the de facto leader running Iran in the weeks after Khamenei’s death; he was replaced by a hardliner.[^44][^45]

• Mojtaba Khamenei was named Supreme Leader on March 8–9.[^14]

Casualties and damage. Figures are preliminary and disputed. The opening wave alone reportedly killed about 170 people when a missile struck a girls’ school near a naval base at Minab.[^39] By mid-to-late March, the Hengaw human-rights group estimated 5,300+ Iranian military personnel killed, while Iran International reported at least 4,700 security forces dead.[^40] Casualties also extended to Lebanon, Israel, and the Gulf states, with millions displaced regionally.[^39]

Consequences for Iran’s leadership. The war produced (a) a dynastic, IRGC-backed succession to a wounded, hidden Supreme Leader;[^12][^14] (b) the physical elimination of moderating figures such as Larijani and the empowerment of the hardest-line IRGC faction;[^44][^31] and (c) the practical sidelining of the elected president — the grievance at the center of the resignation report.[^6]

5. Peace and diplomacy: who wanted a deal?

The factional map on war versus peace is comparatively clear in current reporting.

Pro-diplomacy / de-escalation:

• President Pezeshkian and the reformist camp. Trump himself publicly characterized Iran’s “new regime president” as less radicalized and as having asked for a ceasefire.[^55] Pezeshkian’s de-escalation signals have repeatedly been undercut from within the system.[^55][^6]

• Pragmatic negotiators, including figures who led the Pakistan-mediated talks, sought a negotiated off-ramp but could not commit Iran without the IRGC’s and Leader’s sign-off.[^53]

Pro-confrontation / hardline:

• The IRGC under Vahidi. Analysts warn Vahidi belongs to the hardline core that prefers to keep fighting and may view any ceasefire merely as a chance to regroup.[^31] One commentator quoted by Fox News flatly distrusts him.[^31]

• Hardline factions more broadly have pushed escalation and rejected US terms.[^56]

The Supreme Leader. Mojtaba Khamenei is described as more hardline than his father,[^12][^10] but his reported incapacity and seclusion mean he can barely function as a day-to-day decision-maker, which is exactly how the IRGC clique with access to him has expanded its control.[^20][^45]

Where the talks stand (late May 2026). A fragile ceasefire mediated by Pakistan took hold on April 8, 2026, after Iran rejected an earlier framework and floated its own 10-point plan; both sides have since violated it, and the first round of “Islamabad Talks” (April 11–12) ended without a signed agreement, snagging on the Strait of Hormuz and the nuclear file.[^52][^53] Trump subsequently extended the truce.[^52] By late May, reporting (via Axios) indicated the two sides had reached a tentative 60-day memorandum of understanding to extend the ceasefire, reopen Hormuz, and open a window to negotiate the nuclear program — but with no final sign-off from either Trump or Iran’s leadership, and with Iran disputing any US role over the strait.[^54] The UK House of Commons Library tracks the parallel, unresolved diplomatic effort to reopen Hormuz.[^48][^49]

How to read all this

1. Treat the resignation report as significant but unconfirmed. It is real reporting from a credible-but-partisan outlet, single-sourced, and officially denied — including by IRGC-aligned media.[^6][^7] Don’t state as fact that Pezeshkian has resigned. What would change the picture: an on-camera statement by Pezeshkian, a formal acceptance or rejection by the Leader’s office, or corroboration by a second independent, named source.

2. Watch the IRGC, not the presidency, as the locus of real power. The operative question for any deal is whether Vahidi judges it serves the Guard’s interests.[^31] Concrete tells: whether Hormuz actually reopens,[^48] whether IRGC media endorse or undercut the MOU,[^54] and whether Pezeshkian regains control of appointments.[^6]

3. Use the Supreme Leader’s visibility as a barometer. A verified live public appearance by Mojtaba Khamenei would signal he is reconsolidating personal authority; continued absence signals rule-by-proxy through the IRGC.[^20][^45]

4. On the nuclear question, focus on the uranium stockpile and inspections, which the tentative MOU mostly defers to later talks.[^54]

5. Discount triumphalist framing from all sides — US “obliteration”/”regime change” claims, Iranian “victory” claims, and opposition “collapse” narratives are all interested; weigh them against primary assessments.[^39][^46]

Caveats on sourcing

• Fast-moving, contested situation. This reflects what was reportable as of May 31, 2026; officeholders and the ceasefire status can change quickly.

• Source lean matters. Iran International and JNS are openly anti-regime;[^6][^20] Tasnim/Fars (cited in the denials) are state- or IRGC-aligned;[^7] Fox News and the Times of Israel bring their own editorial lean to Iran coverage.[^31][^8] The steadiest anchors here are the wire services (AP),[^14] the encyclopedic and academic sources (Britannica, NYU GlobalEx, CFR),[^39][^2][^3] and government/parliamentary primary documents (the Iranian constitution itself, the US EIA, the US Treasury, the UK House of Commons Library, a US House Policy Committee memo, the Dallas Fed, and CRS).[^4][^46][^27][^48][^43][^50][^51]

• The “head of state” ambiguity is genuine, rooted in a mismatch between the constitution (which assigns some head-of-state functions to the president) and the political reality of Supreme Leader paramountcy.[^2][^3]

• Claims about Iranian leaders’ deaths, injuries, and whereabouts frequently trace back to unnamed US or Israeli intelligence officials and cannot be independently verified.[^21][^45]

• Casualty and IRGC economic-share figures are estimates that vary substantially across sources.[^40][^29]

References

[^1]: The Conversation, “Iran’s ruling structure explained” (Apr. 21, 2026). https://theconversation.com/irans-ruling-structure-explained-277913 — academic/explainer [^2]: NYU Law GlobalEx, “The Legal System and Research of the Islamic Republic of Iran” (constitutional articles incl. 113–115). https://www.nyulawglobal.org/globalex/iran1.html — academic primary-law reference [^3]: Council on Foreign Relations, “The Islamic Republic of Iran’s Power Centers” (Article 110 powers; ministers). https://www.cfr.org/articles/islamic-republics-power-centers — think tank [^4]: Office of the Supreme Leader (Leader.ir), “Leadership in the Constitution of the Islamic Republic of Iran” (Articles 107–113 etc.). https://www.leader.ir/en/content/14132/Leadership-in-the-Constitution-of-the-Islamic-Republic-of-Iran — Iranian government primary source [^5]: Wikipedia, “Masoud Pezeshkian” (bio, 2024 election, Interim Leadership Council). https://en.wikipedia.org/wiki/Masoud_Pezeshkian [^6]: Iran International, “Iran’s president offers resignation, citing total takeover by IRGC commanders” (May 31, 2026). https://www.iranintl.com/en/202605312204 — originating report; anti-regime [^7]: IBTimes UK, “Did Iran’s President Pezeshkian Resign Over an IRGC Takeover?” (denials incl. Tasnim, presidency). https://www.ibtimes.co.uk/irans-political-turmoil-irgc-takeover-allegations-1799956 [^8]: The Times of Israel (liveblog), “Citing IRGC overreach, Iran’s Pezeshkian reportedly asks Khamenei to let him quit.” https://www.timesofisrael.com/liveblog_entry/citing-irgc-overreach-irans-pezeshkian-reportedly-asks-khamenei-to-let-him-quit-presidency/ [^9]: The Thursday Times, “Iran’s president reportedly submitted a resignation letter. Here is what we know.” https://thursdaytimes.com/2026/06/01/news/iran-pezeshkian-resignation-letter-irgc-khamenei/ [^10]: Encyclopædia Britannica, “Who is Iran’s supreme leader?” (updated Mar. 23, 2026). https://www.britannica.com/question/Who-is-Irans-supreme-leader [^11]: Wikipedia, “Supreme Leader of Iran” (powers, current officeholder). https://en.wikipedia.org/wiki/Supreme_Leader_of_Iran [^12]: Wikipedia, “Mojtaba Khamenei” (profile, injury, IRGC ties, hardline). https://en.wikipedia.org/wiki/Mojtaba_Khamenei [^13]: Wikipedia, “2026 Iranian supreme leader election” (Assembly process, threshold, candidates). https://en.wikipedia.org/wiki/2026_Iranian_supreme_leader_election [^14]: Associated Press via Fortune, “Iran’s Assembly of Experts picks 56-year-old Mojtaba Khamenei as next Supreme Leader” (Mar. 9, 2026). https://fortune.com/2026/03/09/irans-assembly-of-experts-picks-56-year-old-mojtaba-khamenei-as-next-supreme-leader — wire service [^15]: Middle East Monitor, “Iran’s Assembly of Experts selects Mojtaba Khamenei as new Supreme Leader” (NYT: Guards pushed appointment). https://www.middleeastmonitor.com/20260309-irans-assembly-of-experts-selects-mojtaba-khamenei-as-new-supreme-leader/ [^16]: Rudaw, “How Iran chooses a new supreme leader under its constitution.” https://www.rudaw.net/english/middleeast/iran/01032026 [^17]: United Against Nuclear Iran, “Background: Role of the Supreme Leader” (velayat-e faqih). https://www.unitedagainstnucleariran.com/supreme-leader/background-role-of-supreme-leader — advocacy org [^18]: GlobalSecurity.org, “Valy-e-Faqih (Supreme Spiritual Leader)” (Leader’s functions). https://www.globalsecurity.org/military/world/iran/leader.htm [^20]: JNS, “Mojtaba Khamenei’s disappearance fuels uncertainty over who rules the Islamic Republic.” https://www.jns.org/news/world/mojtaba-khameneis-disappearance-fuels-uncertainty-over-who-rules-the-islamic-republic — anti-regime [^21]: Yahoo News, “Missing in action: What we know about Mojtaba Khamenei’s condition.” https://www.yahoo.com/news/articles/missing-action-know-mojtaba-khameneis-110316922.html [^22]: United Against Nuclear Iran, “IRGC (Islamic Revolutionary Guard Corps)” report (structure, six branches, Khatam al-Anbiya). https://www.unitedagainstnucleariran.com/report/irgc-islamic-revolutionary-guard-corps — advocacy org [^23]: American Enterprise Institute, “Eternal Rivals? The Artesh and the IRGC.” https://www.aei.org/articles/eternal-rivals-the-artesh-and-the-irgc/ — think tank [^24]: WION, “Understanding Iran’s military structure: IRGC, Quds Force, Basij, and the Army.” https://www.wionews.com/world/understanding-iran-s-military-structure-irgc-quds-force-basij-and-the-army-who-are-they-and-what-are-their-roles-1749902681411 [^25]: Associated Press via Yahoo, “Iran’s Islamic Revolutionary Guard controls a sprawling business empire that dominates the economy.” https://www.yahoo.com/news/articles/iran-islamic-revolutionary-guard-controls-175230624.html — wire service [^26]: Wikipedia, “Khatam al-Anbiya Construction Headquarters.” https://en.wikipedia.org/wiki/Khatam_al-Anbiya_Construction_Headquarters [^27]: U.S. Department of the Treasury, press release on Khatam al-Anbiya designations. https://home.treasury.gov/news/press-releases/tg539 — US government primary source [^28]: Gulf News, “Iran paramilitaries have chokehold on economy.” https://gulfnews.com/news/mena/iran/iran-paramilitaries-have-chokehold-on-economy-1.2266118 [^29]: Eurasian Times, “Iran’s IRGC & Pakistan Army: How Two ‘States Within a State’ Evolved” (GDP-share range; Khatam ~14% of GDP). https://www.eurasiantimes.com/irans-irgc-pakistan-army-how-two-states-within-a-state-evolved-from-islamic-guardians-to-business-giants/ [^31]: Fox News, “IRGC commander Ahmad Vahidi may decide Iran’s ceasefire, analysts say.” https://www.foxnews.com/world/who-ahmad-vahidi-irans-new-irgc-chief-tied-global-attacks-death-america-ideology [^32]: Wikipedia, “Mohammad Pakpour” (IRGC commander, killed Feb. 28, 2026). https://en.wikipedia.org/wiki/Mohammad_Pakpour [^33]: CBS News, “Pentagon reveals how B-2 bombers struck Iran nuclear sites in ‘Operation Midnight Hammer’.” https://www.cbsnews.com/news/pentagon-briefing-us-strikes-iran-nuclear-sites/ [^34]: Fox News, “A full breakdown of Operation Midnight Hammer.” https://www.foxnews.com/world/full-breakdown-operation-midnight-hammer-largest-b-2-operational-strike-us-history [^35]: The War Zone, “B-2 Strikes On Iran: What We Know About Operation Midnight Hammer.” https://www.twz.com/air/b-2-strikes-on-iran-what-we-know-about-operation-midnight-hammer [^36]: Air & Space Forces Magazine, “World: Operation Midnight Hammer.” https://www.airandspaceforces.com/article/world-operation-midnight-hammer/ [^37]: GlobalSecurity.org, “Operation Midnight Hammer: Iranian Nuclear Facilities (June 2025).” https://www.globalsecurity.org/military/ops/midnight-hammer.htm [^38]: NewsOnAir (Govt. of India), “U.S.–Iran Nuclear Talks Cancelled Amid Escalating Tensions” (June 15, 2025). https://www.newsonair.gov.in/u-s-iran-nuclear-talks-cancelled-amid-escalating-tensions [^39]: Encyclopædia Britannica, “2026 Iran war” (Operation Epic Fury; ~900 strikes; Khamenei killed; Larijani as de facto leader; Minab school). https://www.britannica.com/event/2026-Iran-war [^40]: Wikipedia, “2026 Iran war” (timeline; Hormuz closure; Hengaw 5,300+; Iran International 4,700). https://en.wikipedia.org/wiki/2026_Iran_war [^41]: Wikipedia, “Assassination of Ali Khamenei” (Feb. 28, 2026; decapitation strike). https://en.wikipedia.org/wiki/Assassination_of_Ali_Khamenei [^42]: The Hilltop, “U.S.-Israel Launch ‘Operation Epic Fury’ Against Iran; Khamenei Killed.” https://thehilltoponline.com/2026/03/04/u-s-israel-launch-operation-epic-fury-against-iran-khamenei-killed/ [^43]: U.S. House Republican Policy Committee, “Background on Iran and Operation Epic Fury” (memo, PDF). https://republicanpolicy.house.gov/sites/evo-subsites/republicanpolicy.house.gov/files/evo-media-document/rpc-iran-operation-epic-fury-memo.pdf — US government primary source [^44]: Wikipedia, “Assassination of Ali Larijani” (Mar. 17, 2026; SNSC secretary; de facto leader). https://en.wikipedia.org/wiki/Assassination_of_Ali_Larijani [^45]: The Jerusalem Post, “IDF kills Iran’s Ali Larijani, Basij commander in largest targeted strikes since Khamenei” (Mojtaba wounded; doubts on capacity). https://www.jpost.com/israel-news/defense-news/article-890225 [^46]: U.S. Energy Information Administration, press release (Apr. 7, 2026): Hormuz closure and production outages drive the forecast (shut-ins; Brent prices). https://www.eia.gov/pressroom/releases/press586.php — US government primary source [^47]: Katadata/Databoks (citing EIA), “The closure of the Strait of Hormuz…” (≈20.1 million b/d, ≈one-fifth of global consumption, Q1 2025). https://databoks.katadata.co.id/en/energy/statistics/6858faba4ec9f/the-closure-of-the-strait-of-hormuz-could-disrupt-oil-supplies-from-saudi-arabia [^48]: UK House of Commons Library, “Israel/US-Iran conflict 2026: Reopening the Strait of Hormuz.” https://commonslibrary.parliament.uk/research-briefings/cbp-10636/ — parliamentary research [^49]: UK House of Commons Library, “US–Iran ceasefire and nuclear talks in 2026.” https://commonslibrary.parliament.uk/research-briefings/cbp-10637/ — parliamentary research [^50]: Federal Reserve Bank of Dallas, “What the closure of the Strait of Hormuz means for the global economy” (Mar. 20, 2026). https://www.dallasfed.org/research/economics/2026/0320 — US central bank research [^51]: U.S. Congressional Research Service via Congress.gov, “Iran Conflict and the Strait of Hormuz” (closure declared early March 2026). https://www.congress.gov/crs-product/R45281 — US government primary source [^52]: Wikipedia, “2026 Iran war ceasefire” (Apr. 8 truce; Iran’s 10-point plan; extensions). https://en.wikipedia.org/wiki/2026_Iran_war_ceasefire [^53]: Wikipedia, “Islamabad Talks” (Apr. 11–12; Pakistan-mediated; no MOU; Hormuz & nuclear sticking points). https://en.wikipedia.org/wiki/Islamabad_Talks [^54]: HotAir (summarizing Axios reporting), “Iran Agreed to Discuss Surrendering Its Uranium Stocks” (tentative 60-day MOU; no final sign-off; Hormuz dispute). https://hotair.com/ed-morrissey/2026/05/24/axios-iran-agreed-to-discuss-surrendering-its-uranium-stocks-n3815252 [^55]: Fox News, “Iran’s ceasefire push may be a ‘cycle of deception,’ analysts warn” (Trump on Iran’s president asking for a ceasefire). https://www.foxnews.com/world/irans-ceasefire-push-may-cycle-deception-analysts-warn-shadowy-figure-gains-power.print [^56]: Iran International, “Tehran rejects US terms as hardliners push escalation.” https://www.iranintl.com/en/202605118029

Additional primary/biographical references consulted: Wikipedia, “Aziz Nasirzadeh” (defense minister, killed Feb. 28, 2026), https://en.wikipedia.org/wiki/Aziz_Nasirzadeh ; Wikipedia, “Majid Khademi” (IRGC intelligence chief, killed Apr. 2026), https://en.wikipedia.org/wiki/Majid_Khademi ; The Jerusalem Post, “IRGC’s Vahidi remains defiant as architect of Iran’s military survival,” https://www.jpost.com/middle-east/iran-news/article-891708 ; FIU News mirror of “Iran’s ruling structure explained,” https://news.fiu.edu/2026/irans-ruling-structureexplained ; Rasanah-IIIS, “The Relationship Between the Supreme Leadership and the Presidency” (absolute velayat-e faqih), https://rasanah-iiis.org ; Kiel Institute, “The Cost of Closing the Strait of Hormuz” (tanker-passage collapse), https://www.kielinstitut.de .

Notes: [^48-note] The “above the law and all institutions” characterization of absolute velayat-e faqih is drawn from the Rasanah-IIIS analysis listed under additional references. [^68-note] The June 2025 Iranian parliamentary vote to close the Strait of Hormuz, with the final decision left to the Supreme National Security Council, is reported by Katadata/Databoks citing Iranian Press TV (same source family as [^47]).

• Where it stands (late May 2026): The Stratos Project — a ~40,000-acre, up-to-9-gigawatt AI data center and gas-power campus backed by Kevin O’Leary’s O’Leary Digital and Utah’s Military Installation Development Authority (MIDA) in Box Elder County’s Hansel Valley — has cleared its land-use gates (MIDA board approval April 24; unanimous Box Elder County Commission approval May 4) but has not secured a single binding environmental permit or any water rights. Developers have now twice filed and twice withdrawn water-rights applications amid record public protest, and no replacement application was on file as of late May.

• Executive Order 2026-03 (signed May 29): Gov. Spencer Cox’s new statewide “Data Center Framework” sets a “higher bar” for data-center review — but it binds executive-branch agencies (DEQ, DNR and their divisions), not MIDA, which is an independent authority that “isn’t subject to the governor’s directive.” Cox himself acknowledged the order gives “no extra authority” to those agencies, and critics call it toothless. It was signed after approval and after weeks of Cox praising the project.

• Will it happen? Phase 1 could be operational within two years if everything breaks the developer’s way, but the project faces a gauntlet of five state-agency permitting processes that DEQ says are at least a year — possibly two — from even beginning in earnest, plus the unresolved binding constraint of water rights. It is best read as plausible but far from certain — a multi-year, multi-billion-to-$100B+ bet whose fate turns on water, air permits and capital, not on the approvals already granted.

Key Findings

1. The land-use approvals are done; the hard approvals are not. MIDA’s board approved creating the Stratos Project Area on April 24, 2026, and the Box Elder County Commission unanimously passed Resolutions 26-11 (consent to the MIDA project area) and 26-12 (an interlocal cooperation agreement) on May 4, 2026. Those votes were necessary because the site is private, unincorporated, unzoned land and state law requires local consent for MIDA projects. But as MIDA, the county, and the governor’s office all stress, this is “the beginning, not the end.” No air permit, water-quality permit, drinking-water approval, wildlife review, or water right has been granted.

2. Water rights are the binding constraint — and the developers keep retreating from that fight. The project cannot operate at scale without water for power generation and cooling. Two applications have now been filed and withdrawn: the Bar H Ranch application (1,900 acre-feet, withdrawn May 5–7 after nearly 4,000 protests/letters) and the Murray Hollow L.C. application (about 11 acre-feet, withdrawn May 22 after nearly 700 protests plus hundreds of letters of concern). As of late May 2026 no replacement application was on file, though developers say they will refile.

3. Executive Order 2026-03 pointedly does not reach MIDA. Reporting is consistent: the independent MIDA “isn’t subject to the governor’s directive,” though the order tells executive agencies to “communicate and coordinate frequently” with it. Cox conceded the order “doesn’t give any extra authority” to DEQ or DNR and largely restates existing law.

4. Five agencies hold the real gates. Cox identified five divisions that must sign off on aspects of Phase 1: Air Quality, Drinking Water, Water Quality, Water Rights, and Wildlife Resources.

5. The timeline is long and uncertain. DEQ Commissioner Tim Davis said air- and water-quality permitting is “at least a year away” and that the full review “could take two years”; he noted there will have to be roughly a year of background air monitoring before the developer can even apply for an air-quality permit. Phase 1’s first gigawatt could be operational within two years if construction proceeds quickly; full buildout could exceed $100 billion and take a decade.

Details

The project and its scale

The Stratos Project (branded “Wonder Valley”) would occupy about 40,000 acres across three parcels in western Box Elder County’s Hansel Valley, near the north end of the Great Salt Lake. O’Leary Digital (chaired by “Shark Tank” investor Kevin O’Leary) is developing it with West GenCo. At full buildout it is described as 7.5 to 9 gigawatts of on-site, mostly natural-gas generation — more than double the roughly 4 GW the entire state uses on average. Phase 1 is centered on a data-center building footprint of fewer than 2,000 acres, with a first phase drawing about 3 GW. The site was chosen because the interstate Ruby Pipeline (natural gas) crosses it; the development agreement allows O’Leary Digital to contract with Tallgrass Energy for gas service.

Investors including O’Leary have put about $20 million into the project so far; total cost could exceed $100 billion at full buildout. O’Leary Digital has said it expects to spend upward of $1 billion just on site-preparation work for Phase 1, with early-phase scaling toward 3 GW potentially approaching $20 billion. After the May 4 vote, an O’Leary Digital spokesperson said the company would begin a capital-raising effort over the following 60 days.

Economics and the county’s stake

Backers project the county would receive about $30 million in new annual revenue in the initial phases, rising to as much as $108 million annually at full buildout, plus roughly 2,000 jobs. In rejecting the referendum, the county auditor estimated repealing the resolutions would cost the county guaranteed revenue (about $5.4 million annually initially, up to $108 million at full capacity) and that the project proceeding without the resolutions would expose taxpayers to roughly $5.4 million per year in added fire, ambulance and police costs. MIDA granted Stratos a steep energy-tax reduction (0.5% versus the 6% MIDA is authorized to collect).

The water-rights saga in detail

• Application 1 (Bar H Ranch): filed March 25, 2026 (Change Application a54385) to transfer 1,900 acre-feet from Salt Wells Spring (water right 13-4148, 1904 priority) from irrigation to industrial use for power generation and the data center. It drew a record protest response — nearly 4,000 formal protests and letters, each costing $15 to file (roughly $58,500 total). Bar H Ranch withdrew it (notice May 5, confirmed May 7 by State Engineer Teresa Wilhelmsen), saying it intends to resubmit with more supporting information. Protests do not carry over to a new application.

• Application 2 (Murray Hollow L.C.): filed April 28 for about 11 acre-feet, also converting stock-watering/domestic use to industrial, “primarily for power generation.” It drew nearly 700 protests; a formal hearing was requested. Murray Hollow filed to withdraw May 22; Wilhelmsen confirmed cancellation May 27.

• Developers say they hold roughly 13,000 acre-feet of existing rights on the property, plan about 3,000 acre-feet on-site, and — per West GenCo co-founder Austin Pritchett — already have around 10,000 acre-feet under contract from the nearby town of Snowville (together, enough to supply more than 20,000 Utah households). Critics (Friends of Great Salt Lake’s Rob Dubuc, Center for Biological Diversity’s Deeda Seed) argue the applications lack a complete water budget and that the developers appear to be accruing many small rights piecemeal.

• A crucial legal shift: Utah’s HB 60 took effect May 6, 2026, narrowing what the State Engineer may consider when weighing “detriment to public welfare” largely to water pollution and water scarcity — stripping out broader public-welfare, recreation, and natural-stream-environment considerations. Observers say this materially advantages the developer’s next application. O’Leary Digital president Paul Palandjian said the “water rights are being adjusted to fit the project scope and phasing and will be re-submitted.”

Executive Order 2026-03: what it does and doesn’t do

Signed Friday, May 29, 2026 (effective immediately) at the Eccles Wildlife Education Center after a closed-door Great Salt Lake roundtable, the order creates a statewide “Data Center Framework” directing executive agencies to prioritize: protecting the Great Salt Lake and water resources (ensuring water consumption is not increased and water quality protected); protecting utility ratepayers from infrastructure costs; protecting air quality; providing transparent public comment; supporting rural economic development; mitigating wildlife impacts; and “pro-human AI development.”

The order quotes Cox: “Utahns have expressed legitimate concerns regarding the potential impacts of large data centers on water resources, air quality, utility rates, local communities and quality of life,” and “With Industry as our state motto, Utah is committed to advancing both economic strength and environmental stewardship.”

The MIDA question — the crux. Multiple outlets confirm the independent MIDA, which approved Stratos, is not subject to the order; the order instead directs executive agencies to coordinate with MIDA. Cox acknowledged the order “doesn’t give any extra authority” to DEQ or DNR and said it doubles down on his prior directive barring new water rights in northern Utah. Cox: “The MIDA decision is only about land use. That’s it… MIDA has no authority over air quality, over water. This is the beginning of a long process.” He said the order “could potentially delay the project for sure.” Critics disagree on its force: BEAR’s Brenna Williams called it “just a PR stunt” that “doesn’t affect us at all”; the Center for Biological Diversity’s Deeda Seed said it “lacks teeth.” (Note: it could not be confirmed from the signed text whether the order contains an explicit MIDA-exemption clause or whether MIDA is simply outside executive-branch reach by operation of law; see Caveats.)

Timing criticism. The order came after MIDA and county approvals and after weeks of Cox defending the project (in April he dismissed “really bad information” about data-center water use). Opponents and even some supporters noted the sequencing; Williams’ group said the order “reads like a list of everything we have been saying since April.”

The five gates (gate-by-gate)

Per Cox’s announcement and Deseret News, five divisions must approve aspects of Phase 1 before construction:

1. Utah Division of Air Quality (DEQ): evaluate projected air-pollutant emissions and mandate control techniques. An air permit/Notice of Intent had not yet been filed; roughly a year of background air monitoring is required before the developer can even apply, followed by a multi-month engineering review with a 30-day public comment period and public hearing.

2. Utah Division of Drinking Water (DEQ): review drinking-water systems, which must be physically separate from cooling/fire/power water.

3. Utah Division of Water Quality (DEQ): permits for all surface- and groundwater discharges to waters of the state, including the Great Salt Lake.

4. Utah Division of Water Rights (DNR): evaluate water availability — the binding constraint, with no application currently on file.

5. Utah Division of Wildlife Resources (DNR): review impacts and recommend mitigation once a formal plan is submitted.

Beyond these, MIDA must still form a Design Review Committee (six of seven members chosen by Box Elder County, expected to be appointed at the June or July MIDA meeting), the developer must submit a master plan and site plans, and building-permit/subdivision/site-plan requirements apply. Cox additionally directed DNR to ensure “the most environmentally-sensitive cooling technology,” required the developer to publish a public water plan showing “no degradation” to the Great Salt Lake (with all water use reported publicly and a commitment that “in no event will the developer reduce water going to the Great Salt Lake”), and directed the Public Service Commission (per SB 132) to ensure the project doesn’t raise Utahns’ energy costs.

The referendum fight

Box Elder Accountability Referendum (BEAR), led by Brigham City economist Brenna Williams, filed two referendum applications May 11 to put Resolutions 26-11 and 26-12 to voters. On May 28–29, County Attorney Stephen Hadfield rejected them, ruling the resolutions were administrative actions implementing existing MIDA law, not legislative acts subject to referendum. Hadfield: “Because I am obligated to interpret and uphold the law, I am legally bound to reject the applications for referenda of these resolutions.” BEAR is appealing, citing a 2018 Utah Supreme Court decision that major land-use decisions can be legislative acts. The group had been prepared to gather roughly 5,400 signatures. Separately, Iron County imposed a 180-day moratorium on new data-center applications — a contrast with Box Elder. House Speaker Mike Schultz said he backed letting residents vote.

Environmental context and the controversy

The Great Salt Lake remains far below healthy levels. DNR executive director Joel Ferry said the lake likely peaked around 4,192.6 feet in spring 2026 — about a foot below the prior year’s high and more than five feet shy of its minimum healthy level (~4,198 feet) — amid record-low snowpack and a declared statewide drought emergency (May 21). Scientists warn of large impacts: USU physicist Robert Davies estimated the campus’s combined thermal load at ~16 GW — “the equivalent of about 23 atom bombs worth of energy dumped into this local environment every single day” — and that a 9-GW gas plant “will raise Utah’s current total greenhouse gas production by about 50%.” Utah Clean Energy’s Logan Mitchell estimated around 35 million metric tons of CO2 annually, enough to raise Utah’s emissions by roughly 64%, “not account[ing] for any additional methane leakage.” Originally pitched as “100% off the Ruby Pipeline” natural gas, Cox later insisted later phases would “never” be gas-only, saying they “should be nuclear, and they should be geothermal, and solar.” O’Leary has called concerns “misinformation,” accused opponents of being “bused in,” and even alleged (without releasing evidence) China-linked funding of opposition groups; he has framed the project as a national-security race against China’s power buildout, citing competition as the reason for “rushing” it.

Recommendations (what to watch and how to read it)

1. Watch for the third water-rights application. This is the single best leading indicator. A refiled, well-documented application (especially one structured to exploit HB 60’s narrower review) signals the developer is serious and proceeding; continued absence signals stall. Threshold that changes the outlook: a consolidated application with a full water budget being accepted for review by the State Engineer.

2. Watch the air-quality Notice of Intent (DEQ/DAQ). Because roughly a year of background air monitoring is required before an application, and a multi-month engineering review plus public comment follows, no NOI means construction is not imminent regardless of rhetoric. The gas-vs-renewables question and the Davies/Mitchell emissions estimates will be litigated here.

3. Watch the BEAR appeal to Utah courts. If a court deems the resolutions legislative (referable), the political risk to the project rises sharply. If the appeal fails, local democratic off-ramps largely close and the fight shifts entirely to permits and water.

4. Watch the capital raise and Phase 1 financing. With only ~$20M in and $1B+ needed just for site prep, financing milestones are decisive. Tenant commitments and a closed construction loan would be strong “go” signals.

5. Watch the MIDA Design Review Committee formation (expected June/July) and the master plan. These are procedural but necessary; delays here cascade.

6. Read the EO as signaling, not a hard brake. Because it doesn’t bind MIDA or expand agency authority, it changes tone and coordination more than legal outcomes. The real constraints remain water rights, air permits, and capital.

Bottom line for decision-makers: Treat Stratos as a live but contingent megaproject. The approvals already in hand (MIDA + county land use) were the easy gates; the binding gates (water rights under a drought emergency, air/water-quality permits at least a year — possibly two — out, and tens of billions in financing) remain unresolved. The most likely near-term outcome is continued slow grind, not imminent groundbreaking.

Caveats

• EO 2026-03 verbatim text: This analysis of the order’s operative clauses relies on the Governor’s Office release and reporting by Deseret News, KUER, Utah News Dispatch, FOX 13 and others rather than a confirmed verbatim read of every numbered clause. Whether the order contains an explicit MIDA-exemption clause (versus MIDA simply being outside executive-branch reach by operation of law) and whether it defines a numeric size threshold for “large” data centers could not be independently confirmed from the signed PDF. One lower-reliability outlet claimed “eight distinct principles”; that count is uncorroborated.

• Fast-moving facts: Water-rights, referendum-appeal, and permitting statuses are as of late May 2026 and may have changed.

• Contested figures: Power capacity is variously cited as 7.5 GW (some filings) and 9 GW (developer fact sheet); emissions and heat estimates come from project critics and academics and have not been validated through formal permitting modeling. Developer water-use claims (closed-loop, “less than current agricultural use,” “net positive to the lake”) are disputed and unproven pending a complete water budget.

• Cost/economic projections ($100B+ total, $30M–$108M county revenue, 2,000 jobs) are developer/county projections, not realized outcomes.

• There is no peer-reviewed evidence that flu vaccines cause spasmodic dysphonia (SD). Every reviewed voice/movement-disorder expert, the Dystonia Medical Research Foundation (DMRF), and the published case-control epidemiology converge on this: SD is a focal laryngeal dystonia of multifactorial origin (basal-ganglia/sensorimotor network dysfunction with genetic predisposition and environmental triggers such as upper respiratory infections, gastroesophageal reflux, neck trauma, voice overuse, and stress) — flu vaccination is not on the list of recognized triggers.

• RFK Jr.’s public attribution to flu vaccination is recent and self-acknowledged as speculation. He was diagnosed in 1996, but the first public claim linking his SD to flu vaccines appears in a 2020 Children’s Health Defense webcast — after he began litigating against flu-vaccine manufacturers in 2016. He told NBC News in 2023 the link was “my own speculation”; told a 2021 podcaster he “haven’t been able to figure out any other cause”; and told USA Today on January 16, 2026, “It’s a potential culprit that I cannot rule out. I can’t prove it.” His sister Kerry Kennedy is also affected — consistent with a genetic/familial predisposition. Yahoo! + 2

• The strongest steel-man — that “dysphonia” appears on a flu-vaccine package insert — does not survive scrutiny. Only one flu vaccine (FluLaval) lists “dysphonia” (a generic symptom term meaning “abnormal voice,” distinct from spasmodic dysphonia) in its post-marketing section, with the FDA-mandated caveat that causal relationship cannot be established. FluLaval was approved by the FDA in October 2006 — a year after Kennedy says he stopped taking flu shots, breaking the temporal logic of the claim entirely. Scientific AmericanScientific American

Key Findings

1. Spasmodic dysphonia (now formally called laryngeal dystonia) is a task-specific focal dystonia with prevalence around 1 per 50,000–100,000. It involves involuntary spasms of the laryngeal muscles during speech and is pathophysiologically a disorder of the basal ganglia–thalamo–cortical sensorimotor network, not the larynx itself. Its etiology is unknown but multifactorial. MM+MNIDCD

2. Known/proposed risk factors include genetic predisposition (12–20% have a positive family history; mutations in THAP1, TUBB4A, GNAL, TOR1A in a minority), recurrent upper respiratory infections, gastroesophageal reflux, neck/laryngeal trauma, intensive voice use, childhood mumps/measles, and psychological stress. Flu vaccination is not on this list in any peer-reviewed case-control study.

3. The Dystonia Medical Research Foundation’s panel of movement-disorder experts concluded there is “no evidence” vaccinations cause dystonia. The investigation was prompted by the highly publicized 2009 case of cheerleader Desiree Jennings, whose post–flu-shot “dystonia” the panel concluded was not, in fact, dystonia (it has been widely characterized in the neurology literature as a psychogenic movement disorder). RationalWiki

4. RFK Jr.’s causal narrative is recent and was triggered by reading package inserts during 2016 litigation against flu-vaccine manufacturers, not by his original clinical history. Through the 1996–2019 period he did not publicly attribute his condition to vaccines. He has repeatedly admitted he cannot prove the link and has called it speculation. Voices For Vaccines

5. VAERS, the system most frequently invoked to argue for vaccine–SD links, is by HHS’s own description “not designed to determine if a vaccine caused a specific adverse health event.” Spontaneous reports cannot establish causation; they require corroboration via the Vaccine Safety Datalink, BEST, or controlled epidemiology. Vaccine Adverse Event Reporting System

6. No peer-reviewed case report has documented spasmodic dysphonia/laryngeal dystonia following flu vaccination. A handful of case reports document cervical dystonia after COVID-19 vaccines (proposed autoimmune/molecular-mimicry mechanism), and laryngeal myasthenia gravis (a different, autoimmune disorder) has been reported once after influenza vaccination. Neither is spasmodic dysphonia. Conversely, post-COVID infection laryngeal dystonia has been described.

7. As HHS Secretary (2025–2026), Kennedy has translated his personal suspicion into policy: under his direction, HHS dropped the universal childhood flu-vaccine recommendation in January 2026 (later blocked by a federal judge in March 2026 as “arbitrary and capricious”), reportedly killed flu-vaccine public awareness campaigns, and reshaped ACIP and the federal vaccine schedule. This has unfolded during the 2025–2026 flu season, which CDC’s FluView classified as “moderate across all ages” by its in-season severity framework but which produced the highest peak pediatric hospitalization rate since 2010–11. The Washington Post + 4

Details

1. What spasmodic dysphonia is, scientifically

Spasmodic dysphonia (SD) — the term unanimously replaced with “laryngeal dystonia” by a 2021 NIH/NIDCD expert panel chaired by Kristina Simonyan (Harvard/Mass Eye and Ear) — is a chronic, task-specific focal dystonia characterized by involuntary spasms of the laryngeal muscles during voluntary speech. The NIDCD describes it as “a disorder affecting the voice muscles in the larynx.” ScienceDirect

Three clinical phenotypes are recognized:

• Adductor SD (≈82% of cases): The thyroarytenoid/lateral cricoarytenoid muscles close the vocal folds too tightly, producing a strained, strangled, broken voice with voice breaks during vowels. This is RFK Jr.’s type. ScienceDirect

• Abductor SD: The posterior cricoarytenoid muscle pulls the vocal folds too far apart, producing a breathy, whispery voice with breaks on voiceless consonants. ScienceDirectMedscape

• Mixed / with tremor: Includes voice tremor and overlap features.

Prevalence is on the order of 1 in 50,000–100,000 (≈50,000 cases in North America). Onset is typically between ages 30 and 50, more often in women (≈63–77% of patients are female), and onset is gradual in most patients but sudden in ~35–45%. RFK Jr.’s 1996 onset at age 42 falls squarely in the typical demographic window. Substack + 3

Pathophysiology: Although classically described in older literature as a disorder of the basal ganglia, neuroimaging since the late 2000s — particularly Simonyan and colleagues’ fMRI, DTI, and PET work — has reframed laryngeal dystonia as a large-scale brain network disorder. Documented abnormalities include:

• Reduced cortical inhibition (shortened cortical silent period on TMS) MedLink

• Hyperactivation of primary somatosensory cortex (S1) in the larynx representation Nature

• Hyperfunctional direct basal-ganglia pathway Springer

• Reduced striatal D2/D3 dopamine receptor binding Chinese Medical Journal

• Microstructural white-matter abnormalities (putamen, globus pallidus, ventral thalamus, cerebellum) PubMed Central

• Postmortem evidence of mild inflammation around brainstem reticular formation nuclei, with minor neuronal degeneration in substantia nigra/locus coeruleus Chinese Medical Journal

Genetics: 12–20% of SD patients have a positive family history. Mutations in THAP1 (DYT6), TUBB4A (DYT4, “whispering dysphonia”), TOR1A (DYT1), and GNAL have been implicated in a minority of cases. A 2016 screen of 86 SD patients (Putzel/Frucht et al.) found only 2.3% had THAP1 variants and none had TOR1A or TUBB4A mutations — implying that for most adult-onset focal SD, genetic predisposition is polygenic and probabilistic, not Mendelian. Notably, RFK Jr.’s sister Kerry Kennedy is widely reported to have the same condition (visible in her 2024 public statements), and other Kennedys (per family commentary) have voice tremor — consistent with a familial predisposition. PubMed CentralWikipedia

Established environmental risk factors (from Schweinfurth 2002, Childs/Murry 2011, Tanner/Roy 2011, and the Simonyan group’s 2019 case-control work on 186 patients):

• Recurrent upper respiratory infections (Simonyan 2019 identified this as the dominant extrinsic risk factor; 30% of SD patients in Schweinfurth associated symptom onset with a URI) PubMed

• History of mumps or measles (65% of patients in Schweinfurth vs. 15% national baseline) PubMed

• Gastroesophageal reflux disease

• Neck/laryngeal trauma (including intubation)

• Intense occupational/avocational voice use

• Major life stress (21% in Schweinfurth) PubMed

• Family history of dystonia, tremor, tics, or meningitis

Treatment: The gold standard since the Blitzer/Brin work of the 1980s–90s is EMG-guided botulinum toxin A injection into the affected laryngeal muscles, typically repeated every 3–4 months. Blitzer’s 24-year series of 1,300 patients reports broadly favorable response; adductor patients achieve ≈90% of normal function for ~15 weeks per cycle. Other options include voice therapy, sodium oxybate (alcohol-responsive subset), and selective laryngeal adductor denervation-reinnervation surgery. There is no cure. PubMed

2. The vaccine claim: origins, evolution, and timeline

RFK Jr. was diagnosed with spasmodic dysphonia in 1996, at age 42. By his own account, in interviews with the Los Angeles Times (April 2024), he first noticed a “very, very slow progression” of vocal worsening over years. He told Oprah Winfrey in 2007 that “it began as a mild tremble for a couple of years.” Through this period and into the 2010s, he did not publicly attribute his SD to vaccines. Yahoo!

The pivot occurred around 2016, when Kennedy was preparing litigation against flu-vaccine manufacturers and read package inserts that listed “dysphonia” among reported adverse events. He told the Los Angeles Times: “That was the first I ever realized that.” Yahoo!

A documented timeline of his public statements:

1996 — Personal Onset of SD at age 42.

2005 — Diane Rehm Show (NPR) Discussed Botox treatment. Did not attribute condition to vaccines.

2007 — Oprah “It began as a mild tremble.” No vaccine attribution.

2016 — Litigation prep Read flu-vaccine package inserts while preparing litigation against manufacturers. Began suspecting a vaccine link. Source: Los Angeles Times

2020 — Children’s Health Defense webcast “I got [spasmodic dysphonia]. That’s what’s wrong with my voice, and it is a listed side effect of the flu vaccine.” Source: Yahoo News / Washington Post

2021 — “Here’s to Your Health with Joshua Lane” podcast “I haven’t been able to figure out any other cause.” Source: Yahoo News

2023 — NBC News (Brandy Zadrozny profile) In a follow-up email, called the link “my own speculation.” Source: NBC News / AOL

Feb 2024 — NewsNation / campaign video Described it as a “neurological injury.” Did not claim vaccine causation on-air.

April 2024 — Los Angeles Times Called the flu vaccine “at least a potential culprit.” Admitted no proof. Source: Newsweek / Yahoo News

Feb 2025 — Joe Rogan Experience #2277 (Woody Harrelson guest) Harrelson and Rogan asserted the flu-vaccine theory. The included Kennedy clip says only “It’s a neurological injury in 1996… I never knew what happened” — no vaccine claim in Kennedy’s own clip. Source: Sportskeeda

Jan 16, 2026 — USA Today “Extremely Normal” interview “Do I know whether that was caused by my annual flu shot? I have no idea. It’s a possibility. It’s a potential culprit that I cannot rule out. I can’t prove it.” Source: Scientific American / Yahoo News

The pattern is clear: a roughly 24-year gap between his diagnosis and his first public causal attribution to flu vaccines, with the attribution emerging only after he was financially and professionally invested in litigation against flu-vaccine manufacturers. Kennedy himself oscillates between “potential culprit I cannot rule out” and “my own speculation.”

Other promoters: Actor Woody Harrelson, on Joe Rogan Experience #2277 (February 2025), claimed the connection as established fact, with Rogan reinforcing it (”he believes it’s a vaccine injury, which is very ironic”). Children’s Health Defense, the anti-vaccine nonprofit Kennedy founded, has amplified the claim. USA Today, the Los Angeles Times, Newsweek, the Washington Post, Scientific American, and NBC News have all reported on the claim while flagging that experts reject it. AmericanentertainmentnewsSportskeeda

3. What the actual evidence shows

Peer-reviewed studies linking flu vaccines to SD: zero. A targeted PubMed search returns no case reports of spasmodic dysphonia/laryngeal dystonia following influenza vaccination. The closest entries are:

• Cervical dystonia (not laryngeal) after COVID-19 vaccines: Algahtani 2022 (PMID 35605509) reports a 38-year-old with neck twisting 24 hours after Pfizer-BioNTech; proposed mechanism “autoimmune.” Scorza/Finsterer 2022 (PMID 36174407) discuss broader spontaneous-report signals from VigiBase and EMA databases. These remain spontaneous-report-driven signals, not established causation.

• Laryngeal myasthenia gravis (a different disorder — autoimmune neuromuscular junction failure, not dystonia) after influenza vaccination (PMID 34559977): one case in a 58-year-old woman.

• Post-COVID infection (not vaccine) laryngeal dystonia: Sigal & Raina 2024 (RIICS) describe two women developing adductor laryngeal dystonia ~1 month after COVID-19 infection. Proposed mechanism: peripheral laryngeal nerve inflammation with maladaptive sensorimotor remapping.

• Systematic review in Journal of Movement Disorders of COVID-19 vaccine-related movement disorders identified 10 patients across 8 articles; the most common was hemichorea, with one cervical dystonia case — no laryngeal/spasmodic dystonia.

VAERS limitations. The Vaccine Adverse Event Reporting System is, by the CDC and HHS’s own statements, a passive, unverified, spontaneous-reporting safety-signal system. From the official HHS guidance:

“A report to VAERS generally does not prove that the identified vaccine(s) caused the adverse event… VAERS reports may contain information that is incomplete, inaccurate, coincidental, or unverifiable… The number of reports alone cannot be interpreted as evidence of a causal association.”

Establishing causation requires (per the post-marketing surveillance framework taught to vaccine providers):

• Known biological plausibility

• Consistent temporal relationship

• Positive rechallenge on re-administration

• Specific clinical syndrome

• Confirmation in controlled studies (e.g., Vaccine Safety Datalink, BEST) showing higher risk in vaccinated than unvaccinated populations

None of these criteria are satisfied for flu-vaccine→SD.

The DMRF expert panel. The Dystonia Medical Research Foundation’s “Common Misunderstandings” page states unambiguously:

“There is no evidence that vaccinations cause dystonia. The DMRF asked a panel of preeminent movement disorder experts to examine this closely several years ago when a story surfaced of a woman who claimed to have developed ‘dystonia’ following a flu immunization. It was later established the dystonia diagnosis was not correct. In contrast, preliminary research suggests certain vaccinations may actually be protective against certain forms of dystonia.” Scientific AmericanDystonia Foundation

The “woman” was Desiree Jennings, a 25-year-old Redskins cheerleader who in 2009 was the subject of viral video clips appearing to show her walking only backwards or sideways after a flu shot. Multiple expert neurologists publicly concluded her presentation was inconsistent with true dystonia (most characterized it as a psychogenic/functional movement disorder); follow-up reporting showed her symptoms resolving after non-evidence-based “chelation” and hyperbaric treatment, and observers documented her walking normally afterward — the natural history of a functional movement disorder, not focal dystonia.

Possible protective effects of vaccination. Some preliminary preclinical work suggests certain vaccinations (notably the BCG vaccine in rodent models) may have protective effects against neurological symptoms, including in models of haloperidol-induced tardive dyskinesia. Schweinfurth’s 2002 case-control study also found that 65% of SD patients had a history of measles or mumps versus a 15% national baseline — an association that would, if anything, suggest that the wild-virus exposure (not the MMR vaccine) might be the relevant risk factor for some SD cases. This is the basis of the DMRF’s statement that vaccination may be protective against some forms of dystonia. nihPubMed

The “package insert” argument. The strongest version of Kennedy’s claim is: dysphonia is listed in the FluLaval flu-vaccine package insert, so it must be a known side effect. This collapses on three points:

1. “Dysphonia” ≠ “spasmodic dysphonia.” Dysphonia is a generic descriptive term for “abnormal voice” — hoarseness, breathiness, weakness from any cause (including the local irritation/swelling of an injection-site reaction reaching nearby tissues, or a transient laryngitis from concurrent illness). Spasmodic dysphonia is a specific neurological disorder. Dr. Michael Johns III (USC Voice Center): “Dysphonia just means abnormal voice. That’s all what dysphonia means. So that’s a symptom. I’m hoarse: I have dysphonia. So it’s not the same thing.” Yahoo!

2. Package-insert post-marketing sections are not causality determinations. The FluLaval insert itself states verbatim: “Because these reactions are reported voluntarily from a population of uncertain size, it is not always possible to reliably estimate their frequency or establish a causal relationship to the vaccine.” The FDA requires manufacturers to list these because they were reported, not because they are caused. Dr. Robert Hopkins (National Foundation for Infectious Diseases): listing dysphonia “means that there was at least one voice disorder during clinical trials. That doesn’t constitute proof.” FDA

3. The timing is wrong. FluLaval was approved by the FDA in October 2006 — one year after Kennedy says he stopped taking flu shots in 2005, and ten years after his 1996 SD onset. He could not have received FluLaval before developing SD. A Washington Post review of 11 flu-vaccine package inserts found “dysphonia” listed on only that one product; “spasmodic dysphonia” appears on none of them.

Expert testimony directly on Kennedy’s claim:

• Dr. Steven Bielamowicz (GWU Voice Treatment Center, 30 years, 300+ SD patients): “No one else has ever claimed or felt that they developed this condition due to a vaccination of any kind.”

• Dr. Michael Johns III (USC Voice Center): “We just don’t know what brings it on. Intubation, emotional trauma, physical trauma, infections and vaccinations are all things that are incredibly common. And it’s very hard to pin causation on something that is so common when this is a condition that is so rare.” Yahoo!

• Dr. Robert Hopkins (NFID): “We do not have any known linkage between flu vaccine and voice disorders like this.” Scientific American

• Dr. Timothy Brewer (UCLA epidemiology): “With these influenza vaccines there are real benefits that so far outweigh the potential harm cited here that it’s not worth considering those types of reactions further.” Yahoo!

• Dysphonia International (formerly the National Spasmodic Dysphonia Association) has explicitly stated there is no scientific basis for the claim.

4. Broader context: RFK Jr.’s health claims and his HHS role

Kennedy has a documented history of distinctive medical narratives:

• Mercury poisoning from tuna (≈2010–2012): Disclosed in his 2012 divorce deposition; blood mercury “more than 10 times” EPA safety thresholds; he attributed it to heavy tuna consumption and reported recovery via chelation.

• “Brain worm” (parasite): Also disclosed in the 2012 deposition; Kennedy said a doctor told him a parasitic worm “got into my brain and ate a portion of it and then died.” Medical experts (Hotez, Gounder, Schulder) noted this is most consistent with neurocysticercosis — a calcified tapeworm cyst — not a “worm that ate brain tissue.”

• Atrial fibrillation: Decades-long, reportedly resolved more than a decade ago.

• Past heroin addiction (14 years): Publicly acknowledged. Some voice physicians have noted that drug-related dystonias exist (e.g., neuroleptic-induced), but RFK Jr.’s past stimulant/opioid use does not have a clean mechanistic link to laryngeal dystonia either; it is one of multiple unproven hypotheses.

These claims are relevant because they fit a pattern: distinctive personal-medical narratives offered by Kennedy that are repeatedly reframed or revised over time, and that sometimes outpace what mainstream medicine can corroborate.

HHS Secretary actions 2025–2026:

• Feb 13, 2025: Confirmed as HHS Secretary, 52–48 (Senator Cassidy voted yes after extracting vaccine commitments Kennedy has since broken).

• June 9, 2025: Per the HHS.gov press release, “Under the leadership of HHS Secretary Robert F. Kennedy, Jr., the agency removed the 17 sitting members of the ACIP committee.” CIDRAP

• Aug 2025: Reinstated the Task Force on Safer Childhood Vaccines (disbanded in 1988), at the request of Children’s Health Defense — the anti-vaccine group Kennedy founded and led.

• Jan 5, 2026: Acting CDC Director Jim O’Neill (HHS Deputy Secretary serving in dual role) signed a decision memorandum overhauling the childhood vaccine schedule: vaccines against flu, COVID-19, rotavirus, hepatitis A, hepatitis B, RSV, and certain meningococcal disease moved from universal recommendation to “shared clinical decision-making.” Total recommended childhood vaccines reduced from 17 to 11. HHS.govPBS

• Mar 16, 2026: U.S. District Court Judge Brian Murphy (District of Massachusetts, Boston) temporarily blocked the schedule changes, finding them “arbitrary and capricious” and writing: “the government has disregarded those methods and thereby undermined the integrity of its actions.”

• Throughout 2025–2026: HHS reportedly killed flu-vaccine public-awareness campaigns. David Geier — previously disciplined by state regulators for practicing medicine without a license, and known for the discredited vaccines-cause-autism hypothesis — was hired by HHS to study vaccine–autism connections.

This is all unfolding during the 2025–2026 flu season, which CDC’s FluView Week 6 (ending Feb 14, 2026) classified as “moderate across all ages” by its in-season severity assessment framework — not “particularly severe” as some early reporting suggested — but which set the second-highest peak weekly hospitalization rate overall and the highest peak pediatric hospitalization rate since 2010–11. CDC FluView Week 7 (ending Feb 21, 2026) estimated “at least 25,000,000 illnesses, 330,000 hospitalizations, and 20,000 deaths from flu so far this season,” with 79 reported pediatric deaths at that point. Per FluView Week 6: “approximately 90% of reported pediatric deaths this season have occurred in children who were not fully vaccinated against influenza.” CDC + 3

5. Honest assessment: where is the evidence uncertain vs. clear?

The strongest steel-man for the claim:

1. Spasmodic dysphonia is recognized to have environmental triggers, and upper respiratory infections are one of them. A flu vaccine introduces influenza antigens and (in some formulations) adjuvants that can cause a brief systemic inflammatory response and rarely a Guillain-Barré-Syndrome–like neurological event. If URIs trigger SD via altered laryngeal sensory feedback (as Simonyan 2019 proposed), it is at least biologically conceivable that a flu shot — which can produce flu-like symptoms — could in some idiosyncratic case act as a similar trigger in a genetically predisposed individual.

2. There is precedent in the broader vaccine-and-movement-disorder literature. Peer-reviewed case reports document rare cervical dystonia and other movement disorders following several vaccines, with proposed autoimmune/molecular-mimicry mechanisms. The mechanism is not impossible.

3. Package-insert listings reflect some signal during trials or post-marketing, even if attenuated by the FDA’s causality caveat. Voice disorders did occur in some patients enrolled in some flu-vaccine trials at a frequency the manufacturer thought worth disclosing.

4. VAERS does include a (small) number of dysphonia/dystonia reports following flu vaccination — and absence of a published case report does not equal absence of any case.

Where the steel-man fails:

1. Specificity. Spasmodic dysphonia is a precise neurological diagnosis (with neuroimaging and EMG correlates), not the same entity as “dysphonia.” No peer-reviewed case report has documented post–flu-vaccine laryngeal dystonia. The published vaccine-dystonia reports are overwhelmingly cervical — different muscle group, different network, different proposed mechanism.

2. Epidemiology. Hundreds of millions of flu vaccines have been administered annually for decades. If even a small fraction induced SD, the case-control epidemiology and large-database studies (Vaccine Safety Datalink, BEST, etc.) would have detected a signal. They have not. SD’s incidence is consistent with its baseline rate in unvaccinated populations.

3. Temporal logic for RFK Jr.’s specific case. He developed SD in 1996 over a slow progression starting in the early 1990s. The only flu vaccine that ever listed even generic “dysphonia” in its insert (FluLaval) was approved in October 2006. His sister has the same condition (consistent with familial predisposition). Substack

4. Provenance of his belief. Kennedy himself dates the idea of vaccine causation not to clinical observation in 1996 but to reading inserts in 2016 during litigation preparation. This is post-hoc reasoning in a context of motivated cognition (financial and ideological stakes against flu-vaccine manufacturers). Yahoo!

5. DMRF expert review. A formal expert panel investigated the most prominent claimed case of flu-shot-induced dystonia and concluded it was not dystonia at all.

What real evidence for the claim would look like:

• A statistically significant signal in active surveillance (Vaccine Safety Datalink, BEST, VigiBase) showing elevated SD/LD incidence in flu-vaccinated vs. unvaccinated cohorts, controlled for confounders.

• Peer-reviewed case reports with documented temporal relationship, EMG/laryngoscopic confirmation of dystonia (not generic dysphonia), and a plausible immunological correlate (e.g., anti-neuronal antibodies).

• A positive rechallenge case (which would be ethically problematic to test deliberately, but could emerge from inadvertent re-administration).

• A specific biological mechanism (e.g., molecular mimicry between an influenza protein and a basal-ganglia/laryngeal-cortex antigen, demonstrable in vitro).

None of this evidence exists for flu vaccine→SD. Some of it exists weakly for other vaccine-dystonia associations (mostly cervical, mostly COVID-19), but the analogy is not a substitute for the actual data.

Genuine uncertainties (where intellectual honesty requires acknowledgment):

• The etiology of SD remains genuinely unknown; what causes RFK Jr.’s specific case can’t be known with certainty, and “unknown” is not “vaccine.”

• Vaccines can rarely trigger neurological adverse events (Guillain-Barré with some flu formulations is the textbook example). Categorical denial that any vaccine could ever trigger any focal dystonia in any patient is itself an overreach.

• VAERS underreporting is real; passive surveillance has well-documented limitations on both sides.

• The question of whether any flu vaccine has ever caused a single case of focal laryngeal dystonia in any patient is technically open — but the burden of proof for an affirmative claim is not met.

Recommendations

For the curious reader trying to evaluate the claim:

1. Distinguish the personal speculation from the policy. Kennedy himself does not assert proof; he asserts a possibility he cannot rule out. That is a defensible epistemic stance for a private individual. It is a much weaker basis for federal policy that downgrades flu-vaccine recommendations affecting hundreds of millions of people.

2. Distinguish “dysphonia” from “spasmodic dysphonia.” This conflation is the single most common source of confusion in the public discourse. If you see anyone citing “dysphonia is on the package insert,” ask whether they understand the difference.

3. Treat VAERS data with the skepticism HHS itself prescribes. It is a hypothesis-generating system. Hypotheses get tested in the Vaccine Safety Datalink or BEST.

4. For decision-making about your own flu shot: SD is rare (≈0.001–0.002% population prevalence); per CDC FluView Week 7, flu had caused at least 20,000 deaths and 330,000 hospitalizations in the U.S. by mid-season 2025–2026, with 79 pediatric deaths, ~90% of whom were not fully vaccinated. The risk-benefit ratio overwhelmingly favors vaccination in nearly every demographic.

Benchmarks that would change my assessment:

• A controlled epidemiological study (case-control or cohort, n in the thousands) showing statistically significant elevated incidence of laryngeal dystonia in flu-vaccinated versus unvaccinated populations.

• Two or more peer-reviewed case reports of EMG/laryngoscopically confirmed adductor or abductor SD onset within a defined temporal window of flu vaccination, with autoimmune/inflammatory correlates and exclusion of alternative causes.

• A formal reversal of position by the DMRF, the NIH/NIDCD expert panel, or major movement-disorder neurology bodies.

Until then, the rational position is: flu vaccines have not been shown to cause spasmodic dysphonia, RFK Jr.’s personal case is most parsimoniously explained by the standard etiology (familial predisposition + sporadic adult onset around age 40), and policy decisions on flu vaccination should not be driven by a single anecdotal, self-acknowledged speculative claim.

Caveats

• The chain of inference relies heavily on absence-of-evidence reasoning in some places (e.g., “no peer-reviewed case report exists” is informative but not absolute proof of impossibility for a rare event).

• Some sources cited are journalistic (Scientific American, Washington Post, NBC News, USA Today, Los Angeles Times, Newsweek, CNN, Time) summarizing primary medical and interview material. Where possible, primary sources (NIDCD, DMRF, FDA package inserts, PubMed-indexed journal articles, CDC/HHS official guidance, CDC FluView) are preferred.

• The 2020 Children’s Health Defense webcast quote is attested by Washington Post reporting but the specific episode date is not in the public record I could verify.

• The exact 2021 Joshua Lane podcast episode date is similarly not pinned down in available reporting.

• Some commentary cited (substacks by physicians, partisan outlets like Children’s Health Defense, Infowars, Showbiz411) is clearly labeled as such; their factual content was used only where it could be cross-corroborated with primary or higher-quality sources.

• The political context (a sitting HHS Secretary advancing this claim, a federal lawsuit, the ongoing 2025–2026 flu season) means coverage from any source — including mainstream medical outlets — should be read with awareness of the high-stakes policy implications.

• The 2025–2026 timeline of HHS policy actions is current as of May 21, 2026; Judge Murphy’s March 2026 injunction was characterized as temporary, and final disposition may differ.

• The CDC’s “moderate” classification of the 2025–2026 flu season refers to its formal in-season severity framework comparing across all age groups; this does not contradict the simultaneous fact that pediatric hospitalization peaked at the highest rate since 2010–11. Both can be true, and both are relevant to interpreting RFK Jr.’s contemporaneous public statements downgrading the flu vaccine.

There’s a version of the Domo story where everything works out. Where the bold Utah founder who already sold one company to Adobe for $1.8 billion builds a second act even bigger than the first. Where the audacious bet on a real-time business intelligence platform lands at exactly the right moment — right as enterprises start drowning in data and desperately need someone to make sense of it.

That version didn’t happen.

Instead, Domo became one of the most instructive cautionary tales in modern SaaS: a company with real revenue, a real product, and real customers that is nonetheless worth almost nothing — trading at roughly 50 cents for every dollar of annual revenue — because the math was broken from the very beginning, and nobody fixed it in time.

Here’s how it happened.

Act One: The Man Who Already Won

To understand Domo, you have to understand Josh James first.

James co-founded Omniture in 1996 out of Provo, Utah — back when Utah wasn’t a tech hub and Provo was barely a blip on anyone’s radar. Omniture did web analytics: helping companies understand who was visiting their websites and what they were doing. It wasn’t glamorous. It was infrastructure. But James built it into a real business, took it public on the Nasdaq in 2006, and three years later sold it to Adobe for $1.8 billion.

He personally walked away with roughly $150 million from the sale. He was, briefly, the youngest CEO of a Nasdaq or NYSE-listed company in the country.

This matters because of what it did to the fundraising environment for whatever he did next. In Silicon Valley, a founder who has already returned $1.8 billion to investors is basically a golden ticket. When James started Domo in 2010 and began quietly raising money, VCs didn’t ask many hard questions. They opened their wallets.

That dynamic — a celebrated founder getting essentially unconditional funding — is the first domino.

Act Two: Building a $2 Billion Company Nobody Had Heard Of

Domo spent its first five years in stealth. This was unusual even by startup standards. Most companies want press, customers, visibility. Domo wanted none of it.

What it did want was staff. Lots of it. By the time it came out of stealth in 2015 with a splashy announcement and a $2 billion valuation, it had already hired hundreds of people, spent enormous sums on sales and marketing, and built what James described as a management team suited for a $5 billion revenue company — despite having less than $50 million in revenue at the time.

That quote is worth sitting with. A management team for a $5 billion company, at $50 million in revenue. It’s a tell. It reflects a philosophy that if you build the scaffolding of a giant company, the giant company will come. Sometimes that’s true. For Domo it was not.

By the time of the IPO in 2018, the numbers were stark: Domo had burned through roughly $740 million in funding to reach $107 million in annual recurring revenue — $7 spent for every $1 of recurring revenue generated. Box — famous for its own pre-IPO profligacy — had burned $400 million to reach $233 million in ARR. Domo had burned nearly twice as much for less than half the revenue.

The company also had a deeper structural problem baked in: it cost them somewhere between $300,000 and $650,000 to acquire each net new customer, against an average contract value of around $67,000. At that ratio, you need to hold a customer for roughly eight years just to break even on the cost of signing them — a median payback period of 98 months. And they were signing enterprise customers with all the complexity and churn risk that entails.

Act Three: The IPO That Wasn’t Really a Choice

By spring 2018, Domo was running out of road.

The company had $72 million in cash and was burning roughly $37 million per quarter — less than six months of runway. Buried in their IPO filing was a sentence that made Wall Street wince: “if other equity or debt financing is not available by August 2018, management will then begin to implement plans to significantly reduce operating expenses.”

They were basically saying: we go public now, or we start cutting everything.

This is important context for understanding the IPO. It wasn’t a triumphant market debut. It was a survival mechanism. And it came with a structural trick that most people glossed over.

Before the IPO, Domo had raised $715 million in venture capital, almost all of it with standard liquidation preferences. In plain English: if the company were sold for $500 million, the VCs would get their money back first — all $715 million worth — and common stockholders like James himself would get nothing. The math of a private sale at any reasonable valuation was catastrophic for the founder and employees.

But an IPO changes everything. When you go public, all those preferred shares convert to common stock. The liquidation preferences evaporate. Suddenly Josh James’s equity stake is worth real money at any valuation — whereas in a private acquisition at the same price, he would have received nothing.

So the IPO wasn’t just about raising cash. It was about converting a cap table that would have paid James nothing into one that paid him something. The market priced it accordingly. Domo had been valued privately at $2.28 billion. It IPO’d at $511 million — a 77% haircut before the opening bell.

Act Four: The Treadmill

Here’s the part that explains where Domo is today.

When you have high customer churn and mediocre net revenue retention, your sales and marketing budget isn’t funding growth. It’s funding survival. Domo’s strategy has always been to outsell its churn — and it’s failing. Net revenue retention has sat under 100% — some of the lowest in enterprise SaaS — meaning existing customers are spending less each year on net. You’re spending $150 million a year on sales just to stay flat.

That’s exactly what happened. Revenue has been essentially frozen at $317–319 million for three consecutive fiscal years. Domo is running as fast as it can just to stay in place.

Meanwhile, the competitive landscape shifted underneath them. Microsoft bundled Power BI into Office 365. Salesforce absorbed analytics into its platform. Snowflake, Databricks, and a new generation of cloud data tools commoditized the infrastructure Domo was built on. Customers began churning to broader BI tools already included in larger enterprise suites — analytics bundled into bigger contracts where Domo couldn’t compete on price.

And then there’s the debt. In the frantic pre-IPO period, Domo drew down a $100 million credit facility from BlackRock to stay alive. After the IPO, instead of paying it down, they rolled it. Then rolled it again. And again. It now sits at $126 million, due in August 2028, at an interest rate of roughly 12.2%. That’s about $15 million a year in interest payments — just gone, before a single salesperson gets paid or a single server gets provisioned.

The full income statement in a given year looks like this: $318 million in revenue, roughly $239 million gross profit, then $152 million in sales and marketing, $88 million in R&D, $56 million in administrative overhead, and $15 million in interest. That’s a $59 million net loss. Every year. For fifteen years running.

The accumulated deficit now stands at $1.55 billion.

Act Five: The Endgame

In February 2026, Domo’s board did what boards do when they’ve run out of organic options: they announced a formal strategic alternatives process, hired Jefferies as their investment bank, and started quietly shopping the company.

The translation is straightforward. They have $43 million in cash and no remaining availability under their credit facility. The debt matures in two and a half years. They are not going to generate enough free cash flow to pay it off. A sale is almost certainly the outcome — the only question is to whom and at what price.

Analysts peg a likely takeout in the range of $7 to $11 per share, implying a total enterprise value somewhere between $400 million and $600 million including the debt a buyer would absorb. That’s a 1.2x to 1.8x revenue multiple — extraordinarily cheap by any SaaS benchmark, where healthy companies routinely trade at 5x to 10x revenue.

The stock itself trades around $3.50 — well below even the low end of those estimates — because the market is pricing in the probability that no deal happens at all. Domo’s Altman Z-Score of -10.73 places it in the distress zone, indicating potential risk of bankruptcy within two years. If negotiations drag, cash dwindles further. Equity holders are last in line.

For Josh James personally, it’s a melancholy postscript to what should have been a legendary second act. His Domo stake is worth roughly $5 million today. His real wealth — estimated around $163 million — comes almost entirely from what he made selling Omniture to Adobe in 2009. Domo, for all its years and drama, has been a net destroyer of his wealth, not a builder of it.

Oh, and one more detail that landed in the IPO filing and quietly disappeared: Domo had been spending $700,000 a year leasing a private jet owned by Josh James, and another $600,000 across a catering company and an interior design firm co-owned with his brothers. All terminated just before the IPO when the scrutiny arrived. A small thing in the context of $740 million in losses. But telling.

What Domo Actually Teaches Us

The story isn’t really about Domo. It’s about a set of dynamics that repeat themselves in tech, over and over, with different names attached.

The halo founder problem. When a founder has already had one massive exit, the normal filters on fundraising and spending get suspended. Investors hand over money on faith. That money funds a culture of spending that becomes nearly impossible to unwind once it’s established.

The growth-or-die trap. Domo was always spending to outrun its churn. They could never cut sales and marketing aggressively because doing so would cause revenue to shrink faster than the savings materialized. They were on a treadmill they couldn’t step off.

The IPO as escape valve. The decision to go public wasn’t strategic. It was structural. The liquidation preferences on the cap table made a private sale worthless to the founder and employees. Going public converted a bad cap table into a mediocre one. This is more common than most people realize.

Debt as slow poison. The $100 million BlackRock facility taken out in desperation before the IPO has been rolling and compounding for eight years. At 12.2% on $126 million, it now costs more annually than many companies’ entire engineering budgets. Domo didn’t borrow to invest in growth — they borrowed to survive, and now they’re paying interest on survival.

The outcome, in all likelihood, is that someone — a private equity firm or a larger strategic player — buys Domo for something in the $400–500 million range, takes it private, slashes the sales and marketing spend down to a fraction of what it is today, accepts that revenue shrinks by 20–30%, and tries to run the remaining $200–250 million in revenue as a profitable, if modest, business.

The employees who’ve been there since the early days will get some liquidity. Josh James will move on to whatever comes next. The VCs who backed the 2018 IPO will mostly have lost money. BlackRock will get their loan repaid. And a perfectly good piece of enterprise software will quietly serve a few thousand companies that genuinely rely on it, no longer weighed down by the ambitions of the $2 billion unicorn it was always supposed to become.

It’s not a tragedy. It’s just the market being honest about what a business is actually worth, eventually, no matter how long the pretending goes on.

If you found this useful, the best signal is sharing it with someone who’s building a company or investing in one. The Domo story is not unique — it’s a template. The names change; the dynamics don’t.

Sources

1. Domo Q1 FY2026 Earnings Release — SEC/Domo

2. Domo FY2025 Q4 Earnings Release — SEC/Domo

3. Domo FY2026 Annual Report (10-K) — StockTitan

4. Domo Financial Statements — StockTitan

5. Domo Stock Quote — Morningstar

6. Domo Market Cap History — Public.com

7. Domo Strategic Alternatives Press Release — SEC Filing, Feb 19 2026

8. Domo Explores Strategic Alternatives — GuruFocus

9. Domo Takeout Valuation Analysis — Simply Wall St / Yahoo Finance

10. Domo Full Year 2025 Earnings: S&M Breakdown — Yahoo Finance

11. Domo Debt: Term Loan Extended to 2028 — Domo IR

12. Domo Debt & Liquidity Concerns — Trefis

13. Domo Downgraded to Sell — Seeking Alpha

14. 5 Interesting Learnings from Domo at $320M ARR — SaaStr

15. Domo: Cautionary Tale of Hype & Inefficiency — OnlyCFO

16. Domo IPO S-1 Breakdown — Meritech Capital

17. Domo IPO S-1 Breakdown — Alex Clayton / Medium

18. Domo’s Deal With the Devil — The Hustle

19. Domo IPO Filing Scrutiny — Deseret News

20. Domo IPO: Related Party Transactions — Yahoo Finance / Business Insider

21. Domo IPO Trading Debut — CNBC

22. Domo’s Strategy to Outsell Churn — MarTech

23. Why Domo Stock Is Nosediving — Yahoo Finance

24. Josh James Wikipedia

25. Josh James Insider Trades & Net Worth — Benzinga

26. Josh James Insider Holdings — GuruFocus

27. Domo Launches with $200M Funding — TechCrunch

28. Domo Series A — Benchmark Capital Press Release

29. Domo Funding History — Tracxn

30. Domo Balance Sheet — Simply Wall St

What O’Leary and the project actually say about water

O’Leary’s most-cited public statements form a layered defense. On X on May 10, 2026, he called concerns about draining the Great Salt Lake “ridiculous” and went further: “if anything, we’ll be adding to the Great Salt Lake because the water rights on that land will be used and one of the hyperscalers even said to us, ‘Let’s use air cooled so there’s no water involved.’” On May 5 he claimed the project could blend in “air-cooled turbines,” solar, wind, and batteries — a statement that directly contradicted MIDA Executive Director Paul Morris’s April 22 testimony to Box Elder County that “100% of the power will be generated” from natural gas via the Ruby Pipeline. To CNN Business on May 9, O’Leary repeated the “ridiculous” framing and said developers would “invest in new technologies to reduce the facility’s water usage.”

The most consequential admissions come from CEO Paul Palandjian (O’Leary Digital). On KUER, he said the project would purchase 2,800 AF of water rights and use “just a fraction of it,” with another 10,000 AF available if needed, and that “there’s some rough order of magnitudes that if you go in your AI and ask, on a zero water solution power gen and on a zero water solution closed-loop system, on average, how much acre feet water we use, you’ll get as good a number as I could give you right now.” He could not provide a specific number. Hilary Venable (MIDA Project Director) told the Herald Journal long-term water demand would be “closer to what you would see from a large office campus, not traditional industrial use.” Morris told county commissioners the project would “use less water than ranching” and produce a “net positive to the Great Salt Lake.” Governor Cox’s official FAQ on Stratos states the design is a “closed-loop chilling system combined with dry (air-based) cooling. There is no continuous water draw.”

The numbers the developer has actually committed to paper

Only one water filing exists in the public record. Bar H Ranch, Inc. filed Change Application a54385 on March 25, 2026, seeking to convert 1,900 acre-feet/year from Salt Wells Spring Stream (water right 13-4148, 1904 priority) from agricultural irrigation to year-round industrial use for “Wonder Valley Utah“ — described in the filing itself as a 7.5 GW natural-gas power plant plus a “closed-loop” data center. Public protests reached roughly 3,800 formal filings plus ~2,000 letters of concern, a State Engineer’s office record. Consultant Logan Riley emailed the Division of Water Rights on May 5, 2026 withdrawing the application “in light of the County’s action on Monday” and signaling resubmission “with additional supporting information.” State Engineer Teresa Wilhelmsen confirmed cancellation on May 7. The withdrawal came one day after HB 60 took effect on May 6, narrowing the State Engineer’s authority to consider “public welfare” or “interference with more beneficial use” — a timing critics including the Great Basin Water Network and Center for Biological Diversity called a tactical reset to refile under a more permissive standard.

Beyond the withdrawn 1,900 AF filing, the developer’s publicly described portfolio totals roughly 13,000 acre-feet per year (~4.24 billion gallons): ~3,000 AF on-site plus ~10,000 AF under contract near Snowville. The Snowville rights have not been filed with the State Engineer and exist only as developer statements. Box Elder County Resolution 26-11 (passed 3-0 on May 4, 2026) consented to MIDA designating the 40,000-acre Stratos Project Area; the county explicitly disclaimed authority over water and air, with Commissioner Lee Perry stating “Our vote today had nothing to do with water or air quality.” The accompanying DEQ letter confirmed no Notice of Intent has been filed with the Division of Air Quality — meaning the engineering disclosures that would normally pin down water use (BACT analyses, cooling technology selection) do not yet exist publicly. HB 76 (Data Center Water Transparency Amendments) does apply: Stratos meets the ≥75 AF/year threshold and the “new large data center” definition (operations after July 1, 2026), but the law allows confidentiality claims under §63G-2-305(2) for cooling, discharge, and reuse details, and caps fines at $100/day.

What independent analysts have calculated

Three serious quantitative estimates exist, and they disagree by an order of magnitude depending on assumed technology. Utah Clean Energy (Logan Mitchell et al., May 7, 2026) calculated power-generation water alone — explicitly excluding data-center cooling and upstream gas — at 16.6 billion gallons/year for a combined-cycle gas turbine (CCCT) design (≈25,000 Olympic swimming pools) or 2 billion gallons/year for reciprocating engines (RICE); the RICE scenario uses less water but produces dramatically worse NOx emissions (12,000 tons/yr, double Salt Lake County industrial NOx). Journalist Jonathan P. Thompson (Land Desk) ran the more conventional textbook calculation — 9 GW × 60% capacity factor × 200 gal/MWh ≈ 9.5 billion gallons/year — and noted this “is in line with developers’ statements that they would eventually seek up to 13,000 acre-feet of water rights.” Zach Frankel (Utah Rivers Council) uses a general benchmark of ~5 million gallons/day per large data center, framing two large facilities as equivalent to the annual water use of Sandy, Utah (~40,000 people) or “a city of 100,000 people.” GreenLatinos translated the 13,000 AF portfolio into “more than enough water for over 20,000 households.”

USU physicist Robert Davies added a critical physics constraint: the 9 GW project produces 16 GW of thermal load that must be rejected somewhere. “My understanding is that they’re citing non-disclosure agreements and proprietary technology. That’s a red flag.” Either heat goes to atmosphere via dry coolers (driving local nighttime warming of 8–28 °F) or to water via evaporation (driving large consumption) — physics doesn’t allow both effects to be small. Ben Abbott (BYU / Grow the Flow) noted Utah needs to “decrease our water use by 40%” to restore the Great Salt Lake, calling resubmission “another breach of trust… all the hallmarks of an out-of-state mega-project with little to no concern for the local community.” Friends of Great Salt Lake’s formal protest specifically argued that “federal EIA data shows natural gas power plants at this scale consume hundreds of thousands of acre-feet per year. The application does not explain how 1,900 acre-feet is sufficient.”

What the engineering actually implies

Applying NREL TP-6A20-50900 (Macknick 2011) water-consumption coefficients to 9 GW at 80% capacity factor produces the following scenario table (consumptive water, generation + data-center cooling combined):

Scenario Generation tech DC cooling tech Total water/yr Acre-feet/yr Million gal/day Low (developer claims) Recip engines (radiator) DLC + dry coolers ~1.2 B gal 3,700 3.3 Medium (realistic mixed) Recip + some NGCC DLC + adiabatic assist 8–11 B gal 25,000–33,000 22–30 High (conventional) NGCC + wet tower Evaporative towers ~38 B gal 115,000 104

The Low scenario is technically plausible because reciprocating gas engines (Caterpillar G3520K, Wärtsilä, Jenbacher) use closed-loop radiator cooling consuming only “gallons per week” of makeup, and NVIDIA’s new Vera Rubin platform supports 45 °C warm-water direct-to-chip cooling that allows heat rejection through dry coolers in most climates. Joule Capital’s Millard County campus uses exactly this architecture — Cat G3520K reciprocating engines + DLC — with ~10,000 AF rights for ~4 GW. Novva’s West Jordan facility (1.5 M sq ft) used only 3 million gallons in 2024–25. Meta’s Eagle Mountain campus consumed just 35 million gallons in 2024 across over a terawatt-hour of operation, using air-side economization. Behind-the-meter recip-engine design with DLC dry-cooled heat rejection at Wonder Valley scale would plausibly consume on the order of 3,000–10,000 AF/yr — within the developer’s 13,000 AF portfolio, and 5–10× less than a conventional NGCC+evaporative design would need.

Three caveats matter. First, “closed loop” is marketing shorthand that conceals the dominant engineering choice — outer-loop heat rejection. A closed inner loop coupled to an outer wet cooling tower still evaporates large volumes. The developer has not published which outer-loop method it will use. Second, Hansel Valley summer ambient temperatures (95–100 °F) degrade dry-cooler efficiency and may force adiabatic/wet-assist mode for several weeks per year, adding 10–30% to nominal water use. Third, the 1,900 AF Salt Wells filing is sized only for the Low scenario — Friends of Great Salt Lake’s protest is correct that under any conventional design it would be wildly insufficient. This actually argues, perversely, that the developer intends the Low-water design, but the absence of any disclosed manufacturer, contract, vendor, hydrologic analysis, or air permit BACT filing means none of this is verified.

Where the public claims diverge from reality

Several specific O’Leary and project-team claims do not survive scrutiny. The “net positive to the Great Salt Lake” assertion (Morris, Cox FAQ, Palandjian, O’Leary) rests on the premise that converting historical irrigation use to industrial use reduces consumption — but USU’s David Tarboton notes that “if you use water for irrigation, only a very small fraction reinfiltrates,” meaning the assumption that retiring an irrigation right liberates water for the lake is contested and unsupported by published hydrologic analysis. The “zero water turbine” technology cited at the April 22 commission meeting has no named manufacturer, model, or contract; Grow the Flow’s Samantha Hawkins flagged that “there’s no publicly available hydrologic analysis or independent review to support those claims.” O’Leary’s claim that the project could be partially powered by solar/wind directly contradicts MIDA’s “100% natural gas” statement and the Bar H application’s own language (”power generation”). Palandjian’s admission that he cannot give a water number is fundamentally inconsistent with simultaneous project-team claims that consumption will be “similar to a large office complex” — you cannot know one without knowing the other.

The “adding water to the Great Salt Lake” framing has a narrow defensible reading (closed-loop blowdown reinjection or watershed offset accounting) and a misleading marketing reading (the data center produces water). Only the narrow reading is engineering-credible, and it has not been substantiated. The “closed-loop” framing is technically accurate for the inner cooling loop but is being used to imply zero atmospheric heat rejection — which is impossible at 16 GW of thermal load.

Bottom line

O’Leary’s claims sit in an awkward middle ground: not outright fabrications, but not yet credible disclosures either. The most likely engineering outcome — reciprocating gas engines plus direct-to-chip liquid cooling with dry coolers — would in fact consume far less water than critics’ high-end estimates (which assume conventional NGCC + evaporative towers) and would be roughly consistent with the developer’s 13,000 AF portfolio. In that sense, Frankel’s “city of 100,000” framing is a reasonable upper bound for water-rights control but probably overstates consumption; Utah Clean Energy’s 16.6 billion gallon CCCT scenario is almost certainly not what will be built. Conversely, “zero water,” “net positive to Great Salt Lake,” and “less than ranching” are not currently substantiated — they require engineering disclosures, vendor contracts, and hydrologic studies that do not yet exist publicly, and they rely on accounting choices (irrigation-to-industrial offset) that hydrologists actively dispute. The legitimate residual uncertainty is large: final cooling technology is not locked in, the air permit has not been filed, and HB 76’s reporting requirements will not produce verified numbers until July 2027 at the earliest. Until then, the most accurate characterization of O’Leary’s public claims is plausible if the project builds what it implies it will build, misleading in the framing it has chosen to communicate that intent, and unverifiable on the timeline at which Utah’s water-rights decisions are being made.

What follows is an overview of what hantaviruses are, where they came from, how they make people sick, the major outbreaks on record, and what makes the current situation unusual.

What Hantaviruses Are

Hantaviruses are a group of enveloped, single-stranded, negative-sense RNA viruses with a tripartite genome (small, medium, and large segments). They belong to the family Hantaviridae in the order Bunyavirales, and the species that cause human disease sit in the genus Orthohantavirus. More than 20 viral species have been identified in this genus.

Unlike most other bunyaviruses, hantaviruses are not transmitted by arthropods. They circulate persistently and asymptomatically in their animal hosts — historically thought to be exclusively rodents, though shrews, moles, and bats are now known to harbor their own hantavirus lineages. Each virus species tends to be tightly associated with one primary reservoir host.

Human infection produces one of two clinical syndromes, separated geographically:

• Hemorrhagic Fever with Renal Syndrome (HFRS) is the Old World disease, caused by viruses found mostly across Asia and Europe. The classic agents are Hantaan virus (carried by the striped field mouse, Apodemus agrarius), Seoul virus (Norway rats, Rattus norvegicus), Dobrava-Belgrade virus (yellow-necked field mice in the Balkans), and Puumala virus (the bank vole, Myodes glareolus, in northern Europe). HFRS centers on kidney injury, vascular leak, and hemorrhagic features.

• Hantavirus Pulmonary Syndrome (HPS), also called Hantavirus Cardiopulmonary Syndrome (HCPS), is the New World disease, found only in the Americas. Sin Nombre virus (carried by the western deer mouse) is the dominant cause in North America; Andes virus (carried by the long-tailed pygmy rice rat, Oligoryzomys longicaudatus) is the dominant cause in South America. Other New World agents include Bayou virus (rice rats), Black Creek Canal virus (cotton rats), and the New York virus (white-footed mice).

In HPS, the hallmark is microvascular leak in the lungs leading to severe pulmonary edema, acute respiratory distress, and cardiogenic shock. Without adequate treatment, most HPS deaths occur within 24 to 48 hours of the cardiopulmonary phase onset. Recovery, when it happens, can take months, with breathing problems persisting up to two years.

Discovery: From the Korean War to the Four Corners

Although the disease itself is ancient — Navajo oral tradition records mass illness events tied to wet years and rodent booms in 1918 and 1933, and Chinese, Russian, and Scandinavian medical literature describes recognizable HFRS outbreaks going back to the early twentieth century — the virus that causes it eluded scientists for decades.

The pivotal moment came during the Korean War. By the war’s end, more than 3,000 cases of hemorrhagic fever with renal syndrome had been recorded among UN forces, with an initial mortality of 14.6 percent. The disease was called Korean hemorrhagic fever, and despite extensive military and academic investigation, the cause remained unknown for more than two decades.

In 1976, Korean virologist Ho Wang Lee finally cracked it. Working at Korea University in Seoul, Lee and his team isolated what they called Hantaan virus from the lungs of striped field mice (Apodemus agrarius) trapped near the Hantan River, in the area where infections had been clustered. He went on to identify Seoul virus in urban Norway rats in 1980 and helped develop the first hantavirus vaccine, earning him the nickname “Korea’s Pasteur.”

The Western Hemisphere did not appear on the hantavirus map until 1993. That spring, several young, otherwise healthy people from the Navajo Nation along the Arizona–New Mexico border developed a mysterious, rapidly fatal respiratory illness. Within weeks, CDC and university investigators identified a previously unrecognized hantavirus that they initially called Muerto Canyon virus, then Four Corners virus, and finally — after community objections — Sin Nombre virus, “the nameless one.” Its reservoir turned out to be the deer mouse.

The 1993 outbreak was unusually lethal. Investigators eventually confirmed 33 infections in the Four Corners states, and more than half of those infected died. The trigger appears to have been ecological: an El Niño-driven wet period produced a bumper pinyon nut crop, which fueled a population boom in deer mice and pushed humans into closer contact with rodent excreta. Navajo elders had recognized the same pattern decades earlier.

How Hantaviruses Spread

For nearly every hantavirus, transmission is zoonotic and one-directional: from rodent to human. People become infected mainly through contact with infected rodents’ urine, droppings, or saliva, most often by inhaling aerosolized particles when disturbing contaminated dust in barns, sheds, cabins, or rarely-used buildings. Bites can also transmit the virus, though that route is uncommon.

Hantaviruses are surprisingly hardy outside their hosts. They can survive for around ten days at room temperature and longer in cooler conditions, which is why disturbing old droppings in an enclosed space is the classic exposure scenario. The two best-documented North American outbreaks both involved exactly that — buildings that had been heavily used by mice and then re-entered or cleaned by people.

The crucial exception to the rodent-to-human-only rule is Andes virus. Andes virus is the only hantavirus known to spread person-to-person, and even then the spread is generally limited to people in close contact with the ill person. The first signal came from a 1996 cluster in El Bolsón, Argentina, where epidemiologic and molecular evidence pointed to direct human-to-human transmission across a chain of cases that eventually reached Buenos Aires more than a thousand kilometers away.

The clearest demonstration of Andes-virus transmission came in 2018–2019 in the Patagonian town of Epuyén. The index case was a 68-year-old resident likely infected from rodent excreta near his home, who attended a 90-minute birthday party with about 100 guests on November 3, 2018. From November 2018 through February 2019, person-to-person transmission resulted in 34 confirmed infections and 11 deaths in Chubut Province. A landmark New England Journal of Medicine analysis showed that three “super-spreader” patients accounted for roughly 64 percent of secondary cases. The median reproductive number was about 2.12 before control measures and dropped to 0.96 once isolation and contact tracing were enforced.

What that outbreak also revealed is how short the infectious window seems to be. Investigators found that people are at peak infectiousness on the day they develop a fever, and the transmission window is roughly a day, though the virus can pass after only brief proximity within that window. Most of the 82 health-care workers exposed during the outbreak did not become infected, suggesting that even with Andes, sustained human-to-human spread requires close contact.

Major Outbreaks on Record

A handful of episodes stand out historically.

The Korean War HFRS epidemic (1951–1954) is the founding event of modern hantavirology — roughly 3,000 UN troops sickened, hundreds dead, and a mystery that drove decades of research.

China’s ongoing HFRS burden dwarfs everything else in absolute numbers. Between 1950 and 2007, China recorded more than 1.55 million HFRS cases and over 46,000 deaths, with the disease reported in 29 of 31 provinces. A peak year of 115,807 cases and 2,569 deaths was recorded in 1986. More than 100,000 cases of HFRS still occur each year worldwide, with China the most affected country in Asia and Finland the most affected in Europe. Hantaan and Seoul viruses are the dominant agents in China, and a national vaccination program has driven cases down sharply since the early 2000s.

The 1993 Four Corners outbreak identified the New World branch of the family and HPS as a clinical entity. Its 52–56 percent case fatality rate remains one of the highest ever observed for a viral respiratory infection.

The Yosemite outbreak (2012) was the most prominent recent point-source HPS event in the United States. The CDC ultimately confirmed ten cases of Sin Nombre infection among Yosemite visitors during the summer of 2012, with three deaths; nine of the patients had stayed in “signature” tent cabins in Curry Village. A subsequent investigation pointed to foam insulation in the cabin walls that had attracted nesting deer mice, an unintended consequence of a design upgrade meant to make the cabins more comfortable. The park notified roughly 260,000 overnight guests as a precaution.

The 2017 U.S./Canada Seoul-virus outbreak among pet-rat owners illustrated a different transmission pattern. Investigators identified 31 facilities — primarily home-based ratteries — across 11 states with human and/or rat Seoul-virus infections. Seventeen people had recent infections, eight became ill, and three were hospitalized; all recovered. The outbreak traced back to a network of rat breeders and showed that hantaviruses can travel through the pet trade, not just wild rodent populations.

The Epuyén outbreak (2018–2019), discussed above, remains the largest documented person-to-person hantavirus event.

In 2025, hantavirus briefly captured American public attention when Betsy Arakawa, the wife of actor Gene Hackman, died of hantavirus pulmonary syndrome in New Mexico in February 2025. The case was a reminder that the virus remains a low-frequency but ongoing rural risk in the U.S. Southwest. As of CDC reporting, 890 hantavirus cases had been reported in the U.S. since surveillance began in 1993 through the end of 2023.

The MV Hondius outbreak (2026) is the most recent and the most striking. The Dutch expedition ship had departed Ushuaia, Argentina on April 1, 2026, with 147 passengers and crew for an Antarctic itinerary. Sequencing of patient samples confirmed the Andes strain, and the WHO has noted that some human-to-human transmission may have occurred on board. The WHO’s director-general has said the overall public health risk from this event remains low and dismissed comparisons to COVID-19, but contact tracing across the 23 nationalities represented on board has been complex, with infected passengers turning up in South Africa, Switzerland, and elsewhere.

Smaller but historically important outbreaks include nephropathia epidemica epidemics in Scandinavia (Puumala virus in bank voles, often during years of vole population peaks) and HFRS clusters during the Balkan wars of the 1990s, when Dobrava-Belgrade and Puumala viruses sickened soldiers and civilians living in field conditions.

Mortality

Mortality varies dramatically by virus species.

• Sin Nombre / HPS in North America: roughly 35 to 40 percent on average. A CDC analysis of 510 U.S. HPS cases between 1993 and 2009 found a 35 percent case-fatality rate, with no clear trend of increasing or decreasing fatality over time. The National Park Service cites a 38 percent figure.

• Andes virus / HPS in South America: comparably severe. Case-fatality rates for Andes-virus HPS range from 21 to 50 percent, with the Argentine experience tending toward the higher end.

• Hantaan virus / HFRS: historically the most lethal Old World hantavirus, with case-fatality rates that ran around 10 to 15 percent in the Korean War era and have fallen to roughly 1 to 2 percent in modern Korea with supportive care and vaccination.

• Dobrava-Belgrade virus: up to about 12 percent mortality in the Balkans.

• Seoul virus: generally milder, with mortality on the order of 1 to 2 percent.

• Puumala virus / nephropathia epidemica: the mildest of the major HFRS agents. Puumala virus causes a relatively mild form of HFRS with a mortality rate of roughly 0.1 to 0.3 percent in Europe.

The dominant cause of death in HPS is acute cardiogenic shock during the cardiopulmonary phase; in HFRS it is acute kidney injury and circulatory collapse.

Evolution and Origins

For most of the late twentieth century, hantaviruses were thought to have co-evolved with rodents over millions of years, with each rodent lineage carrying a co-divergent virus lineage. The current picture is messier and more interesting.

Phylogenetic analysis suggests hantaviruses may have first appeared in Chiroptera (bats) or Soricomorpha (moles and shrews) before emerging in rodent species. Recent discoveries have shown that hantaviruses infect a much more diverse range of mammals than previously believed, and that cross-species transmission has played an important role in their evolution. The first hantavirus ever isolated, in fact, was Thottapalayam virus from an Asian house shrew in India in 1964 — but it was an outlier ignored for decades because the rodent-borne paradigm dominated.

The implication is that hantaviruses are an old, broadly distributed family that has repeatedly jumped between mammalian orders. Whether the human-pathogenic strains evolved their virulence specifically as a result of host-switching, or whether it is simply an accidental side effect of replication strategies adapted to the original host, is still being worked out. The genus has not changed dramatically since its discovery — there is no evidence of rapid antigenic drift comparable to influenza — but new species and host associations continue to be reported every year.

Treatment, Prevention, and Vaccines

There is no specific antiviral approved for hantavirus disease in the United States or Europe. Care is supportive: aggressive fluid management, vasopressors, mechanical ventilation, and in the most severe HPS cases, extracorporeal membrane oxygenation (ECMO), which has become an important rescue therapy. Intravenous ribavirin has been tested in HPS patients but was not shown to be effective; it has somewhat better evidence in HFRS, particularly Hantaan-virus disease, when given early.

Vaccines exist, but only for HFRS and only in two countries. Ho Wang Lee produced an inactivated whole-virus vaccine against HFRS in 1990 that has only been approved in China and South Korea due to limited evidence for its efficacy by Western regulatory standards. Approximately two million doses of inactivated Hantaan/Seoul vaccines are administered every year in China, and since 2007 these vaccines have been included in China’s expanded immunization program, contributing to a sharp decline in incidence. No HPS vaccine is licensed anywhere; DNA-vaccine candidates and monoclonal antibody therapeutics against Andes virus are in clinical development.

Prevention focuses on rodent control: sealing buildings, removing food sources, eliminating nesting habitat, and using safe cleanup techniques (wetting down droppings with bleach solution rather than sweeping or vacuuming, wearing N95 respirators and gloves, and ventilating spaces before entry).

Other Questions Worth Asking

Why is Andes virus the only hantavirus that transmits between humans? This is genuinely not well understood. The Andes-virus glycoproteins appear to bind human respiratory and endothelial cells particularly well, and infected patients can have extremely high viral loads in blood and respiratory secretions in the days around symptom onset. Patients with high viral load and liver injury were more likely than others to transmit during the Epuyén outbreak. The Epuyén strain (designated ARG-Epuyén) is genetically very close to the 1996 El Bolsón strain, suggesting that the trait is stable in certain Andes-virus lineages rather than a recent mutation.

Could a hantavirus drive a pandemic? The current scientific consensus is that the risk is low. Even Andes virus has only produced limited, local clusters; sustained human-to-human transmission has never been observed beyond a few generations of cases, and the infectious window is short. The WHO has been explicit that the MV Hondius situation is not analogous to the early phase of COVID-19. That said, hantaviruses are on the WHO’s broader list of priority pathogens for research because their case-fatality rates are high and effective therapeutics are limited.

How does climate change figure in? Most rodent-borne diseases are sensitive to weather, and hantavirus outbreaks have repeatedly been linked to wet years that increase rodent food supply. The 1993 Four Corners outbreak followed a strong El Niño; the Patagonian Andes-virus outbreaks have come during years of unusually high rice-rat populations. As climate variability increases, the conditions favoring rodent population booms — and human exposure — are expected to recur more often.

Are pets and livestock at risk? There is some evidence that dogs and cats can be infected with hantaviruses but do not develop clinical illness and do not transmit the virus to humans. The 2017 pet-rat outbreak showed that rats themselves, kept as pets, can be a meaningful exposure route — but this is a Seoul-virus phenomenon tied to the Norway rat as the natural host.

Where in the world is hantavirus a real risk? Old World hantaviruses are concentrated in East Asia (especially China and Korea), the Russian Far East, the Balkans, and Northern Europe (Finland, Sweden, parts of Germany). New World hantaviruses occur from Canada down to Argentina and Chile, with HPS hot spots in the U.S. Southwest, parts of Brazil, Paraguay, and Patagonia. African hantaviruses have been identified in shrews and bats but have not been clearly associated with human disease.

Sources

• World Health Organization, Disease Outbreak News: Hantavirus cluster linked to cruise ship travel, Multi-country, May 2026.

• U.S. Centers for Disease Control and Prevention, “About Hantavirus” and “Clinician Brief: Hantavirus Pulmonary Syndrome (HPS).”

• CDC, “Reported Cases of Hantavirus Disease,” surveillance data through 2023.

• MacNeil A, Ksiazek TG, Rollin PE. “Hantavirus Pulmonary Syndrome, United States, 1993–2009.” Emerging Infectious Diseases, 2011.

• Martínez VP et al. “’Super-Spreaders’ and Person-to-Person Transmission of Andes Virus in Argentina.” New England Journal of Medicine, 2020.

• Alonso DO et al., virological characterization of the ARG-Epuyén Andes virus strain, PLOS Neglected Tropical Diseases, 2025.

• Núñez JJ et al., “Hantavirus Infections among Overnight Visitors to Yosemite National Park, California, USA, 2012,” Emerging Infectious Diseases.

• CDC MMWR, “Outbreak of Seoul Virus Among Rats and Rat Owners — United States and Canada, 2017.”

• Lee HW et al., “Isolation of the etiologic agent of Korean hemorrhagic fever,” Journal of Infectious Diseases, 1978; obituary and biography in The Lancet Infectious Diseases, 2022.

• Zhang YZ et al., “Phylogeny and Origins of Hantaviruses Harbored by Bats, Insectivores, and Rodents,” PLOS Pathogens, 2013.

• Holmes EC, Zhang YZ, “The evolution and emergence of hantaviruses,” Current Opinion in Virology, 2015.

• Zhang YZ et al., “Hantavirus Infections in Humans and Animals, China,” Emerging Infectious Diseases, 2010.

• Chinese Center for Disease Control and Prevention, “Hemorrhagic Fever with Renal Syndrome and Diversity and Distribution of Hantaviruses — China, 2014–2023,” China CDC Weekly, 2025.

• Liu R et al., “Vaccines and Therapeutics Against Hantaviruses,” Frontiers in Microbiology, 2020.

• Mayo Clinic and U.S. National Park Service hantavirus information pages.

• Wikipedia, “1993 Four Corners hantavirus outbreak,” “Hantavirus pulmonary syndrome,” “Hantavirus vaccine,” and “MV Hondius hantavirus outbreak.”

• Time, CNN, France 24, KRQE, and Scientific American reporting on the 2025 Arakawa case and the 2026 MV Hondius outbreak.

The basic facts are not in dispute. Reporting says old online posts attributed to Blouin from roughly 2009 to 2015 included anti-Latter-day Saint insults, slurs, crude remarks about women, and comments that joked about or minimized sexual assault. Blouin did not deny writing them. He said there was “no excuse,” called the posts “vulgar” and “stupid,” said they reflected a version of himself from his early twenties that he is ashamed of, and apologized to the people he hurt. (Axios)

The backlash was immediate. Democratic leaders in the Utah House and Senate called the posts “beyond unacceptable,” and Salt Lake County Mayor Jenny Wilson said she did not believe Blouin should continue to serve in public office or seek Congress. At the same time, not every Democrat joined the pile-on. Utah Rep. Sahara Hayes said he had taken full responsibility, and Salt Lake County Councilmember Natalie Pinkney said his apology showed how much he had grown. (KUER)

That split is the whole story. No serious person has to pretend the posts were harmless. They were ugly. Voters are entitled to judge them harshly. But the public reporting so far centers on decade-old posts, followed by a public record in which Blouin became a renewable-energy advocate and then an elected official with a conventional progressive issue portfolio. That does not erase what he wrote. It does matter if the question is whether Democrats believe people can actually change. (Axios)

Bad past conduct is not the same thing as an unchanged present

This is where Democrats keep getting themselves in trouble. They often talk as if accountability has to mean permanent disqualification. But those are not the same thing. If a candidate admits the conduct, apologizes, and has spent years building a materially different public life, then the relevant question is not “Was the old conduct bad?” Of course it was. The relevant question is whether the old conduct still best describes the person now.

In Blouin’s case, the answer is at least debatable. The reporting is about old posts, not a current pattern in office. His defenders are not saying the comments were fine; they are saying the man who wrote them is not the man they know now. A party that insists growth is real in theory but meaningless in practice is not enforcing standards. It is enforcing purity. (Axios)

Purity tests are especially destructive in races like this one

The strategic problem is obvious. AP has described Utah’s new 1st District as a rare Democratic foothold in a red state and a race that could matter in the fight for House control. KUER reported that the Democratic primary has taken a “mean” turn, with analyst Leah Murray arguing there is not much policy daylight between many of the candidates, which creates an incentive to manufacture difference through mudslinging and opposition research. In other words, this is exactly the kind of race where a party can talk itself into losing by turning every moral failing into a campaign-ending event. (AP News)

That matters because Democrats are not choosing between saints and villains. They are choosing between imperfect human beings competing in a high-stakes primary. If the party’s working rule becomes “find the oldest offensive thing and declare the candidate unfit forever,” then it will keep converting winnable races into circular firing squads. Politics is not clergy selection. Coalitions are built out of flawed people, not moral finalists.

Democrats should know how to distinguish categories of wrongdoing

Blouin’s own rise in Utah politics makes the contrast clearer. He took office after defeating longtime Democratic state Sen. Gene Davis in a primary that was shaped by sexual harassment allegations against Davis. Democrats were right to take those allegations seriously. But that is precisely the point: present-day allegations of misconduct against a sitting officeholder are not the same thing as decade-old anonymous posts followed by admission, apology, and years of different public conduct. A party that cannot distinguish between those categories stops looking principled and starts looking simplistic. (KUER)

Democrats do not need fewer standards. They need better ones. A serious standard would ask five questions: how severe was the conduct, how recent was it, was it part of a continuing pattern, how did the person respond when confronted, and what has the person’s public record looked like since? That is not softness. That is judgment.

Applied here, the case is hard but not mysterious. The posts were serious. They were also old. Blouin did not deny them. He apologized. And some Democratic officials, including women who know his record now, say he has taken responsibility and changed. Reasonable voters can still decide that is not enough. But there is a big difference between saying “I am not persuaded” and saying “no amount of repentance or contrary conduct can ever matter.” The first is politics. The second is a purity test. (KUTV)

The party has to decide whether it believes in redemption or just rhetoric

This is the broader Democratic problem. The party often speaks the language of empathy, rehabilitation, and human complexity. But when one of its own becomes politically inconvenient, it can revert to a zero-defect standard that assumes the worst version of a person is the only version that counts. That may feel righteous in the moment. It is also a great way to shrink your bench, demoralize your voters, and waste rare openings like Utah’s new 1st District. (AP News)

None of this requires anyone to like Nate Blouin, excuse what he wrote, or pretend the controversy is fabricated. It requires something more difficult and more adult: proportional judgment. Democrats can believe the posts were ugly, demand accountability, and still refuse to turn every resurfaced sin into automatic exile. If they cannot do that, they will keep purity-testing themselves out of the very victories they say they need. Character matters. But if a party cannot weigh character alongside time, repentance, and the life lived since, it is not selecting the best candidates. It is selecting the least recoverable past.

• Axios

• AP News

• AP News

Blouin’s path into Utah politics was unconventional. Brown University says he grew up in New Hampshire, moved to Utah in 2009 to ski, worked odd jobs, later attended Salt Lake Community College, earned a bachelor’s degree from the University of Utah in 2017, and completed a Master of Public Affairs at Brown in 2019. Before elected office, Weber State and Future Caucus both describe him as a renewable-energy policy advocate who worked on Utah campaigns and then handled legislative and regulatory advocacy for Interwest Energy Alliance across western states. (Watson Institute)

Since taking office in 2023, he has built a public record that looks fairly conventional for a young Utah progressive. The Utah Senate Democrats’ official bio says he represents parts of Sugar House, Millcreek, Murray, West Valley City, and South Salt Lake, serves on energy- and environment-related committees, and sponsored legislation in 2023 aimed at Great Salt Lake restoration and commercial energy storage. Brown’s 2025 profile added that he was the youngest member of the Utah Senate when he took office. (Utah Senate Democrats)

He got there by beating a longtime incumbent. In the 2022 Democratic primary, Blouin defeated Gene Davis 75.89% to 24.11%, then won the general election 71.76% to 28.24%. KUER notes that Davis had been accused of sexual harassment in the lead-up to that primary, which is part of why Blouin’s own controversy now lands with extra force inside Utah Democratic politics. (KSL)

By the time he launched his congressional bid, Blouin had real momentum. He qualified for the June 23 primary ballot, remains scheduled to compete at the April 25 Utah Democratic convention, and has rolled up endorsements from Bernie Sanders, Pramila Jayapal, Ro Khanna, Jill Tokuda, Maxwell Frost, Greg Casar, and the Congressional Progressive Caucus PAC. The Associated Press described him in March as one of the prominent progressives in a district where the Democratic primary winner will likely win the general election. (KSL)

Nate Blouin controversy explained

The controversy itself is not hard to summarize. Punchbowl News reported that old posts attributed to Blouin on Reddit and other forums from roughly 2009 through 2015 included anti-Latter-day Saint insults, explicit slurs, crude remarks about women, jokes minimizing sexual assault, and comments referencing child sexual exploitation. Coverage from Axios, KUER, Deseret News, KUTV, and the Standard-Examiner all converged on the same bottom line: these were ugly, politically explosive posts that touched two especially sensitive constituencies in Utah politics, women and Latter-day Saints. (Axios)

One question that keeps surfacing in online debates is whether the posts were really his. Blouin largely settled that issue himself. KUER reported that he conceded the authenticity of the comments, and KUTV reported that he did not dispute writing them. In his public statement, he said there was “no excuse,” called the posts “vulgar” and “stupid,” said he was ashamed of them, said he would not “minimize” what he wrote, and apologized to the people he hurt. (KUER)

That apology mattered, but it did not end the fight. Utah House and Senate Democratic leaders called the comments “beyond unacceptable” and said there is no room in public service for language that demeans people or makes light of sexual assault. Salt Lake County Mayor Jenny Wilson went further, saying that even if the apology reflected genuine growth, she did not believe Blouin should continue to serve in public office or seek Congress. Former Salt Lake County councilman Jim Bradley also urged him to leave the race. (Deseret News)

Other Democrats did not join the pile-on. Utah Rep. Sahara Hayes said Blouin had taken full responsibility and argued that “the Nate of today” had proven himself as a tireless advocate for constituents. Salt Lake County Councilmember Natalie Pinkney said his apology and accountability showed how much he had grown. And in the latest Salt Lake Tribune reporting surfaced on April 16, Blouin said he had no plan to leave the race. (KUTV)

Why Utah Democrats are split

The split makes more sense once the race itself is understood. This district exists because of Utah’s Proposition 4 anti-gerrymandering fight and a court order keeping most of Salt Lake County in one congressional seat. In practical terms, that turned what would normally be a symbolic Democratic primary in Utah into the main event. AP reported that whoever wins the Democratic primary will likely win in November, and KUER described the field as a contest between a younger, more progressive wing and a more traditional set of Democrats led by former congressman Ben McAdams. (League of Women Voters)

When primaries become high-stakes and the candidates mostly agree on policy, character fights become irresistible. Leah Murray of Weber State told KUER that there is “not a lot” of policy difference among the candidates, which creates pressure to show separation by getting “mean.” She said the primary had already taken a “mean” turn, while former Utah Democratic operative Theo Gardner-Puschak said the money and attention now flowing into the race make opposition research far more likely than Utah Democrats are used to. (KUER)

That is the backdrop for the Blouin argument now spilling across social media. It is not only a story about what he wrote. It is also a story about a crowded, ideologically compressed primary in a newly winnable district, where the easiest way to create daylight between similar candidates is often to weaponize biography rather than policy. (KUER)

The deeper issue: accountability or self-sabotage?

There are two plainly defensible ways to read Blouin’s case. The first is straightforward accountability: the posts were serious enough, especially the remarks touching sexual assault and the LDS faith, that voters can reasonably decide they are disqualifying. Axios noted that the comments were especially likely to inflame women and Latter-day Saints, overlapping groups that could be decisive in a Utah Democratic primary. That is not a frivolous objection. (Axios)

The second reading is the one that makes this bigger than a single candidate. A party still needs a workable standard for growth. If the rule is that any ugly resurfaced statement means permanent exile, even when the conduct is old, admitted, apologized for, and followed by a different public record, then “accountability” starts to blur into something critics describe as Democratic self-cancellation. Blouin’s case is useful precisely because his post-forum public record is not hidden: it is a trackable one of campaign work, renewable-energy advocacy, and legislative attention to housing, air quality, and the Great Salt Lake. Whether that proves change is a judgment call, but it plainly gives voters more to examine than a pile of screenshots. (KUER)

That is also why some Democrats are resisting the idea that the story should be dispositive by itself. Pinkney’s defense of Blouin was blunt: taking accountability and apologizing, she said, showed how much he had grown. Hayes made the same case in institutional language, saying the relevant question is who he is now as a public servant. A neutral observer does not have to accept those defenses to see the larger problem they point to: a party that says people can change needs some credible rule for recognizing change when it happens. (Deseret News)

Axios observed that the story is landing as Democrats nationally try to distinguish themselves from Republicans on accountability. That instinct is understandable. But accountability without a theory of rehabilitation is where the strategic problem begins. In a race AP described as one that could help determine House control, and in a district the court’s map gave Democrats a stronger shot to win, the danger is not that Democrats have standards. The danger is that they apply them with no limiting principle, turning every opposition-research dump into a possible act of self-disqualification. (Axios)

A more durable standard would ask five questions at once: how serious was the conduct, how old was it, was it a one-period pattern or a continuing one, what has the person done since, and how did they respond when confronted? By that measure, Blouin’s case is hard precisely because it cuts both ways. The comments were serious. They were also old, admitted, apologized for, and followed by a public record that supporters say points the other direction. (KUER)

Reasonable people can still land in different places. Some will decide the posts are disqualifying because certain lines, once crossed, cannot be uncrossed. Others will decide that a decade of contrary conduct deserves real weight. But the larger lesson for Democrats is not that they should drop standards. It is that they need a serious doctrine of judgment, one that can distinguish between an ugly past and an unchanged present. Without that, a party can start mistaking moral seriousness for political self-harm, especially in the very races it most needs to win. (AP News)

Quick answers for people debating online

Who is Nate Blouin?

He is a Utah state senator from District 13, a clean-energy-focused Democrat, and a congressional candidate in Utah’s new Democratic-leaning 1st District. He studied at Salt Lake Community College, the University of Utah, and Brown, and previously worked in renewable-energy policy. (Utah Senate Democrats)

What did Nate Blouin say?

Reporting says his old forum posts included anti-LDS insults, vulgar slurs, crude remarks about women, jokes minimizing sexual assault, and comments involving child sexual exploitation. (Axios)

Did Nate Blouin apologize?

Yes. He did not dispute the posts and said there was “no excuse,” that the comments were “vulgar” and “stupid,” and that he would take full accountability. (KUER)

Is Nate Blouin dropping out?

As of the latest Salt Lake Tribune reporting surfaced on April 16, he said he had no plans to leave the race. (The Salt Lake Tribune)

Why are Democrats divided over him?

Because the seat is suddenly winnable, the primary is high-stakes, the field is split between progressive and more traditional Democrats, and analysts say there is not much policy daylight between the candidates, making personal attacks and opposition research more potent. (AP News)