Standard Metrics is an AI-driven portfolio monitoring and investor reporting platform built for VC and private equity firms. It helps automate portfolio company data collection, standardize reporting, and track performance across the fund, with 150+ institutional managers using the platform across $400B+ in assets and 10,000+ companies.

Learn more about Standard Metrics →

Quick Answer

Josh Kushner founded Thrive Capital in 2009 while still at Harvard and has built it into one of the most powerful venture capital firms in the world. Thrive manages approximately $65 billion across its venture funds and Thrive Holdings, with investments including Instagram, Spotify, Stripe, GitHub, OpenAI, Anduril, Figma, Robinhood, and Oscar Health.

Kushner’s own wealth has grown alongside the firm. His estimated net worth jumped from $5.4 billion in 2024 to $16.7 billion in 2025 — an $11.3 billion increase in roughly 12 months, driven largely by the private-market appreciation of companies Thrive had spent years accumulating positions in.

But the venture portfolio is only part of the story.

Alongside its traditional funds, Thrive has quietly built Thrive Holdings, an acquisition and operating business that controls more than 70 companies generating over $1 billion in combined annual revenue. It has turned Thrive into something that increasingly looks less like a traditional VC firm and more like a hybrid of venture capital, growth equity, private equity, and a long-term technology holding company.

Most coverage of Kushner still leads with his family name or celebrity marriage. That misses what is actually interesting. Over the past 17 years, he has built one of the strongest investment track records in technology while creating an investment platform that looks increasingly different from the rest of venture capital.

This is the operating breakdown of how Thrive actually works.

Investor Snapshot

Name: Joshua Kushner
Firm: Thrive Capital — Founder & Managing Partner
Founded: 2009, while at Harvard
AUM: ~$65B across venture funds and Thrive Holdings
Fund Count: 10+ funds across venture and growth
Notable Investments: Instagram, Spotify, Stripe, OpenAI, GitHub, Anduril, Oscar Health, Robinhood, Figma
Notable Exits: Instagram ($1B → Facebook), GitHub ($7.5B → Microsoft), Glassdoor ($1.2B), Jet.com ($3.3B → Walmart)
Estimated Net Worth: $16.7B, Forbes 2025
Background: Harvard, co-founded Oscar Health in 2012, brother of Jared Kushner

How Kushner Got Started

Kushner started angel investing as a Harvard undergraduate, writing small checks into startups while most of his classmates were interviewing for traditional finance jobs. He launched Thrive Capital in 2009 at age 24, initially as a small seed fund backed largely by family capital. The early portfolio was modestly sized, with checks reportedly in the tens of thousands of dollars into companies including Instagram, Warby Parker, and Kickstarter.

Instagram was one of the investments that put Thrive on the map. Kushner invested extremely early, before the company became one of the fastest-growing consumer products in the world, and remained an investor through Facebook’s $1 billion acquisition in 2012. The return was extraordinary, but the more important takeaway was the pattern it established: Kushner was willing to back product-obsessed founders building consumer products that looked risky at the time but obvious in retrospect.

At the same time, Kushner was becoming an operator himself. He co-founded Oscar Health in 2012, building a health insurance company in one of the most regulated and difficult industries in the country. Oscar eventually went public in 2021.

Running an operating company while managing a venture fund is unusual, and it gave Kushner a perspective most pure investors never develop. He had to think about unit economics, regulation, hiring, product development, customer acquisition, and the difference between a compelling demo and a company that can actually operate at scale.

That founder experience would become an important part of Thrive’s identity.

Investment Thesis & Strategy

Thrive’s strategy has evolved through three distinct phases. Understanding that progression is important because the Thrive of today bears very little resemblance to the seed fund Kushner started in 2009.

Phase 1: Consumer Product Picking (2009–2015)

Early Thrive was a conviction-driven seed investor. The firm wrote relatively small checks, focused heavily on consumer internet companies, and placed enormous weight on product quality and founder taste.

The portfolio from this period included Instagram, Warby Parker, Kickstarter, Oscar Health, and early exposure to what would eventually become a much broader fintech thesis through companies such as Stripe.

The edge was largely Kushner’s product instinct. He developed a reputation for actually using the products Thrive invested in and giving founders detailed feedback rather than approaching investments purely through financial models or market sizing.

The strategy was relatively simple: find exceptional founders building products people genuinely wanted to use, get involved early, and stay involved as the companies grew.

Phase 2: Multi-Stage Platform (2015–2021)

As Thrive raised larger funds, the strategy expanded. The firm moved from primarily seed investing into Series A through D, growth-stage, and eventually pre-IPO rounds.

This period produced investments in companies including Spotify, GitHub, Robinhood, Figma, Notion, and Discord. Thrive was no longer simply trying to discover unknown startups before everyone else. It increasingly had to win allocations in the most competitive private technology companies in the world.

That required a different advantage.

Kushner became known for speed. He has developed a reputation for being willing to commit to investments after a single meeting, sometimes during the meeting itself. In competitive rounds where a founder may have five or ten major firms interested, the ability to understand the company and make a decision quickly becomes a meaningful advantage.

Thrive was evolving from a small seed investor into a true multi-stage platform, but it retained a relatively concentrated decision-making structure.

Phase 3: AI Mega-Bets + Thrive Holdings (2022–Present)

The current version of Thrive is fundamentally different.

The firm is now capable of deploying enormous amounts of capital into individual companies, with OpenAI becoming the clearest example. Thrive has reportedly invested more than $1 billion into OpenAI across multiple transactions while also building significant exposure across the broader AI ecosystem.

At the same time, Thrive Holdings has created an entirely separate growth engine alongside the venture funds. Rather than simply purchasing minority stakes in high-growth startups, Holdings acquires controlling positions in established businesses and operates them over the long term.

The combination changes the economics of Thrive.

Traditional venture capital produces highly uneven, long-duration returns. A fund may wait a decade for its largest positions to become liquid, and a small number of companies typically generate the majority of the returns.

An operating holding company behaves differently. Acquired businesses generate revenue and potentially cash flow every year.

Thrive now has both.

That makes it increasingly difficult to categorize the firm as simply a venture capital fund.

Portfolio & Track Record

Thrive’s portfolio reads like a list of many of the most valuable technology companies created over the past 15 years. What is unusual is not simply the quality of the names, but the range of stages and categories where the firm has been successful.

The common thread is not really a sector or even a stage. Thrive has invested across consumer, enterprise, fintech, healthcare, defense, developer tools, and AI, from pre-launch startups through enormous late-stage rounds.

The more interesting pattern is the type of company.

Many of Thrive’s biggest winners began as products and eventually became infrastructure for their respective markets. Instagram became a default layer for photo sharing and social media. Stripe became a default payments layer for internet businesses. GitHub became core infrastructure for software development. Spotify became one of the dominant platforms for global music distribution.

Thrive has repeatedly backed products that begin as tools or applications and have the potential to become platforms.

The Thrive Holdings Shadow Conglomerate

This is the part of Thrive that gets far less attention, and it may eventually become the most important part of the entire operation.

Alongside the venture funds, Thrive runs a separate acquisition arm called Thrive Holdings. It acquires controlling stakes in businesses — often established, recurring-revenue companies — and operates them as part of a growing portfolio.

The scale is already significant: more than 70 acquired companies generating over $1 billion in combined annual revenue.

The easiest comparison is something like Constellation Software combined with a major venture capital platform. The VC funds generate returns through high-growth equity investments and eventual exits. Holdings owns operating businesses that generate recurring revenue and cash flow.

The two models complement each other particularly well. Venture returns are inherently lumpy and back-loaded. A firm can make an extraordinary investment and still wait ten years before realizing the majority of the value. An operating portfolio creates a much more consistent economic engine alongside those investments.

It also potentially changes Thrive’s relationship with fundraising.

Traditional VC firms live and die by their ability to raise the next fund. Management fees fund the organization, and carried interest creates the long-term upside. Thrive is building another source of economic value alongside that model.

That raises a bigger question: At what point does Thrive stop being primarily a venture capital firm and become a technology conglomerate that also manages venture funds?

The distinction matters because that model could give Thrive advantages that traditional venture firms simply do not have.

Fund Performance & Economics

Thrive does not publicly disclose complete fund-level returns, so there is no clean public dataset showing the performance of every vintage. But the scale of the platform, its fundraising trajectory, and the appreciation of its largest positions provide a useful picture of how the economics have compounded.

AUM: ~$65B across the broader Thrive platform
Kushner Estimated Net Worth: $16.7B
Thrive Holdings Revenue: $1B+
Holdings Companies: 70+
Recent Growth Fund: $5B

The roughly $65 billion figure puts the broader Thrive platform in the same conversation as some of the largest technology investment firms in the world. That is remarkable for an organization that was founded only 17 years ago and remains closely associated with a single founder and managing partner.

The fundraising trajectory itself tells part of the story. Thrive began with relatively small pools of capital and progressively raised larger funds as its portfolio matured and LP confidence increased. Today, the firm can compete for transactions requiring hundreds of millions or even billions of dollars.

Kushner’s personal wealth illustrates the same compounding effect at an individual level. His estimated net worth increased from approximately $5.4 billion to $16.7 billion in roughly a year, largely reflecting appreciation across Thrive and some of its largest private-company positions.

That is the venture capital power law operating at an extreme scale. You do not need every investment to become Instagram, Stripe, or OpenAI. You need enough ownership in a small number of extraordinary companies for those winners to overwhelm everything else.

Investment Style & Reputation

Kushner is not a media VC.

He rarely gives interviews, does not constantly publish investment theses on social media, and has not built the kind of public-facing media operation that has become common among large venture firms. In an industry where personal brand increasingly matters, Kushner has largely taken the opposite approach.

Instead, founder accounts of working with him tend to emphasize two characteristics: speed and product depth.

Kushner is known for moving quickly on deals and, unusually for an investor deploying growth-stage capital, spending meaningful time with the actual products. Founders have described receiving detailed product feedback from him rather than conversations focused exclusively on financial metrics, fundraising strategy, or market sizing.

The firm’s structure reflects that approach. Thrive remains relatively concentrated compared with some firms managing similar amounts of capital. It has not tried to replicate the enormous platform-service organizations built by some competitors.

The underlying bet appears to be that concentrated judgment by a relatively small group of investors with strong product instincts can outperform a much larger institution built around layers of process.

The challenge will be maintaining that model as the amount of capital grows.

The AI Paradox: Bear Thesis, Bull Portfolio

This may be the most interesting contradiction in Kushner’s current investment strategy.

Kushner has argued publicly that AI could compress venture capital returns broadly. The logic makes sense: leading AI companies require enormous amounts of capital, valuations have risen extremely quickly, and a small number of winners may capture a disproportionate share of the economic value.

If companies require billions of dollars to build and investors enter at enormous valuations, even successful outcomes may generate lower multiples than traditional early-stage venture investors expect.

Then Thrive invested more than $1 billion into OpenAI.

The contradiction disappears when you look at the strategy through the power law.

Kushner’s argument is not necessarily that AI will be a bad investment category. It is that AI could be a bad category for the average venture investor because the market may become extraordinarily concentrated.

If two or three companies ultimately capture the majority of the value, owning a broad basket of AI startups may produce mediocre returns. Owning meaningful positions in the two or three dominant companies could produce extraordinary ones.

That appears to be what Thrive is attempting to do.

The strategy is not to avoid AI because valuations are high. It is to concentrate capital in the companies Thrive believes have the highest probability of becoming foundational platforms and accept that the rest of the market may struggle.

Bearish on average AI venture returns. Bullish on the small number of companies that could capture most of the value.

Those are not contradictory positions. They are the same thesis applied at different levels.

Beyond Venture: The Expanding Empire

The broader Kushner portfolio is also becoming increasingly difficult to describe as simply venture capital.

Kushner invests through Thrive, has operating experience through Oscar Health, acquires businesses through Thrive Holdings, and has expanded into major assets outside traditional technology investing.

Most successful venture capitalists eventually diversify the wealth generated from carried interest into public equities, real estate, or other passive investments. Kushner’s approach is somewhat different. He has continued building or acquiring operating platforms around the capital.

That creates several different economic engines under the broader Thrive ecosystem. Venture investments create long-duration capital appreciation and carry. Holdings creates operating revenue and potentially recurring cash flow. Oscar provides exposure to a public operating business. Other investments create exposure to entirely different asset classes.

Rather than simply diversifying a portfolio, Kushner is diversifying the ways the portfolio itself generates value.

That could ultimately make Thrive more resilient than a pure venture platform — but it also makes the organization more complicated and raises the question of how much can remain centralized around one investor as the platform continues to expand.

What to Watch

OpenAI trajectory: This is one of Thrive’s largest and most consequential bets. If OpenAI continues to increase in value or eventually becomes a major public company, the impact on Thrive’s returns could be enormous.

Thrive Holdings scale: The acquisition arm already controls more than 70 businesses generating over $1 billion in combined revenue. If that figure continues growing, Holdings could become as strategically important as the venture funds themselves.

Next fund raise: Managing tens of billions of dollars creates a very different problem from managing an early-stage venture fund. Thrive will need to continue finding opportunities large enough to matter without sacrificing the returns that allowed it to reach this scale in the first place.

Oscar Health: Oscar remains the clearest test of Kushner’s operating judgment rather than his investing judgment. Its long-term performance adds another dimension to evaluating his overall track record.

AI concentration: Thrive’s strategy creates enormous upside if its biggest AI bets become the dominant platforms of the next decade. Concentration works in both directions, however, and a major correction in private AI valuations would have an outsized impact on the portfolio.

The Bottom Line

Josh Kushner has built something that does not have a clean analogy in venture capital.

Thrive started as a small seed investor making early bets on companies like Instagram. It evolved into a multi-stage venture platform capable of backing companies like Spotify, Stripe, GitHub, Figma, and Robinhood. It then evolved again into an enormous private-market investor capable of putting more than $1 billion into OpenAI.

Now Thrive Holdings is adding another layer entirely: acquiring and operating dozens of companies that generate more than $1 billion in combined annual revenue.

The common thread across each phase has been concentrated decision-making, strong product instincts, a willingness to make large bets when conviction is high, and an unusual ability to evolve the investment model as the firm grows.

Most venture firms that reach this scale eventually become institutions. They add partners, committees, processes, platform teams, and increasingly distributed decision-making.

Thrive appears to be making a different bet: that it can reach institutional scale without completely institutionalizing the way it invests.

That may ultimately be the most interesting part of what Kushner has built.

The question is no longer whether the strategy worked at $1 billion or $10 billion. It clearly did.

The question is whether the same model can work at $65 billion — and eventually $100 billion — without losing the concentrated judgment that made Thrive successful in the first place.

https://valueaddvc.com/blog/josh-kushner-how-thrive-capital-quietly-became-the-most-important-firm-in-tech

A quick thank you to the 5,400+ of you reading Value Add VC. It’s been really exciting to watch this grow, and over the last 90 days alone, the site has generated 30,000+ clicks and more than 7 million Google Search impressions. I really appreciate everyone who reads, replies, sends me feedback, or shares these deep dives with friends and colleagues. If you have thoughts on what I should cover next, just reply t@nyvp.com — I read everything. And if your company is interested in partnering with or sponsoring Value Add VC, you can reply here as well. Most importantly, if you enjoyed this one, please share it with someone who might enjoy it too. Thank you for being part of it

In partnership with Standard Metrics

Standard Metrics is an AI-driven portfolio monitoring and investor reporting platform built for VC and private equity firms. It helps automate portfolio company data collection, standardize reporting, and track performance across the fund, with 150+ institutional managers using the platform across $400B+ in assets and 10,000+ companies.

Learn more about Standard Metrics →

Founders love telling VCs, “We’re raising $3M, maybe we’ll do one more round, and then we’ll never need outside capital again.”

It sounds capital efficient, but to a VC it can signal that you don’t fully understand how venture works.

VCs aren’t just investing for the eventual exit. They’re investing in companies they believe can compound quickly, attract additional capital and raise subsequent rounds at significantly higher valuations. A Seed at $15M, Series A at $50M and Series B at $150M creates external validation and new pricing along the way. Those markups aren’t cash, but they matter to funds and their LPs because they demonstrate portfolio progress years before an exit.

That’s part of the venture model. Funds need a handful of companies capable of becoming enormous outcomes, and if deploying another $10M or $20M can help a company grow significantly faster, raising more capital is often the rational decision.

So when you say, “We’ll raise $3M, become profitable and never need another dollar,” you may be describing a great business. You’re just not necessarily describing the type of company venture capital was designed to fund.

There’s also a simpler problem: investors probably don’t believe you anyway. If everything works, revenue takes off and you suddenly have the opportunity to expand internationally, acquire a competitor or aggressively scale distribution, are you really going to turn down $20M at an attractive valuation because your Seed deck said you’d never raise again?

Capital efficiency is great. Profitability is great. The better pitch is optionality: we’re building efficiently enough that we don’t have to raise, but if additional capital can accelerate a massive opportunity, we’ll be in a position to use it.

That’s a much more venture-backable story.

In partnership with Standard Metrics

Standard Metrics is an AI-driven portfolio monitoring and investor reporting platform built for VC and private equity firms. It helps automate portfolio company data collection, standardize reporting, and track performance across the fund, with 150+ institutional managers using the platform across $400B+ in assets and 10,000+ companies.

Learn more about Standard Metrics →

We are currently seeing one of the biggest bifurcations in early-stage startup funding I can remember.

Over the last few weeks, I’ve had dozens of conversations with founders, VCs and family offices about what is actually “fundable” right now. The consistent takeaway is that there is no shortage of capital. There is a shortage of companies investors believe deserve venture-scale capital.

The headline funding numbers make the market look incredibly healthy. Global startup funding hit a record $510 billion in the first half of 2026, already exceeding the $440 billion invested during all of 2025. But OpenAI and Anthropic alone accounted for $217 billion, or 43% of the entire market.

In the U.S., the concentration is even more extreme. Roughly 73% of venture dollars this year have gone into billion-dollar-plus rounds, while PitchBook found that just five deals represented 73% of all U.S. venture investment in Q1.

That distortion is now happening at the earliest stages too.

More than 40% of seed and Series A investment globally in early 2026 went into $100M+ rounds, and in the U.S. it was more than half. Carta says the 95th percentile seed valuation hit $200.4 million in Q2, versus $72.2 million just one year earlier. That is a 177% increase in twelve months.

At the same time, the traditional startup fundraising market is getting harder.

Carta saw roughly the same amount of pre-seed capital deployed in Q2 as a year ago, $3.19 billion versus $3.22 billion, but spread across 11,500 instruments instead of 14,825, a roughly 22% decline in deal volume.

Crunchbase found that sub-$5 million U.S. seed deals fell roughly 20% year over year, while the part of the market that actually grew was rounds of $10 million or more.

This is the bifurcation.

On one side, investors are willing to fund the companies they believe can become category-defining businesses with enormous amounts of capital, increasingly at valuations and round sizes that would have looked like Series B or Series C financing only a few years ago.

On the other side are thousands of legitimately good startups with customers, revenue, solid products and respectable growth that are discovering those attributes alone are no longer enough.

Even the benchmarks have moved. The median U.S. seed round reached roughly $3 million last year, 3x its 2018 level, while the median Series A reached $15 million and continued moving higher in 2026. More importantly, investors interviewed by Crunchbase described the old ~$1 million ARR Series A benchmark migrating toward $2 million, $3 million or even $4 million ARR, depending on growth and category.

So the middle is getting squeezed.

A company can have $3 million of revenue, grow 50% or 70%, generate attractive margins and ultimately become a fantastic $100 million or $300 million business.

That does not necessarily make it a venture business anymore.

The venture model mathematically needs outliers. As more capital concentrates into the perceived winners, investors are increasingly underwriting around some combination of exceptional speed, enormous market size, technical differentiation, founder quality and the possibility of a multi-billion-dollar outcome.

The paradox is that it has arguably never been cheaper or easier to build a software company, while simultaneously becoming harder to convince institutional investors that the company needs venture capital.

That distinction matters.

Founders should not interpret a difficult fundraising environment as meaning their company is bad. They should ask a different question:

Is this a good business, or is this a venture-scale business?

Those used to overlap much more frequently.

In 2026, they are becoming two increasingly different things.

Capital has not disappeared.

The definition of what deserves venture capital is getting much narrower.

More from Value Add VC

A few other things worth reading from Value Add Pulse, where I track startup funding, AI, IPOs and the deals shaping the venture market.

What $67B in AI Buyouts Says About Where Value Is Moving

Stripe’s $7.5B OpenRouter deal and SpaceX’s $60B Anysphere acquisition show how aggressively strategic value is shifting toward AI distribution, infrastructure and proprietary usage data.
Read the analysis →

Rillet Hits a $1B Valuation After Just Two Years

The AI-native accounting company raised a $100M Series C, its third funding round in a year, pushing total capital raised above $200M. It is a perfect example of the other side of today’s venture bifurcation.
Read more →

Nvidia’s Moat Is Moving From Silicon to Its Balance Sheet

With $48.5B of quarterly free cash flow, a $500B GPU financing pact and billions invested across the AI ecosystem, Nvidia is increasingly using capital itself as a competitive advantage.
Read the breakdown →

Unitree Robotics Jumped 629% in Its Public Debut

The humanoid robotics company briefly reached a $66B valuation after listing in Shanghai, giving public markets one of their first major price-discovery moments for humanoid robotics.
Read more →

Value Add Pulse is updated throughout the day with the startup, VC, AI and public-market news I think is actually worth paying attention to.

Browse Value Add Pulse →

Over the last three months, ValueAddVC went from relatively little search visibility to 4.79M Google impressions, 18.1K clicks, a peak of roughly 850+ clicks in a day, and an average position of 7.5.

The biggest shift was building a system around Google Search Console that tells me what to work on next. Each week, I look for pages ranking between positions 5 and 20, high-impression pages with weak CTR, queries where Google is already testing us, declining pages, stale content, overlapping topics, and pages with weak internal linking. That gives me a much better roadmap than simply guessing what keywords to target or publishing more for the sake of it.

Once the site reached scale, the data started surfacing problems that were hard to see manually. We rewrote 92 titles based on ranking and CTR data, fixed 21 cannibalization clusters where multiple pages were competing for the same searches, found 191 orphan pages with little or no internal linking, and removed more than 500 repeated AI-style phrases across the content library.

That changed the way I think about SEO. A page sitting in position 8 with meaningful impressions is often a better opportunity than starting from zero. Two decent pages competing for the same query may be stronger when merged. A good page with weak internal linking may simply need more authority flowing into it.

We eventually created a simple framework for every page: kill, merge, update, promote, or keep.

CTR also became more nuanced. Overall CTR is now around 0.4%, even as impressions continue to climb, because a lot of the new visibility is coming from broader searches and AI-driven search experiences. Higher-intent formats still perform much better, with ranked lists around 6.8% CTR and question-led pages around 3.2%.

The system now works as a loop:

Publish → collect search data → identify opportunities → update, merge, link, or rewrite → measure → repeat.

I open-sourced the full playbook, including 14 scripts for Search Console reporting, striking-distance pages, query gaps, cannibalization, title rewrites, content audits, orphan pages, stale content, broken links, and more.

GitHub:
https://github.com/TraceCohenTech/ai-seo-playbook

If you want to dig into how it works, use the scripts, adapt them to your own site, star the repo, fork it, or build on top of it.

The biggest takeaway from going from near zero to 4.79M impressions in three months is that once content production gets cheap, the real advantage shifts to measurement, prioritization, site architecture, and iteration. That is where most of the growth came from.

Want to reach this audience?

I send this newsletter to 5,000+ founders, investors, operators, and people building in and around tech. If your company is relevant to this audience and you are interested in sponsoring a future issue, partnering on research, or getting in front of readers in a useful way, feel free to reach out.

I am especially interested in working with companies across AI, SaaS, fintech, data, developer tools, venture, and startup infrastructure.

Reply to this email if you want to discuss a sponsorship or partnership.

I analyzed 500 of the biggest startup exits since 2000, representing $10.5 trillion in combined value, and built an interactive dashboard to break down where that value came from.

valueaddvc.com/startup-exits

252 IPOs. 208 acquisitions. 36 SPACs. 37 countries. 21 sectors. 17 interactive sections.

A few things stood out.

The power law is even more extreme than it looks.

The top 10 companies account for 61% of all value. The bottom 250, half the entire dataset, account for just 3.5%.

The average top-10 company is worth 439x the average company in the bottom half. Venture outcomes are not just skewed. A very small number of companies drive almost everything.

IPOs created the overwhelming majority of value.

IPOs generated $9.0 trillion, or 86% of total value.

Average IPO value: $35.6B Average acquisition value: $3.7B

That is a 9.6x difference. Acquisitions generate liquidity, but the largest wealth creation overwhelmingly comes from companies that make it to the public markets.

The 2000–2004 founding cohort was extraordinary.

96 companies founded during that five-year period are worth a combined $6.2 trillion today, more than the next three cohorts combined.

Facebook. Tesla. Palantir. Shopify. LinkedIn.

An enormous share of modern technology value was created by companies founded within the same five-year window.

Capital efficiency matters more than total capital raised.

Splunk raised $40M → $28B Wix raised $60M → $32B WhatsApp raised $70M → $21.8B

Some of the most successful companies produced hundreds of dollars of exit value for every dollar raised, while others raised billions and ultimately exited below their final private valuations.

Solo founders performed surprisingly well.

Companies with solo founders averaged $22.4B in exit value versus $13.2B for two-founder companies.

That does not mean solo founders are inherently better, but it is an interesting result given how strongly venture investors have historically preferred founding teams.

Late-stage private valuations were often poor predictors of eventual value.

35% of companies with comparable private-round data are now worth less than their last private valuation.

That represents roughly $250B of value destruction.

The median markup from the final private valuation to exit value was only 1.2x.

The 2021 SPAC boom is clearly visible in the data.

27 of the 36 SPAC exits in the dataset occurred in 2021 alone.

Average SPAC value: $6.5B, less than one-fifth the average IPO.

A financing window opened, an enormous number of companies rushed through it, and it effectively disappeared shortly afterward.

The United States still dominates startup value creation.

359 of the 500 companies are American, and they account for roughly 90% of total exit value.

The next largest countries:

Canada: $175B Sweden: $144B Israel: $90B

Startup creation has globalized far more quickly than massive startup outcomes have.

And despite all the stories about overnight success, liquidity usually takes a decade.

Median time from founding to exit: 10 years.

Only 42 of the 500 companies reached liquidity in less than four years.

The dashboard lets you explore all of this yourself:

📊 Exit Timeline: when the biggest liquidity events happened
🗺️ Industry Treemap: where $10.5T of value is concentrated
⚡ Capital Efficiency: who turned the least capital into the most value
🍷 Golden Vintages: which founding years produced the strongest cohorts
🏆 Investor Leaderboard: which investors backed the most winners
⏱️ Speed to Liquidity: companies that exited in under four years
💰 Bootstrapped Billionaires: massive outcomes with minimal fundraising
🏢 Acquisition Map: the most active buyers across the dataset
🌐 Global Breakouts: the largest outcomes outside the US
📉 2021 SPAC Boom: what happened during the peak SPAC window
🧬 Founder DNA: how founding-team size relates to exit value
📐 Unicorn-to-Exit Markup: what happened after the final private round
🗓️ Billion-Dollar Decade: every $10B+ exit on a single timeline
📋 Full sortable dataset of all 500 companies

Every chart is interactive and the underlying data is sourced.

→ valueaddvc.com/startup-exits

Airtable raised $735 million in December 2021 at an $11 billion pre money valuation, bringing total funding to roughly $1.36 billion. The round came at the peak of the software cycle, when interest rates were near zero, public SaaS multiples were elevated and investors were willing to pay aggressively for future growth.

That valuation was real because investors invested real capital at that price, and the markups were real because venture funds used that round to value their Airtable positions. The outcome is now much lower. Bending Spoons is acquiring Airtable at a $1.285 billion enterprise value and an implied equity value of approximately $2.25 billion after accounting for nearly $1 billion of net cash.

The key numbers

• $11 billion pre money valuation in 2021
• $1.36 billion in total funding raised
• $2.25 billion implied equity value today
• $1.285 billion enterprise value
• $480 million of annual recurring revenue
• 2.7 times ARR acquisition multiple
• Approximately 80 percent decline from the 2021 equity valuation

For investors that entered near the peak, the transaction will likely produce substantial losses. Earlier investors may still generate positive returns, particularly if they invested at much lower valuations or sold shares in later secondary transactions. The exact amount of prior liquidity is not publicly disclosed.

The broader story is that Airtable was only one investment inside much larger portfolios. Thrive also backed Slack, Robinhood, Nubank, Instacart, Figma and Wiz. Benchmark backed Uber, Asana and Confluent. CRV backed DoorDash, Ring and Zendesk. ICONIQ backed GitLab, Samsara, Klaviyo, Rubrik and ServiceTitan. Greenoaks backed Robinhood, Toast, Nubank and Figma. Coatue backed Snowflake, DoorDash, Reddit, Instacart and CoreWeave.

Those outcomes help explain why Airtable’s investors continued to raise significant capital even as the company’s paper valuation ultimately failed to hold.

Capital raised since Airtable’s 2021 round

• Thrive: more than $18 billion
• ICONIQ: $5.75 billion
• Greenoaks: $4.6 billion
• Benchmark: roughly $2.4 billion
• CRV: roughly $2.25 billion
• Coatue: roughly $1.7 billion in traditional private market vehicles
• WndrCo and Caffeinated: more than $620 million combined

In total, Airtable’s major investors raised more than $35 billion after the company’s 2021 financing.

LPs were not underwriting Airtable alone. They were evaluating complete portfolios, realized distributions, public holdings, remaining private positions, access to future deals and each manager’s long term track record.

The Airtable acquisition now converts a large unrealized markup into a real transaction value. It will hurt some funds more than others, but it does not define the performance of the firms involved. Venture capital is built around uneven outcomes, where a small number of major winners can offset companies that exit below their peak valuations.

Airtable’s markdown matters. Portfolio construction matters more.

The money’s real. The cash isn’t. Every week I dig into the numbers that actually moved the market — not headlines, the underlying data. Here’s what stood out.

Let’s get into it. — Trace

Value Add VC Research

AI Just Took a Majority of European VC Dollars for the First Time

European startups raised more in H1 2026 than a year earlier, but the number of companies actually getting funded fell to a six-year low as AI absorbed the gains.

AI absorbed 60% of European VC dollars in H1 2026 — its first-ever majority share

The Numbers

• H1 rebound in dollars: European startups raised €44.1 billion in H1 2026, a 27% jump from the same period a year earlier.

• But fewer companies funded: the number of companies receiving investment fell to its lowest level in six years, even as total dollars rose.

• AI crossed a majority: AI absorbed 60.3% of all European venture deal value in H1 2026, up from 37.9% across all of 2025 — the first time AI has taken an outright majority of a six-month reporting period.

• AI already ahead of last year: of the €44 billion raised, €26.5 billion went to AI companies, already surpassing all of 2025’s full-year AI total.

• Mega-rounds crowd out the rest: deals over €100 million accounted for more than half of Q2 2026 deal value, up from roughly 37% in 2025.

✈️ Key Takeaways

For early-stage European founders outside AI, this is a real warning sign: capital breadth is drying up, and non-AI companies are fighting for a shrinking slice of a growing pie. GPs raising European vehicles right now should expect LPs to press hard on deal-count risk feeding into 2027-2028 fund vintages — fewer companies funded today means a thinner pipeline of graduates two years out. If you’re building outside frontier AI in Europe, the check that would have shown up easily in 2023 isn’t there anymore; reposition the pitch or look elsewhere for capital.

Value Add VC Research

AI Took 86% of Every US Venture Dollar Raised in H1 2026

AI companies absorbed the overwhelming majority of US venture capital in the first half of 2026, while funding to non-AI startups actually shrank.

AI captured 86% of all US venture dollars in H1 2026

The Numbers

• US concentration: AI startups captured roughly 86% of all US venture dollars raised in H1 2026.

• Global share close behind: globally, AI companies took in the range of 70-80% of total VC dollars over the same period, depending on how broadly “AI company” is defined.

• Q1 alone hit $242B: in Q1 2026, AI funding totaled roughly $242 billion, about 81% of that quarter’s entire global venture total.

• Non-AI funding actually fell: funding to non-AI startups declined by roughly 10% year-over-year even as total venture volume set records.

• One company, one round: a single AI lab raised $65 billion in one round in Q2 2026 — on its own larger than most countries’ entire annual venture totals.

✈️ Key Takeaways

At 86% of US dollars, “AI startup” isn’t really a category anymore — it’s most of the asset class. For angels and early-stage funds, real diversification now requires deliberately writing non-AI checks, because the market isn’t doing that work for you anymore. The bigger risk isn’t that AI itself is overvalued — it’s that everything adjacent to it, from non-AI SaaS to consumer to fintech, is starved of capital regardless of its own fundamentals.

Value Add VC Research

The Biggest IPO Ever Is Now Trading Below Its Own Debut Price

SpaceX’s June 2026 IPO briefly created over $2 trillion in market value at a 19% first-day pop; seven weeks later the stock trades below where it priced.

SpaceX is trading roughly 20% below its own $135 IPO price

The Numbers

• Largest IPO in history: the June 2026 offering priced at $135 a share, raising $75 billion at a $1.75 trillion valuation — the largest public offering ever.

• Big first-day pop: shares opened around $150, an 11% gain over the IPO price, and market cap briefly topped $2 trillion.

• Then it kept climbing — briefly: the stock peaked intraday at $225.64 on June 16, days after the debut.

• Now it’s underwater: as of August 2, 2026, shares trade around $108 — roughly 20% below the original $135 IPO price and over 50% off the June 16 peak.

• Broader listing pickup: 113 tech IPOs priced through May 2026, up over 10% year-over-year, raising $34.2 billion — up 164% from the same period a year earlier.

• Bigger names still waiting: both OpenAI and Anthropic have confidentially filed for IPOs, either of which could dwarf SpaceX’s offering if it prices in 2026.

✈️ Key Takeaways

The largest IPO in history is a clean case study in why a single trading day is a terrible valuation anchor — SpaceX topped $2 trillion intraday, then gave nearly all of it back within seven weeks. Any LP or fund marking a pre-IPO SpaceX stake off the peak comp just took a real markdown, whether or not it’s reflected on paper yet. Expect OpenAI and Anthropic’s eventual IPOs to get priced, and re-priced, just as violently once they hit the public market.

Value Add VC Research

LPs Have Gotten Back 8 Cents for Every Dollar They Put In

2021-vintage VC funds show a median DPI of just 0.08x even as paper valuations hold steady, revealing how little of the 2026 boom has turned into actual cash.

0.08x median DPI for 2021-vintage VC funds in 2026

The Numbers

• The DPI gap: 2021-vintage VC funds show a median DPI of just 0.08x — 8 cents in actual cash returned per dollar invested.

• Paper looks fine: market-wide TVPI has stabilized near 1.1x, meaning most of the reported value exists only as unrealized markups, not cash.

• Exits are concentrated: stripping out the five largest 2026 exits cuts total exit value by more than 70%.

• Secondaries as the release valve: the LP secondary market hit roughly $120 billion in H1 2026 volume, up about 20% year-over-year.

• But at a discount: GP-led continuation vehicles now account for over half of that secondary volume, while LP-led stake sales often clear at 40-60% discounts to reported NAV.

• Vintage divergence: 2022-vintage funds, which deployed at valuations 40-60% below 2021 peaks, are outperforming 2021 vintages by an estimated 20-30% at the same stage.

✈️ Key Takeaways

0.08x DPI on a 2021 vintage means most LPs backing that era haven’t gotten real money back — just a nicer number on a capital call statement. If you’re an emerging manager pitching a new fund right now, expect every LP conversation to start with “when do we actually see cash,” not IRR. The secondary market clearing at 40-60% discounts is the honest price of that illiquidity, not the TVPI the fund is currently showing.

Read more on Value Add VC →

Value Add VC Research

Crossover Funds Set a Record, Then Six Firms Ate the Whole Thing

Hedge fund and crossover deal value hit a record $220.9 billion in Q1 2026, more than four times the 2021 peak, but concentrated in a fraction of the deals — and by Q2, most of those firms had stopped writing new checks.

$220.9B in Q1 2026 crossover deal value, spread across just 178 rounds

The Numbers

• Record dollars: crossover deal value hit $220.9 billion in Q1 2026, more than 4x the $50.3 billion 2021 peak.

• Way fewer deals: that record volume was spread across just 178 rounds, versus 425 rounds at the Q4 2021 peak.

• Then it stopped: hedge funds pulled back sharply on new venture deals in Q2 2026 as scrutiny over AI valuations increased.

• One fund manager shrank 82%: a major crossover investor’s newest fund closed at just $2.2 billion, down 82% from the $12.7 billion predecessor fund raised in 2022.

• One firm stayed active: a single crossover fund closed roughly a dozen venture deals in Q2 2026 alone and co-led both of the year’s largest AI mega-rounds.

• Where the money went: one AI company alone absorbed a $65 billion round during that same pullback quarter, at a $965 billion valuation.

✈️ Key Takeaways

The headline crossover number is a record, but the real story is that six firms are now writing the checks a much wider set of hedge funds wrote in 2021. For founders raising a Series C or later, crossover capital isn’t a reliable fallback the way it was three years ago unless you can credibly claim frontier-AI status or a near-term IPO story. The firms best positioned built direct relationships with two or three actual winners rather than spreading smaller checks across the field — a model most funds simply can’t replicate.

Read more on Value Add VC →

Read Value Add Pulse →

By Trace Cohen · @Trace_Cohen · t@nyvp.com

Good morning —

Priced like it’s over. It isn’t. Every week I dig into the numbers that actually moved the market — not headlines, the underlying data. Here’s what stood out.

Let’s get into it. — Trace

Value Add VC Research

VC deal count hit a decade low while dollars hit a record

Q2 2026 venture funding topped $200 billion for a second straight quarter, but the number of individual deals fell to its lowest level in roughly ten years.

81% of H1 2026 venture dollars went into $100M+ mega-rounds

The Numbers

• Record half-year total: global startup funding hit $510 billion in H1 2026, already ahead of all of 2025’s full-year total of $440 billion

• Deal count cratered: the number of individual venture deals fell to its lowest level in roughly a decade, even as total dollars invested kept climbing

• Mega-rounds took over: deals of $100 million or more captured roughly 81% of all venture dollars deployed in the first half of 2026

• Sixteen billion-dollar rounds: sixteen rounds of $1 billion or larger accounted for $108.6 billion of Q2’s $205 billion total on their own

• Asia’s growth streak: Asia-based startups raised $50.8 billion in Q2 2026, a fifth consecutive quarter of growth for the region

• US still dominates: US startups pulled in $144.9 billion in Q2 2026 across roughly 3,644 deals, the largest single-country share of global volume

✈️ Key Takeaways

The gap between the headline number and the median founder’s raise has never been wider. A decade-low deal count means fewer new relationships are being started industry-wide, not just fewer checks written — if you’re not in the mega-round lane, you’re competing for a shrinking pool of actual decisions. For GPs, this is the environment to underwrite selectively: capital is concentrating hard, so returns outside the AI mega-deal lane will need real differentiation, not a rising tide.

Value Add VC Research

Startup M&A just had its biggest quarter ever — $113B in one wave

24 startups sold for $1 billion or more in Q2 2026, the largest quarter for startup M&A on record, headlined by the single biggest startup acquisition in history.

$113 billion in startup M&A in a single quarter — a record

The Numbers

• Record M&A quarter: 24 startups were acquired at price tags of $1 billion or more in Q2 2026, worth a combined $113 billion — the largest quarter for startup M&A on record

• Biggest startup acquisition ever: the quarter’s largest deal alone was valued at $60 billion, the biggest startup acquisition in history

• Old record nearly doubled: the prior largest startup acquisition on record was a $32 billion enterprise security deal — the new record is close to 2x that

• IPOs surged in parallel: 32 venture-backed companies went public above a $1 billion valuation in the same quarter

• Strongest exit market since 2021: combined IPO and M&A activity in Q2 2026 marked the strongest quarter for startup exits since the 2021 boom

✈️ Key Takeaways

A real exit market changes the DPI math for every fund sitting on aging vintages — distributions have been the industry’s weakest spot for years, and this is the first quarter in a while where the exit data actually backs up the optimism. For founders, the size of the record-setting acquisition is a reminder that strategic buyers are now willing to pay up for category leaders rather than wait for an IPO. Don’t assume the pace holds, though — a handful of outsized deals are doing most of the lifting here, same as on the funding side.

Value Add VC Research

Four AI mega-rounds this spring totaled $87 billion in six weeks

A frontier model lab, an industrial-engineering AI startup, and a defense-tech company each closed billion-plus rounds within weeks of each other in spring 2026.

One AI company’s valuation hit $965B just three months after a rival hit $852B

The Numbers

• A $65B round in six weeks: a leading AI lab closed a $65 billion round in May 2026 at a $965 billion post-money valuation, overtaking a rival’s prior $852 billion mark from March

• Revenue kept pace: that company’s annualized run-rate revenue hit $47 billion in May 2026, up from $9 billion just five months earlier

• A Bezos-backed Series B: an industrial-engineering AI startup raised $12 billion in June 2026 at a $41 billion valuation

• Defense tech kept raising too: a leading defense-tech company raised $5 billion in May 2026, doubling its valuation to $61 billion in under a year

• And it’s still climbing: that same defense-tech company is reportedly now in talks for a valuation near $100 billion, more than 3x last year’s mark

• Two companies, 43% of the world: two frontier AI labs together absorbed 43% of every venture dollar raised globally in the first half of 2026

✈️ Key Takeaways

When two companies eat 43% of global venture dollars, the rest of the market isn’t really competing in the same asset class anymore — it’s a separate, much smaller pool of capital. For early-stage investors, that’s actually a reason for optimism: the mega-round capital isn’t coming for your seed deals, it’s chasing a handful of balance sheets big enough to absorb $10B+ checks. The real signal to watch is the revenue growth curve, not the valuation — a company tripling run-rate revenue in five months is a very different bet than one just riding multiple expansion.

Value Add VC Research

Harvey AI tripled its ARR in nine months selling seats, not APIs

Harvey AI generated $300 million in ARR as of May 2026, up from $195 million at year-end 2025, almost entirely through per-seat licenses sold to law firms and corporate legal departments.

Harvey AI’s ARR tripled from $100M to $300M in nine months

The Numbers

• ARR nearly doubled in five months: Harvey generated roughly $300 million in ARR as of May 2026, up from $195 million at the end of 2025 and $100 million in August 2025

• Per-seat pricing model: mid-market law firm pricing runs roughly $1,000-$2,000 per attorney per month with 20-seat minimums, scaling down to $100-$200 per user at large enterprise firms

• 1,500+ customers: as of March 2026, Harvey serves more than 1,500 customers across 60+ countries, including roughly 1,000 law firms and 500+ corporate legal departments

• Half the Am Law 100: Harvey cites coverage of roughly half of the Am Law 100, the ranking of the 100 highest-grossing US law firms

• Valuation vs. revenue: an $11 billion valuation on $300 million in ARR puts Harvey at roughly a 37x revenue multiple, high even by 2026 AI-software standards

✈️ Key Takeaways

Harvey’s model is the one worth studying if you’re evaluating vertical AI bets: no self-serve motion, no API, just enterprise seats sold into a spending category that’s enormous and price-insensitive. That’s a very different risk profile than usage-based AI startups whose revenue can evaporate the moment a model gets cheaper. For GPs pricing vertical AI deals right now, the multiple here is a useful anchor — 37x ARR is what the market will pay for a company that’s positioned itself as the default layer inside a sticky, high-value workflow.

Read more on Value Add VC →

Value Add VC Research

Kalshi and Polymarket are priced like the winner is decided. The data disagrees.

Kalshi is valued at $22B and Polymarket at $15B, but June trading volume and revenue data point in a different direction than either valuation suggests.

Kalshi did $31.5B in June volume — nearly 3x Polymarket’s $10.8B

The Numbers

• Kalshi’s valuation doubled in five months: Kalshi is valued at $22 billion after a $1 billion Series F in May 2026, double its $11 billion valuation from a Series E just five months earlier

• Polymarket’s valuation jumped 67%: Polymarket is raising $400 million at a $15 billion valuation, up 67% from the roughly $9 billion valuation implied by a strategic investor’s initial stake in October 2025

• Volume gap is wide: Kalshi did $31.5 billion in trading volume in June 2026, up 87.4% month over month, versus Polymarket’s record $10.8 billion — roughly a third of Kalshi’s total

• Category grew fast: combined monthly volume across both platforms hit $44.8 billion in June 2026, up from a niche crypto curiosity about eighteen months earlier

• Kalshi’s revenue run: Kalshi has generated roughly $1.15 billion in total fee revenue since launch, with about $850 million of that in 2026 alone, much of it from sports contracts

• Kalshi eyeing a much bigger round: Kalshi is reportedly in talks for a new round targeting a $40 billion valuation as soon as Q3 2026 — nearly 2x its current mark

✈️ Key Takeaways

Both companies are priced as if the prediction-markets category has already picked a winner, and the volume data says it hasn’t — Kalshi leads by a wide margin today, but regulatory positioning can flip fast in a category this new. For anyone underwriting either name at these marks, the real bet isn’t on current volume, it’s on which platform’s regulatory structure holds up as the category matures past a sports-betting boom cycle. Sports-driven volume spikes are real revenue today, but they’re also the least defensible kind — worth discounting accordingly in any valuation model.

Read more on Value Add VC →

Read Value Add Pulse →

By Trace Cohen · @Trace_Cohen · t@nyvp.com

Let’s get into it.

Value Add VC Research

AI took 86% of every US VC dollar in a record $412.7B first half

US venture funding hit a record $412.7 billion in H1 2026, but the dollars are more concentrated in AI mega-rounds than in any prior half-year on record.

86% of every US venture dollar in H1 2026 went to AI startups

The Numbers

• Record half-year total: US startups raised $412.7 billion in H1 2026, the largest half-year total on record

• AI’s share of dollars: AI companies captured roughly 86% of every US venture dollar deployed in the first half

• Q2 alone: Q2 2026 investment totaled $137.2 billion in North America

• Mega-round concentration: seven rounds above $1 billion closed in Q2 2026, totaling $87.2 billion, and five of the seven went to AI companies

• Deal count fell: overall deal volume dropped to its lowest point in roughly a decade even as dollar totals hit new highs

• Global total: worldwide startup investment hit a record $510 billion in H1 2026

✈️ Key Takeaways

The gap between the headline number and the median founder’s experience has never been wider. If you’re not raising an AI mega-round, the dollars in these stats aren’t your dollars — deal count at a decade low means fewer new relationships are being started industry-wide, not just fewer checks written. For GPs, this is the environment to underwrite selectively: capital is chasing a narrow set of outcomes, so returns outside that lane will need real differentiation, not a rising tide.

https://valueaddvc.com/pulse/venture-half-ai-concentration-record-2026

Value Add VC Research

Traditional IPOs raised 7x more in H1 2026 than the same span last year

The US IPO market reopened hard in the first half of 2026, with proceeds and deal count both climbing well past 2025’s pace.

$114.1B raised via traditional IPOs through June 2026 — 7x 2025’s pace

The Numbers

• Traditional IPO proceeds: US IPOs raised roughly $114.1 billion through June 30, 2026, more than 7x the $14.8 billion raised over the same span in 2025

• SPAC issuance also up: 118 SPAC IPOs raised about $20.9 billion in H1 2026, versus 66 SPACs raising $11.8 billion in H1 2025

• Deal count climbed: IPO count rose roughly 10% year-over-year through the first half

• Broad-based sectors: deal flow spanned AI infrastructure, industrials, healthcare, and consumer names, not just one sector

• Record single listing: the half’s standout debut priced at a valuation of just under $2 trillion, the largest IPO debut ever

• Second-best half on record: H1 2026 is already tracking as the second-most-active half-year on record for US IPO proceeds

✈️ Key Takeaways

A reopened IPO window changes the calculus for every fund sitting on aging vintages — distributions have been the industry’s biggest weak spot, and a real exit market is the only thing that fixes DPI. For founders, this is the first time in years it’s worth actually modeling an IPO path rather than assuming M&A or a secondary sale is the only realistic exit. Don’t over-read the mega-listing at the top, though — the broad-based sector mix is the more durable signal than any single record-setting debut.

https://valueaddvc.com/ipo-wave-2026

Value Add VC Research

90 new unicorns were minted in H1 2026, more than double last year’s pace

Unicorn formation more than doubled year-over-year in H1 2026, and AI startups are commanding a clear valuation premium at every stage behind the surge.

90 new unicorns minted in H1 2026 — more than double last year’s pace

The Numbers

• Unicorn formation doubled: roughly 90 new unicorns were minted globally in H1 2026, more than double the same period in 2025

• AI-native dominance: AI-native startups made up the clear majority of the newly minted unicorn cohort

• Top new valuation: the highest-valued new entrant in the cohort reached a $41 billion valuation

• Seed-stage AI premium: AI startups now command a median $17.9 million seed pre-money valuation, versus roughly $16 million across all sectors

• Premium size: that puts AI seed valuations about 42% above non-AI peers at the same stage

• Premium persists to Series B: median AI company valuations reach roughly $143 million by Series B

✈️ Key Takeaways

A 42% seed premium for AI companies means the valuation gap isn’t closing as rounds mature — it’s compounding. If you’re writing seed checks into AI at these prices, your ownership math needs to assume the premium holds (or grows) through Series B, not fades. For founders outside AI, this is real evidence the market is pricing your category differently, which is worth naming directly in your own pitch rather than pretending the comp set is neutral.

https://valueaddvc.com/pulse/unicorns-2026-record-pace

Value Add VC Research

Crossover deal value hit $220.9B in Q1 2026 — 4x the 2021 peak, in 178 rounds

Our analysis of crossover investor activity found record Q1 2026 deal value alongside a near-record-low deal count, meaning capital is concentrating rather than broadening.

$220.9B in Q1 2026 crossover deal value — 4x the 2021 peak, spread across just 178 rounds

The Numbers

• Record deal value: crossover deal value hit $220.9 billion in Q1 2026, more than 4x the prior 2021 peak of $50.3 billion

• Deal count stayed low: just 178 rounds closed in Q1 2026, versus 425 in Q4 2021 at the height of the last crossover cycle

• Share of the market slipping: crossover firms’ share of total US VC deal value fell from 42% in 2024 to 38% in 2025

• Two rounds, most of the total: two AI mega-rounds together raised roughly $140 billion in February 2026 alone, eating most of the quarter’s crossover dollars

• AUM pulled back: one major crossover fund’s assets under management sit at $50-60 billion in 2026, down from a roughly $95 billion peak

• Newest vehicle shrank 82%: that same fund’s newest private vehicle closed at about $2.2 billion, an 82% drop from its $12.7 billion 2022 fund

✈️ Key Takeaways

Crossover capital isn’t broadly back — it’s parked in a handful of AI names, and the deal-count collapse is the tell. If you’re a founder outside frontier AI hoping a hedge fund writes your next check, this data says don’t count on it; that capital pool has never been more selective. For GPs, the shrinking fund sizes among the crossover set is worth watching too — it’s a leading indicator of how much dry powder is actually left for late-stage rounds outside the AI mega-deal lane.

Read more on Value Add VC →

Value Add VC Research

87% of LPs have rejected a fund manager over due diligence alone

Our review of LP fund-selection criteria found DPI has replaced IRR as the first metric checked, with distributions stuck near 14-15% of NAV since late 2022.

87% of LPs have rejected a fund manager over due diligence concerns alone

The Numbers

• Distributions stalled: the average VC fund has distributed back just 14-15% of NAV to LPs since late 2022

• DPI beats IRR as gatekeeper: DPI is now the first number institutional LPs check, ahead of IRR or TVPI, because of that stalled distribution rate

• GP commitment doubled: median GP commitment for first-time funds reached 3.0% of fund size in 2026, up from 1.5% in 2020

• Due-diligence rejections: an estimated 87% of LPs have rejected a manager over operational due diligence alone, even with a strong thesis and returns

• Re-up rate spread: top-quartile funds see 80-90% LP re-up rates versus roughly 45% for bottom-quartile funds

• Concentration among established firms: established firms captured 90.9% of total VC fundraising dollars in Q1 2026, leaving a thin slice for new managers

✈️ Key Takeaways

If you’re raising a fund right now, your GP commitment and operational infrastructure are getting scrutinized before anyone even looks at your thesis — that’s a real shift from five years ago. Emerging managers should read the 90.9%-to-established-firms number as the actual competitive landscape, not the friendlier-sounding fundraising narratives. And if your last fund’s DPI is thin, get ahead of that in the pitch; LPs are actively screening for it now, not asking about it as an afterthought.

Read more on Value Add VC →

So what are your thoughts on the current tech startup ai world!? Let me know, please back!

By Trace Cohen · @Trace_Cohen · t@nyvp.com

So I (with a lot of AI agents) analyzed all 1,481 episodes spanning 11 years.

https://valueaddvc.com/20vc

How I did it:

• Pulled the complete RSS archive dating back to 2015.

• Downloaded and transcribed ~180 older episodes with Whisper where transcripts didn’t exist.

• Analyzed 2.3 million words.

• Had every episode independently scored by two AI reviewers (blind to each other’s ratings) across preparation, pushback, depth, and tone.

• Separately analyzed 49 weekly roundtable episodes with Jason Lemkin and Rory O’Driscoll.

What stood out

• Harry’s assertiveness more than doubled, rising from 2.8/10 (2015) to 6.2/10 (2026). Early episodes featured almost no pushback; today he regularly challenges guests’ assumptions.

• His warmth never changed. Across every year, reviewers consistently described him as warm, curious, and conversational. He became tougher without becoming colder.

• His questions got shorter. Early interviews averaged 25–30 word setups. Today they’re closer to 12 words. Confidence is brevity.

• The bigger the guest, the softer the pushback. Rishi Sunak scored 3/10, Marc Andreessen 4/10, while founders and operators often received much stronger questioning.

• 2022 was the inflection point. Episodes expanded from roughly 25 minutes to 73 minutes, titles became multi-part theses, and the show evolved far beyond its original format.

• The guest mix shifted with the industry. The show went from mostly venture capitalists to founders, operators, and AI CEOs as technology itself changed.

• Even his vocabulary evolved. “Incredible” became less common, while “dude” went from virtually nonexistent to a regular part of his interviews.

The project also includes:

• 930 searchable insights from 147 guests.

• 171 predictions from the Lemkin/O’Driscoll roundtable.

• 33 documented Jason vs. Rory debates.

The biggest takeaway

The most meaningful change wasn’t Harry’s personality.

It was preparation.

As his archive grew, he became dramatically better at connecting ideas across hundreds of prior conversations and pressure-testing guests in real time. Your accumulated knowledge compounds—and your archive can become one of your biggest competitive advantages.

Explore the full interactive analysis:

https://valueaddvc.com/20vc

Read Value Add Pulse →

By Trace Cohen · @Trace_Cohen · t@nyvp.com

Also a personal angel investment from 2018 just released their Fi Ultra - the first dog tracker that runs on the Starlink network. It’s amazing and every dog needs it.

It is a talent, morale, budget, incentive, infrastructure, data, and organizational design problem all happening at the same time.

The largest public companies are being told they need to completely reinvent themselves around AI while many are simultaneously dealing with slower growth, falling valuations, layoffs, tighter budgets, burned-out employees, decades of technical debt, and organizational structures that were never designed for this kind of transformation.

We keep asking why enterprises are struggling to adopt AI.

The better question might be: How could they not be struggling?

The best talent is leaving.

The best engineers, product managers, and technical operators have more opportunities than ever. They can join AI labs, fast-growing startups, well-funded private companies, or start companies themselves.

Why spend the next five years modernizing a 20-year-old enterprise software stack when you can work at the frontier of AI with better compensation, more equity upside, and potentially more interesting problems?

Companies cannot retain the people they desperately need.

Many public software companies have seen their valuations fall significantly from their 2021 highs. That means employee equity is worth less, promotions are harder to get, compensation budgets are tighter, and the perceived upside of staying has declined.

The exact people needed to lead an AI transformation are often the people with the most attractive opportunities to leave.

They may have fired the people who understood how everything worked.

Companies have spent the last several years cutting 10%, 20%, 30%, and sometimes significantly more of their workforce.

Layoffs look clean on a spreadsheet. Institutional knowledge does not.

The person who understood the obscure integration between two critical systems may be gone. The engineer who built the original architecture may have retired. The product manager who understood why a workflow was designed a certain way may have been laid off three reorganizations ago.

Now the company wants an AI agent to automate the entire process.

The remaining employees have more work and less incentive.

Companies cut headcount and distribute the remaining responsibilities across fewer people.

Employees are managing larger workloads, more systems, more meetings, and more responsibilities, often without significantly more compensation or upside.

Then management introduces another priority: AI transformation.

The people responsible for reinventing the company are frequently the same people already struggling to keep the existing company running.

There is an agency problem.

Who actually benefits from successfully deploying AI?

The employee who automates a workflow may eliminate part of their own job, reduce the size of their team, or create significantly more responsibility for themselves.

Meanwhile, the financial benefits accrue primarily to shareholders.

Companies want employees to aggressively automate their own organizations without always creating meaningful incentives for employees to do it.

Budgets are moving in the wrong direction.

Real AI transformation requires investment.

Companies need better infrastructure, cleaner data, new tooling, technical talent, security, governance, experimentation, and time.

But many of the companies most desperate for AI-driven productivity are simultaneously cutting costs.

They are trying to fund a technological transformation while operating under an austerity program.

The technical stack is too old.

A significant amount of enterprise infrastructure was built years or decades ago.

These systems were not designed for AI agents, real-time data access, modern APIs, or automated workflows that operate across multiple departments.

AI may be advancing exponentially.

Enterprise infrastructure is not.

Technical debt compounds everything.

Every old integration, undocumented API, manual workaround, abandoned database, duplicate system, and temporary fix becomes another obstacle.

Technical debt was manageable when humans were navigating these systems individually.

It becomes a much larger problem when companies want intelligent agents to autonomously operate across them.

The data is fragmented, inconsistent, and poorly prepared.

Enterprise AI is ultimately dependent on enterprise data.

Unfortunately, that data is often spread across dozens or hundreds of systems, stored in inconsistent formats, owned by different departments, governed by different permissions, and maintained at dramatically different levels of quality.

The AI model may be incredibly intelligent.

It still cannot reliably answer a question if the company itself does not know where the correct information lives.

The organization is completely siloed.

The most valuable AI workflows frequently cross organizational boundaries.

A customer workflow might involve sales, finance, legal, customer support, product, and engineering.

But companies are still organized into departments with separate budgets, separate leadership, separate systems, separate incentives, and sometimes directly conflicting priorities.

Agents operate across processes.

Enterprises operate across silos.

That is a fundamental mismatch.

Nobody actually knows how the entire company works.

This may be one of the most underestimated problems.

Large enterprises are enormously complicated systems built incrementally over decades.

No single person understands every database, workflow, integration, permission structure, business rule, and internal dependency.

The people who built the original systems may have left years ago. Documentation is incomplete. Institutional knowledge is fragmented across thousands of employees.

Before an AI agent can automate the company, someone first has to figure out how the company actually works.

They know they need better talent, but the talent may not want to work there.

Every company now says it needs AI engineers.

But the supply of genuinely talented AI builders is limited, and those people have options.

The companies most in need of transformation are often the least attractive destinations for the people capable of leading it.

Even when the right people apply, companies may not know how to hire them.

Traditional enterprises are still hiring against traditional job descriptions, compensation structures, interview processes, and organizational charts.

AI transformation requires unusual combinations of engineering, product, data, operations, domain expertise, and organizational influence.

Many companies do not even know what they are looking for.

Leadership wants AI outcomes without changing the operating model.

Companies want AI productivity while maintaining the same budgeting processes, approval structures, procurement cycles, organizational hierarchies, and technology deployment models they used for SaaS.

They want transformational outcomes without transforming the organization.

Nobody knows who actually owns AI.

Is it the CIO? CTO? Chief Data Officer? Chief AI Officer? Business unit leader? Product team? Engineering? Operations?

Who controls the budget?

Who owns the data?

Who approves deployment?

Who measures ROI?

Who is responsible when the agent makes a mistake?

Everyone wants AI.

Nobody wants to own the entire mess required to make it work.

This is why forward-deployed engineers and AI transformation operators are becoming so important.

Enterprise AI is not traditional plug-and-play software.

Someone has to go inside the organization, understand how the company actually operates, find the data, map the systems, navigate the politics, align the stakeholders, rebuild the workflows, deploy the technology, and convince employees to actually use it.

In many ways, the forward-deployed engineer is becoming a combination of engineer, product manager, consultant, systems integrator, organizational psychologist, and internal political operator.

The models will keep getting better.

The infrastructure will eventually improve.

The cost of intelligence will continue falling.

But none of that automatically fixes the organization trying to deploy it.

The hardest part of enterprise AI is not the AI.

It is the enterprise.

The IPO window just opened wider — Lime rings the Nasdaq bell today as scooters get their public-market redemption arc, while Big Tech’s $725B AI capex guidance is finally starting to spook the same public investors who cheered it a quarter ago.

A few things worth tracking today: Chip M&A returns: Qualcomm buys Modular for ~$4B in its biggest-ever acquisition, aimed at Nvidia’s CUDA moat; Frontier AI, gated: OpenAI ships GPT-5.6 Sol only to ~20 partners after coordinating with the US government; Fintech re-rate: Ramp hits $44B on $750M as Iconiq, GIC and OTPP telegraph the next major fintech IPO.

Let’s get into it. — Trace

Quick Hits

• Chip M&A returns: Qualcomm buys Modular for ~$4B in its biggest-ever acquisition, aimed at Nvidia’s CUDA moat.

• Frontier AI, gated: OpenAI ships GPT-5.6 Sol only to ~20 partners after coordinating with the US government.

• Fintech re-rate: Ramp hits $44B on $750M as Iconiq, GIC and OTPP telegraph the next major fintech IPO.

IPO · $174M IPO at $1.8B

Lime Rings the Nasdaq Bell at $25 as Uber-Backed Scooter Giant Reaches $1.8B Market Cap

Lime — the world’s largest shared micromobility operator, formally Neutron Holdings — priced its Nasdaq IPO at the $25 midpoint on June 30 and begins trading July 1 under ticker LIME, raising roughly $174-182M at a $1.63-1.8B market cap. Uber, which owns a meaningful stake and operates Lime bikes and scooters through its ride-hail app, is a key backer testing whether the scooter category still has a public-market future after Bird’s bankruptcy.

The VC Read · Trace’s Take

Lime’s IPO is the most important shared-mobility print since Uber went public — because it says a category the market wrote off in 2023 is investable again after 70% cost cuts and disciplined city-by-city retreat. The real story is Uber keeping its stake — that’s a bigger vote of confidence than any anchor investor because Uber has the data on ride-substitution and knows if this is a real business or a subsidy from the app. Founders operating in un-sexy hardware-adjacent categories should study the Ting-era playbook: 30+ city exits, tender-offer talent retention, and a 4-year IPO patience trade. Watch the first two days of trading and then Uber’s next 13G filing — those tell you if the reset stuck or if this is just a liquidity event.

Read the full story →

AI · 3-tier launch

OpenAI Previews GPT-5.6 Sol, Terra and Luna — Restricted to Trusted Partners at Washington’s Request

OpenAI unveiled a three-tier GPT-5.6 family — Sol (flagship, $5/$30 per M tokens), Terra (balanced, $2.50/$15) and Luna (fast/cheap, $1/$6) — on June 26 and released it to only ~20 partner orgs after sharing the deployment plan with the US government. It is the first frontier model gated by explicit federal government coordination before wide release.

The VC Read · Trace’s Take

The three-tier ladder is now the standard architecture — Sol/Terra/Luna maps 1:1 to Opus/Sonnet/Haiku — because customers demanded price optionality and every lab that ignored it lost budget. But the real signal is the White House gating: this is the first time a US frontier lab has coordinated a public release with federal officials before shipping, and that changes the risk calculus for every AI-native founder who now has a new reason to worry about API disruptions when Washington gets nervous. Watch which 20 companies are on the trusted-partner list — that’s the emerging enterprise AI whitelist. And watch Anthropic’s response; Claude Sonnet 5’s discount was already priced to fight Terra, and this preview just accelerated the pressure.

Read the full story →

BIG TECH · $3.92B

Qualcomm Agrees to Buy Modular for ~$4B in All-Stock Deal, Taking Aim at Nvidia’s CUDA Moat

Qualcomm agreed on June 24 to acquire Modular, the AI-infrastructure company founded by Chris Lattner (LLVM, Swift), in an all-stock deal valued near $4 billion — its largest ever. The play: use Modular’s MAX platform to run AI models across Qualcomm silicon without hand-tuning code for each chip, directly attacking the software stack that has cemented Nvidia’s data-center dominance.

The VC Read · Trace’s Take

This is Qualcomm finally admitting that owning great silicon without owning the compiler stack means you lose the AI enterprise budget to Nvidia every single quarter — and $4B in stock is a bargain if Modular’s MAX becomes the default portability layer. Chris Lattner is the highest-value acquihire in a decade; compiler founders who ship at scale are almost extinct, and getting him locked into Qualcomm changes the roadmap for every ARM-based device from smartphones to Windows Copilot+ PCs. The bigger read is that CUDA is now being priced as attackable — every AI infra investor pitching an anti-Nvidia thesis just got a $4B comp. Watch whether AMD counters with a Modular-adjacent acquisition and whether Nvidia responds by open-sourcing more of the CUDA stack.

Read the full story →

BIG TECH · $725B combined

Big Tech’s 2026 AI Capex Will Hit $725B, up 77% Year-Over-Year

Alphabet, Amazon, Meta and Microsoft collectively guided to roughly $725 billion in 2026 capital expenditure — up 77% from $410B last year — as the AI infrastructure buildout enters its third and most expensive year. Amazon leads at ~$200B, Microsoft ~$190B, Alphabet $175-185B and Meta $115-135B, according to freshly restated guidance from the four companies’ most recent updates.

The VC Read · Trace’s Take

The $725B number is either the biggest infrastructure moat any four companies have ever built or the biggest coordinated capital misallocation since the fiber overbuild of 1999 — and honestly nobody knows yet because monetization is running 12-18 months behind the spend. What I’m watching is the shift in how founders are pitching: ‘we save hyperscalers 30% on training’ now beats ‘we build a new frontier model,’ because every startup near the buildout has priced in that the Big Four will keep spending regardless. For LPs, the second-order private opportunity is grid infra, cooling, and power PPA aggregators — those are the arms dealers to the AI arms race, and they’re still under-covered. Watch which hyperscaler blinks first on 2027 guidance; that’s the top-tick signal.

Read the full story →

FUNDING · $750M

Ramp Raises $750M at $44B Valuation Led by Iconiq, GIC and Ontario Teachers’

Ramp closed a $750 million primary financing round on June 4 led by ICONIQ, GIC and Ontario Teachers’ Pension Plan, tripling its valuation to $44 billion in under a year as spend-management demand explodes with enterprises trying to rein in AI budgets.

The VC Read · Trace’s Take

The $44B mark essentially prices Ramp as a public-market fintech, and the syndicate — Iconiq, GIC, OTPP, Goldman — is a textbook ‘we’re 12 months from S-1’ lineup. Ramp is the operational winner of the AI-native fintech era because they invested three years earlier in agentic approvals when Brex was still selling cards to founders, and that head start compounds every quarter. For founders, the lesson is that AI-native rebuilds of boring workflows (procurement, expense, close) are still open — the incumbents haven’t figured out how to ship agentic products safely, and every 12 months of delay makes the disruption gap bigger. Watch Brex’s response; the valuation gap is now large enough that a merger conversation is arithmetically compelling.

Read the full story →

FUNDING · $600M

Cyera Raises $600M at $12B Valuation as AI Data-Security Demand Doubles Again

Data security posture management leader Cyera raised $600 million at a $12 billion valuation on June 2, doubling its mark from January and quadrupling since late 2024. Evolution Equity Partners led with Temasek, Cyberstarts, Accel, AT&T Ventures, Blackstone and Coatue participating — cementing Cyera as the runaway DSPM category leader.

The VC Read · Trace’s Take

Cyera’s 4x re-rate in 18 months is the cleanest ‘AI created a real security budget’ story of 2026 — enterprises finally realized that shipping LLMs without knowing where their sensitive data lives is a resignation-generating career risk. DSPM was a boring category two years ago and is now the fastest-growing cyber sub-sector because every CISO needs the answer before their AI committee approves a single model deployment. For founders in adjacent security categories, the play is clear: pick a narrow AI-adjacent risk (prompt injection, agent authorization, model supply chain) and race to build the Cyera of that lane while budgets are still expanding. Watch for a Q4 S-1 filing — this cap table screams ‘IPO ready.’

Read the full story →

AI · 750 tok/sec

Cerebras Runs OpenAI GPT-5.6 Sol at 750 Tokens per Second, Setting a New Frontier-Model Speed Record

Cerebras Systems will deploy OpenAI’s GPT-5.6 Sol on its wafer-scale WSE-3 chips at up to 750 tokens per second in July — roughly 10x faster than any Nvidia GPU deployment of a frontier model in production. The partnership is a strategic proof point for Cerebras ahead of its planned 2026 IPO.

The VC Read · Trace’s Take

This is the most strategically important non-Nvidia inference deployment of 2026 — because if Sol at 750 tok/sec ships and holds, it proves that specialized inference silicon can beat H100 clusters on the workloads customers care most about, and that unlocks the anti-Nvidia infra thesis for a dozen private startups. Cerebras timed this perfectly: land OpenAI as a reference logo weeks before public-market investors read the S-1. Watch whether the S-1 amendment discloses OpenAI-specific revenue commitments; that number moves the IPO by a full turn of multiple.

Read the full story →

AI · ~$800M est.

Adobe Acquires Topaz Labs to Bring Best-in-Class AI Image Enhancement Native to Creative Cloud

Adobe agreed on June 25 to acquire Topaz Labs, the pro-grade AI photo and video upscaling and denoising firm favored by photographers and filmmakers. Terms weren’t disclosed but industry estimates put the deal in the $700M-$1B range, folding Topaz’s Gigapixel, DeNoise and Video AI models directly into Photoshop, Lightroom and Premiere.

The VC Read · Trace’s Take

Adobe buying Topaz is a defensive move dressed as strategy — Topaz was the one AI image product that Adobe’s own Firefly team couldn’t out-ship, and every quarter that pros paid Topaz on top of Creative Cloud was a quarter that Adobe’s AI narrative looked incomplete. The interesting read is what happens to independent creator-tool startups: Adobe just proved they’ll pay $1B for anything that becomes a real second bill on a Creative Cloud subscription. Watch the DxO and Skylum reaction; either they raise fast or they get bought before year-end at a discount to this deal.

Read the full story →

IPO · S-1 filed

Ambiq Micro Files S-1 to List on NYSE, Betting Ultra-Low-Power AI Silicon on Public Investors

Austin-based Ambiq Micro filed an S-1 on June 22 to list on the NYSE, marking the largest ultra-low-power AI chip IPO filing of 2026. Ambiq’s Apollo family powers ~200 million shipped smartwatches, wearables and IoT sensors and is a leading candidate to define on-device AI silicon in consumer electronics.

The VC Read · Trace’s Take

Ambiq is one of the few semi IPOs where the story is real deployed volume, not roadmap slides — 200M shipped units into wearables and hearing aids is a base most chip startups would kill for. The under-appreciated angle is that Apple Intelligence and Google Gemini Nano created a whole new customer class overnight, and Ambiq’s Apollo generation is one of maybe three credible options for anything smaller than a phone. For public-market investors, the read is whether they treat this like a specialized fabless like MPS or a generic ARM licensee — the multiple gap is meaningful. Watch the customer concentration disclosure in the S-1; single-customer risk is the biggest overhang if too much revenue traces to one wearables OEM.

Read the full story →

Read Value Add Pulse →

By Trace Cohen · @Trace_Cohen · t@nyvp.com

Quick gut check before you read another “AI is transforming SaaS” take this week: what if it isn’t transforming software at all — what if it’s replacing it?

That’s not my line. Crunchbase made the argument out loud this week: SaaS isn’t coming back. I wanted to disagree. Then I lined it up against where capital actually moved in the last seven days — and it’s hard to.

The tell is in the term sheets

Look at the rounds that closed this week. Almost none of them are seat-based software:

• Skild AI — $1.4B at $14B+, for an “omni-bodied” robot foundation model.

• Saronic — $1.75B at $9.25B, for autonomous maritime systems.

• Mecka AI — $60M, to train robots on human body-sensor and iPhone data.

• Groq — $650M, rebuilding leadership after Nvidia’s $20B “not-acqui-hire.”

The one classic-software exception — AppsFlyer’s $1B+ raise at $2.7B — got bought into by Google, Meta, Unity and Moloco precisely because it sits underneath everyone’s AI ad spend. Even the SaaS winner of the week is really an AI-infrastructure play.

Why this is structural, not a hype cycle

The SaaS model priced access: pay per seat, per month, forever, whether or not anyone logs in. Agentic AI prices work done — and that quietly breaks the comp table founders and LPs have leaned on since 2015.

For founders: “build a tool people open” is becoming “build an agent that does the job nobody has to open.” Your wedge isn’t a better UI anymore — it’s a better outcome per dollar.

For GPs: your SaaS multiple benchmarks are measuring a category being repriced in real time. An 8x revenue multiple on a seat-based business and an 8x on an outcome-based agent are not the same asset.

For LPs: the durable revenue you underwrote on net-retention math assumes humans keep renewing seats. Agents don’t renew — they get swapped for the next, better agent. Retention curves built for SaaS may not survive contact with this.

What I’m watching

• Whether any pure-seat SaaS company posts accelerating net-new logos in H2 2026 — or whether growth quietly migrates to usage and outcome pricing.

• How fast this week’s ~$3.5B in robotics and agent-foundation rounds turns into revenue, not just valuation.

• The OpenAI and Anthropic S-1s — the first time the public market gets to price “AI that does the work” at scale.

The uncomfortable version: if your 2026 thesis still rhymes with “vertical SaaS plus a copilot,” you may be underwriting the losing side of this trade.

See where the money’s actually going →

I keep a live tracker of AI company valuations against the old SaaS multiples — it’s the clearest picture I’ve found of the repricing as it’s happening. (And if you want the other side, the SaaS multiples tracker shows just how far the correction has already run.)

— Trace

Reply and tell me which side you’re on — “SaaS is fine” or “it’s already over.” I read every one.

• Defense tech mints another unicorn: Twenty raises $100M at a $1B valuation for automated offensive cyber, as 2026 defense funding hits a record.

• The picks-and-shovels trade gets specific: Architect Labs ($24M) attacks the custom-chip design bottleneck while Nvidia seeds the open-weights frontier with a 550B model.

• Capital broadens beyond AI: biotech and fintech both land big rounds (Triveni Bio $65M, Interchecks $50M) as crossover investors re-engage.

FUNDING · $100M

Twenty Raises $100M at $1B -- America’s First VC-Backed Cyber Warfare Startup

Twenty Technologies, an Arlington-based startup building automated offensive cyber operations for the U.S. military and intelligence community, raised a $100 million Series B at a $1 billion valuation led by Accel, with Point72 Ventures, Caffeinated Capital, and Friends & Family Capital participating. The round, which brings total funding to $138 million, mints a defense-tech unicorn in a category most VCs avoided a few years ago.

The VC Read · Trace’s Take

Offensive cyber was the third rail of venture for a decade -- classified buyers, murky exits, reputational risk. Accel leading a billion-dollar round into it is the clearest sign yet that the Anduril playbook has fully normalized selling sharp-edged tech to the government. The moat here is the clearance and the contracts, not the code, which makes it nearly un-disruptable once you’re in -- but also slow, lumpy, and politically exposed. Watch whether the record defense-tech vintage actually produces the exits VCs are now underwriting, because a lot of 2026 marks assume it will.

Read the full story →

BIG TECH · ~$84.75B

Alphabet Upsizes Equity Raise to ~$85B -- With $10B From Berkshire -- to Fund the AI Buildout

Alphabet upsized its equity raise to roughly $84.75 billion to bankroll an AI infrastructure buildout it now guides at $180-190 billion of capex this year. The package pairs a $30 billion underwritten offering with a $40 billion at-the-market program starting in Q3 and a $10 billion private placement to Warren Buffett’s Berkshire Hathaway -- a striking move for a company that prints free cash flow every quarter.

The VC Read · Trace’s Take

The number everyone quotes is $85B; the number that matters is $180-190B of capex -- Alphabet is raising equity it doesn’t strictly need because the buildout has gotten too big to fund comfortably from cash. The Buffett placement is the real tell: when the most disciplined value investor alive anchors an AI-infrastructure raise, the ‘bubble’ framing gets much harder to defend. The AI race has quietly become a balance-sheet contest, which is brutal for anyone hoping to out-compute the hyperscalers. For founders, go build where being capital-light is the whole advantage.

Read the full story →

FUNDING · $310M

Odyssey Raises $310M at $1.45B -- Amazon and AMD Bet on World Models for Physical AI

Odyssey, the Los Angeles startup founded by self-driving veterans Oliver Cameron and Jeff Hawke, raised a $310 million Series B at a $1.45 billion valuation to build ‘world models’ -- AI systems that simulate physical environments with real-world physics for robotics, autonomy, and interactive media. The round was led by Natural Capital, with Amazon, AMD Ventures, GV, Jeff Dean, Elad Gil, Garry Tan, Guillermo Rauch, and Kyle Vogt participating, and notably routes the company onto AWS Trainium after it had earlier taken Nvidia’s money.

The VC Read · Trace’s Take

The number is big, but the cap table is the story. Amazon plus AMD plus a pivot to AWS Trainium -- after taking Nvidia’s money -- is a deliberate shot at the Nvidia-default stack, and it tells you the hyperscalers are now recruiting frontier startups onto their own silicon. The deeper signal: capital is rotating off the screen and into the physical world. Language models are getting commoditized, and the smart money is paying up for the simulation layer that robots and autonomous systems actually need. The risk is timeline -- world models are early, and the distance from a jaw-dropping demo to a shipping product is exactly where a $1.45B mark gets stress-tested. Watch which robotics and autonomy companies adopt Odyssey’s models; that’s the real proof, not the raise.

Read the full story →

FUNDING · $260M

Dream Raises $260M at $3B to Sell ‘Sovereign AI’ Cyber Defense to Governments

Dream, the Israeli cybersecurity startup founded by former NSO Group CEO Shalev Hulio and former Austrian Chancellor Sebastian Kurz, raised $260 million at a $3 billion valuation -- roughly tripling its mark in a year as it sells AI-driven national cyber defense directly to governments and critical-infrastructure operators. The new round was co-led by Bicycle Capital and Group 11.

The VC Read · Trace’s Take

Read the founding team, not the round size. An ex-NSO CEO and a former Austrian Chancellor building a $3B company that sells AI cyber defense to nation-states is the clearest signal yet that ‘sovereign AI’ is the next defense-tech land grab -- and the relationships, not the tech, are the moat. The discipline I like: ~$300M in real contracts underwrites the mark. The risk: government revenue is sticky but slow, and a cap table this geopolitically loaded cuts both ways.

Read the full story →

FUNDING · $24M

Architect Labs Raises $24M to Speed AI Custom-Chip Design With Kindred Ventures

Architect Labs raised $24 million led by Kindred Ventures to build AI tools that accelerate custom-chip design, compressing the time and cost of taping out specialized silicon. As every hyperscaler and AI startup races to design its own accelerators, the software that shortens the chip-design cycle is becoming strategically valuable.

The VC Read · Trace’s Take

The custom-silicon wave has a hidden chokepoint: design takes too long and costs too much. Architect Labs is selling the pickaxe that lets more than just the hyperscalers credibly tape out their own accelerators. If AI can meaningfully shrink the design cycle, the field of who can build bespoke chips widens fast -- and that’s quietly bad news for Nvidia’s default position. Small round, but it’s pointed at exactly the right bottleneck.

Read the full story →

FUNDING · $50M

Radical Numerics Raises $50M to Build AI That Simulates Biology

Radical Numerics raised $50 million led by Emergence Capital to build AI models that simulate biological systems -- a bet that foundation-model techniques can model cells and living processes the way world models simulate physics. It’s part of a clear 2026 wave of capital flowing toward AI-for-science applied to specific scientific domains.

The VC Read · Trace’s Take

World models for physics, simulation models for biology -- 2026’s frontier theme is teaching AI to model reality, not just text. Radical Numerics is a clean expression of that thesis pointed at the slowest, most expensive lab work in the economy. The bar to clear is the same one LifeSciBench just exposed: simulate to augment the scientist, don’t over-claim autonomous discovery. Get the framing right and this is a durable category; get it wrong and it’s another computational-biology cautionary tale.

Read the full story →

AI

Grok 5 Slips to Q3 -- xAI’s 6-Trillion-Parameter Flagship Still in Training

xAI’s Grok 5, the rumored ~6-trillion-parameter flagship, has missed both its Q1 and Q2 2026 windows and remains in training, with full API access now looking like a Q3 story. Prediction markets price a public release before June 30 at roughly 7%, even as xAI ships smaller updates like Grok Voice and Grok Imagine Video.

The VC Read · Trace’s Take

In this market, shipping is the strategy and delay is the cost. Every quarter Grok 5 stays in training, Anthropic and OpenAI bolt another cohort of enterprise customers to their models as the default -- and defaults are nearly impossible to dislodge. A 6T-parameter MoE is an audacious swing, but the real question isn’t the benchmark, it’s whether it trains and serves at a price anyone will pay. Grok Voice and Imagine are smart momentum plays, but they don’t win the deals that are already being signed.

Read the full story →

AI

Nvidia Ships Nemotron 3 Ultra -- a 550B Open Model to Seed Its Own Ecosystem

Nvidia released Nemotron 3 Ultra, a roughly 550-billion-parameter open model, extending its strategy of giving away strong models to drive demand for the hardware that runs them. As the open-weights frontier heats up, Nvidia is using openness as a moat around its silicon rather than its software.

The VC Read · Trace’s Take

Nvidia giving away a 550B model looks generous until you remember the business model: every open weight it ships sells more GPUs. Openness is a complement to hardware, not a competing product -- and that’s a structurally unbeatable position. The smart read for founders is that the open-weights frontier now has the world’s most powerful company actively feeding it, which makes self-hosting a more credible strategy than ever. Watch whether Nemotron and the Mistral partnership pull developers onto the Nvidia stack by default.

Read the full story →

IPO · $85.7B

SpaceX (SPCX) -- Record IPO Surges Past Tesla in Market Cap

After the largest IPO in history, SPCX’s post-listing rally pushed SpaceX’s market cap above Tesla’s -- proof that mega-cap tech can price above private marks and trade up sharply, reopening the late-stage liquidity market that’s been frozen since 2022.

The VC Read · Trace’s Take

Watch the secondary market, not the ticker. Pre-IPO SpaceX shares traded at 20-30% discounts through 2023-24; they’re now at premiums, and the cap just blew past Tesla’s. That repricing -- not the day-one print -- is what tells every LP sitting on illiquid AI and space positions that the bid is back. The next phase is already visible: newly public mega-caps with rich stock spending it on M&A. The merger chatter is noise for now, but the liquidity signal is real.

Read the full story →

Read Value Add Pulse →

By Trace Cohen · @Trace_Cohen · t@nyvp.com

Below roughly $10M ARR, an AI startup’s revenue multiple is meaningless — early rounds price at 50–150x ARR, and the most extreme labs raise $1B+ with zero revenue. That’s the short answer. The longer answer is more interesting.

I’ve made 65+ investments and sat on both sides of this table. The single most common mistake founders make in 2026 is anchoring their raise on a revenue multiple. At the early stage, no one investing in your company is doing that math — and if they are, you probably don’t want their money.

AI company valuation in 2026: why revenue doesn’t set the price

AI company valuations in 2026 are set by team quality, proprietary data or model access, and growth rate — not by a revenue multiple — until a company crosses roughly $10M ARR. Below that threshold, revenue is too small to forecast from reliably, so investors underwrite the trajectory and the people. A $2M ARR company growing 25% month-over-month is priced as the $40M ARR business it could become, which is how you get 80x headline multiples that look insane on a spreadsheet.

The math is simple once you stop fighting it. Seed and Series A pricing is an option on a future, not a discount on a present. When the present is $1M of revenue, the option premium is almost the entire valuation. That’s not a bug in 2026 — it’s the whole product being sold.

AI startup valuation vs revenue: the multiple by stage

Here is how the revenue multiple actually behaves across the AI funding ladder in 2026. Notice that the multiple is highest exactly where revenue is smallest — and that it compresses as the number gets large enough to trust.

StageTypical ARRARR MultipleWhat Sets the PricePre-seed$0–$0.5MN/A / 200x+Founders, thesis, demoSeed$0.5M–$2M50–150xTeam, growth slope, dataSeries A$2M–$10M40–80xGrowth rate, retention signalSeries B$10M–$30M20–40xNRR, gross margin, efficiencySeries C$30M–$75M15–25xDurable growth, unit economicsGrowth / late$75M+10–18xRule of 40, path to profitFrontier lab$0–variesOff the chartTalent, compute, scarcity

Ranges reflect 2026 deal data across Carta, PitchBook, and primary round reporting. Compare current public-software multiples on the SaaS Valuations dashboard and AI-specific comps on AI Valuations.

Why the multiple is meaningless before $10M ARR

There are three structural reasons a revenue multiple tells you almost nothing about an AI company below $10M ARR.

Revenue is statistically noisy

A $1.5M ARR figure swings 40% on two enterprise deals slipping a quarter — too volatile to anchor a price.

Growth dominates the math

At 20%+ MoM growth, this year’s ARR is <10% of the value created over the next 24 months.

Quality of revenue is unknown

Pilots, design partners, and discounted credits all show up as ‘revenue’ but don’t predict durability.

The comp set is the future, not now

Investors price against where the category leader will be, not this startup’s trailing twelve months.

This is also why two AI startups at identical $3M ARR can raise at a $30M and a $300M valuation in the same month. One is growing 8% MoM with 95% gross-revenue retention; the other is growing 30% MoM with 140% net retention and a proprietary dataset. The 10x valuation gap has nothing to do with the revenue line they share.

The pre-revenue extreme: $1B+ before a dollar

If the multiple were meaningful, pre-revenue companies couldn’t exist at scale. Yet 2025–2026 produced the largest pre-revenue rounds in venture history, priced entirely on talent density, compute access, and strategic scarcity.

Safe Superintelligence Ilya Sutskever’s lab — no public product, no revenue ~$32B

Thinking Machines Lab Mira Murati’s team — raised before shipping ~$12B

xAI Compute + talent + distribution via X ~$80B (2025)

Mistral AI European frontier lab, early monetization ~$14B

These aren’t revenue bets — they’re fund-returner bets on foundational technology. An investor putting $100M into a $30B pre-revenue lab is buying a 0.3% slice of optionality on something that could be worth $300B or zero. That underwriting logic simply doesn’t map to a revenue multiple, and pretending it does is how people convince themselves the whole market is irrational.

What actually sets AI startup valuations early

If revenue isn’t the input, what is? After 65+ checks, here’s the honest order of what moves an early AI valuation in 2026.

What Pushes Valuation Up

• ✓ Net revenue retention above 120%

• ✓ Growth rate of 20%+ month-over-month

• ✓ Proprietary data or workflow no one else has

• ✓ Gross margin above 70% after inference cost

• ✓ A team that’s shipped at frontier scale

What Pulls Valuation Down

• ✕ Revenue that’s mostly pilots or credits

• ✕ Gross margin gutted by API/inference spend

• ✕ Thin wrapper with no defensibility

• ✕ Churn that signals a vitamin, not a painkiller

• ✕ A model dependency a foundation lab can erase

The brutal part: gross margin is where a lot of 2026 AI revenue dies. A company can show $5M ARR and 40% gross margin because inference cost eats the rest — and that revenue is worth a fraction of $5M of clean 80% software margin. This is the detail that separates a real AI business from an expensive demo, and it’s why I push founders to report margin alongside ARR before anyone asks.

When the multiple finally starts to matter

Around $10M ARR, the revenue line gets large enough to forecast from, and the conversation flips from narrative to numbers. Past $50M ARR, AI companies start to trade like premium SaaS — a 2026 software business at scale lands roughly 10–20x forward revenue, with the premium reserved for those clearing a Rule of 40 (growth rate plus profit margin above 40%). The frothy 100x multiples don’t survive contact with a number big enough to model.

For founders, the practical takeaway is sequencing. Below $10M ARR, raise on your story, your team, and your slope — and don’t let an investor talk you into a revenue-multiple frame that caps your price. Above $10M ARR, do the opposite: lean into retention, margin, and efficiency, because that’s now what you’re being paid for. Track where the public comps sit on the SaaS Valuations dashboard so your ask stays tethered to reality as you scale.

https://valueaddvc.com/blog/ai-startup-valuation-vs-revenue-why-the-multiple-is-meaningless-before-10m-arr

I built the Market Event Correlator to answer a simple question: do major macro events actually move markets, or does the market just move? After annotating 81 events across 857 trading sessions — Fed decisions, tariff shocks, AI launches, IPOs, earnings beats — the answer is surprisingly structured.

Event days are 2.55× more volatile than quiet days. Knowing which kind of event is coming is more valuable than knowing whether one is coming. Here’s what the data shows.

01 · AI Events Are the Only Reliable Directional Edge

Nine AI-annotated events. Eight of nine were positive for QQQ. Average return: +0.46%. Win rate: 89%. Every major product launch — Sora, GPT-4o, Apple Intelligence, Stargate — produced a green day.

The singular exception was DeepSeek on January 27, 2025, which dragged QQQ −3.8% as Nvidia fell 17% intraday. That exception is more instructive than the rule: AI news is bullish for the broad tech index unless it threatens the infrastructure layer. Product capability improvements are priced as tailwinds. Cost compression at the model level is priced as a threat to capex.

VC implication: The market has consistently rewarded AI announcements since 2023. This is a sentiment tailwind for portfolio marks — but it also means “AI” as a category signal is increasingly crowded and mean-reverting.

02 · The Fed Decision Is Noise. The Language Is Everything.

Fourteen FOMC meetings in the dataset. Seven up days. Seven down days. Perfect coin flip. Average return: +0.07% — essentially zero.

The worst Fed day in the entire dataset was December 2024 (QQQ −3.5%). It was a rate hold — not a hike, not a cut. The selloff came from the dot plot projecting two 2025 cuts instead of the expected four. One word in the accompanying statement moves more than the actual basis-point decision.

VC implication: For portfolio companies raising in rate-sensitive windows, the direction of the cut matters less than the forward signal. Build rate sensitivity analysis around Fed language, not Fed action.

03 · The Best and Worst Days in Three Years Happened 24 Hours Apart

The April 2025 tariff sequence is the defining macro event of this dataset.

• Apr 4 — China 34% retaliation tariff → QQQ −3.5%

• Apr 7 — Pause rumors + short squeeze → QQQ +3.7%

• Apr 8 — China 104% tariff escalation → QQQ −5.0% (worst day in 857 sessions)

• Apr 9 — 90-day tariff pause announced → QQQ +12.1% (best day in 857 sessions)

The 17.2 percentage-point swing between April 8 and April 9 is the largest 48-hour whipsaw in this dataset. Both sessions were driven by a single social media post. Political event volatility (σ 4.32%) is more than double the dataset average — and the outliers dwarf every other category.

The political event average return is technically positive (+0.32%) — but strip April 9 and political events average −0.80%. The distribution is: escalation = sell, pause/rollback = massive buy. The pattern is asymmetric and driven by a single session.

VC implication: Runway decisions and close timings around policy windows carry hidden tail risk. Single-event vol is structurally underpriced in standard models.

04 · Monday Is the Best Trading Day. Thursday Is the Only Losing Day.

The Mon–Thu spread of 0.40% is the most persistent time-of-week pattern in 857 sessions. Monday absorbs weekend news and opens bullish. Thursday catches pre-announcement anxiety — FOMC decisions fall on Wednesdays, making Thursday the reflexive second-guess day. Earnings also disproportionately drop Thursday evening, adding pre-weekend risk premium.

05 · IPO Days Are Bad for Your Existing Portfolio Marks

Fifteen IPO events. Average QQQ return on IPO days: −0.20%. The mechanism: capital flows to the new listing, meaning it flows from existing tech holdings. The more compelling the IPO story, the harder existing names get hit.

Case in point: CoreWeave (CRWV) priced March 28, 2025 — below its $47–55 target range at $40. QQQ fell 2.3% that session as the subdued pricing raised questions about AI infrastructure demand more broadly. The secondary effect on AI comps was immediate.

VC implication: IPO calendar awareness isn’t just about direct exposure to the new listing. It’s about managing the temporary headwind to your existing AI/tech positions during allocation windows. Don’t mark up on IPO day.

06 · 2026 YTD Is Quietly the Best Year in the Dataset

Through 105 trading days, 2026 is averaging +0.147% per day — the highest in the dataset — at σ 0.87% volatility, well below 2025’s tariff-shock-inflated σ 1.29%. April and May 2026 have been standout months at +0.45% and +0.37% average daily returns respectively.

2025 had 2× the volatility of 2023 — almost entirely because of April. One month of tariff shock doubled annual vol. That’s the cleanest illustration of single-event tail risk this dataset produces.

The Full Dataset

All 857 sessions, 81 annotated events, and the interactive correlator are live at https://valueaddvc.com/market-event-correlator

t@nyvp.com · @Trace_Cohen NYVP · ValueAdd VC

Last week Viktor wrote a brief, built a landing page, and opened a pull request.

Last week, Viktor wrote a campaign brief, built a landing page, opened a pull request, generated a board-ready PDF from live Stripe data, and sent a follow-up email to a churned customer. All from Slack. Same colleague that also pulls your reports and monitors your dashboards. 5,700+ teams. 3,000+ integrations.

Start free. $100 in credits →

The complete startup and VC tech stack — tools I actually use, recommend, and trust. From global HR to sales outreach to privacy protection.

Best Tech, Marketing & Sales Tools for Startups & VCs | Value Add VC

Curated list of the best tools for startups, VCs, and founders

Trace Cohen

I analyzed 3 years of Central Park rain data to see if the feeling was real.

It was https://valueaddvc.com/nyrain

27 of 38 summer weekends had measurable rain.

That’s 71%.

Friday was the rainiest day of the week, with rain on 43.6% of summer Fridays. Basically worse than a coin flip for your evening plans.

Sunday had the highest rainfall volume of any day, averaging 0.17 inches. It may not rain every Sunday, but when it does, it commits.

Thursday was objectively the best day to be outside in NYC: lowest precipitation, clearest skies, least weekend-related misery.

The wildest stat:

70.8% of rainy Sundays were preceded by a gross Friday or Saturday.

The weekend basically telegraphs its own downfall.

I built a full dashboard using NWS Central Park data with every weekend tracked and every raindrop counted

It’s all here https://valueaddvc.com/nyrain

NYC Rain Dashboard — Live & Historical Rainfall Data

Live and historical rainfall data for New York City by borough. Free interactive dashboard.

Trace Cohen

Attio is the AI CRM for high-growth teams.

Connect your email, calls, product data and more, and Attio instantly builds your CRM with enriched data and complete context. Whether you’re running product-led growth or enterprise sales, Attio adapts to your unique GTM motion.

Then Ask Attio to plan your next move.

Run deep web research on prospects. Update your pipeline as you work. Find customers and draft outreach emails. Powered by Universal Context, Attio's intelligence layer, Attio searches, updates, and creates across your data to accelerate your workflow.

Ask more from your CRM.

Ask Attio

A comprehensive suite of venture capital analysis tools. Benchmark your fund, track unicorns, calculate SPV returns, and model liquidity waterfalls.

Value Add VC | 45+ Free VC Tools Hub

45+ free venture capital tools for fund managers and investors. Fund benchmarking, unicorn tracking, IPO dashboards etc

Value Add VC • Trace Cohen

SpaceX’s S-1 gives investors a consolidated look at the company as it prepares to enter the public markets. This is not a Starlink spinout or a theoretical future IPO. The filing presents SpaceX as one combined company across launch, Starlink/connectivity, government and defense, Starship, and related infrastructure initiatives.

The important framing is that SpaceX is not a traditional aerospace company. It combines multiple financial models inside one entity: recurring connectivity revenue, launch services, government contracts, defense infrastructure, satellite deployment, and long-duration R&D.

That makes the S-1 more useful as a segment-by-segment financial analysis than a simple revenue multiple exercise.

Key Financial Snapshot

Based on the S-1, SpaceX is already operating at major public-company scale.

• Q1 2026 revenue: $4.694B

• Annualized revenue run-rate: approximately $18.8B

• Starlink/connectivity revenue: $3.26B in Q1 2026

• Starlink share of total revenue: approximately 69%

• Starlink operating profit: $1.19B

• Starlink operating margin: approximately 36.5%

• Starlink subscribers: approximately 10.3M worldwide

• IPO valuation context: approximately $1.75T

• Implied annualized revenue multiple: approximately 93x

The main takeaway is that Starlink is already the financial center of the business. SpaceX is not simply a launch company with a broadband side business. The filing shows a scaled connectivity platform supported by a vertically integrated launch operation.

Segment Breakdown

Starlink / Connectivity

Starlink is the clearest segment to model because it has recurring revenue, subscriber growth, and operating profit.

• Q1 2026 revenue: $3.26B

• Share of total revenue: approximately 69%

• Q1 2026 operating profit: $1.19B

• Operating margin: approximately 36.5%

• Subscribers: approximately 10.3M globally

• Revenue model: consumer broadband, enterprise, mobility, aviation, maritime, and government connectivity

The investor case for Starlink depends on whether SpaceX can continue growing subscribers while improving unit economics across terminals, satellites, launch costs, and network infrastructure. Public investors will focus on ARPU, churn, customer mix, international penetration, capex requirements, and whether operating leverage continues as the network scales.

Launch Services

Launch is smaller than Starlink as a revenue contributor, but it is central to SpaceX’s strategic advantage.

The segment supports:

• Commercial customer missions

• Government launches

• Internal Starlink deployment

• Future Starship-related economics

The key point is that launch is not just a standalone revenue line. It is the infrastructure layer that enables Starlink’s economics. SpaceX’s launch capability gives the company lower satellite deployment costs, faster constellation expansion, higher launch cadence, reuse-driven cost advantages, and less dependence on third-party launch providers.

Investors will focus on launch cadence, backlog, revenue by customer type, gross margin per launch, reusability economics, failure risk, insurance exposure, and the Starship transition timeline.

Government and Defense

Government and defense revenue gives SpaceX durability and strategic relevance. This includes NASA, DoD, NRO, national security launch, defense communications, and related government programs.

This revenue should be analyzed differently from Starlink subscription revenue. It can be durable and high-quality, but it is tied to procurement cycles, agency budgets, contract renewals, classified programs, and political risk.

Key investor questions include:

• How concentrated is revenue by agency?

• What is the contract duration and backlog?

• What is the margin profile?

• How much revenue is tied to defense budgets?

• How important are classified programs?

• How does Starlink/Starshield fit into government demand?

Government and defense work positions SpaceX as critical infrastructure, not just a commercial broadband or aerospace company.

Starship

Starship is the long-term option value in the filing. It is not the primary current revenue engine, but it could materially change SpaceX’s future cost structure if it reaches commercial scale.

The main investor questions are:

• Development cost

• Testing cadence

• Regulatory approvals

• Launch reliability

• Reusability milestones

• Commercial readiness

• NASA and Artemis-related usage

• Impact on future launch costs

• Impact on future Starlink deployment economics

Starship could improve SpaceX’s cost structure and expand the company’s addressable market over time. It could also continue consuming significant capital before producing predictable revenue. Public investors will likely treat it as long-duration upside with execution risk.

Consolidated Business Model

SpaceX is best understood as a vertically integrated infrastructure company with several revenue types.

• Recurring revenue: Starlink subscriptions and connectivity services

• Transactional revenue: commercial and government launch services

• Contract revenue: NASA, DoD, NRO, and other government programs

• Strategic infrastructure revenue: defense communications, enterprise connectivity, aviation, maritime, and mobility

• Long-term option value: Starship, orbital logistics, and future space infrastructure

This mix makes SpaceX harder to value than a standard aerospace, telecom, defense, or software company. Investors will need to evaluate revenue quality, margin structure, capex intensity, and free cash flow conversion across the full company.

Valuation Framework

At a $1.75T valuation and approximately $4.694B of Q1 revenue, SpaceX would be valued at roughly 93x annualized Q1 revenue.

That is a significant premium multiple. The valuation debate will likely center on whether investors view SpaceX as a conventional aerospace company, a satellite broadband company, a defense infrastructure company, or a vertically integrated technology infrastructure platform.

The segment-level framing matters:

• Starlink likely receives the highest multiple because it has recurring revenue, scale, subscriber growth, and operating profit.

• Launch may receive a lower standalone multiple, but it provides strategic cost advantages.

• Government and defense revenue may receive a durability premium, but also a customer concentration and procurement-cycle discount.

• Starship is long-duration upside, but comes with high capex and execution risk.

• Consolidated capex will be a major factor in determining the appropriate public-market multiple.

The central financial question is whether SpaceX can convert growth into durable free cash flow while continuing to fund satellites, launch infrastructure, Starship, and other capital-intensive initiatives.

Investor and VC Winners

The S-1 also makes SpaceX one of the largest potential venture liquidity events in history. Exact outcomes will depend on IPO pricing, lockups, share sales, and cost basis, but several groups appear positioned for major gains.

• Elon Musk: If Musk owns roughly 42%–43%, his stake would be worth approximately $735B–$752B at a $1.75T valuation.

• Valor Equity Partners / Antonio Gracias: Valor-linked entities reportedly hold approximately 7.3% of SpaceX Class A stock. At a $1.75T valuation, that stake would be worth approximately $128B.

• Founders Fund: One of SpaceX’s most important early venture backers. Reported potential returns could exceed $60B, making this one of the largest single-company venture outcomes ever.

• Sequoia Capital: Reported potential proceeds are approximately $2B, still a major venture-scale liquidity event.

• Alphabet / Google: Alphabet invested alongside Fidelity in 2015, reportedly around $900M. If it retained meaningful exposure, the mark-to-market gain could be substantial.

• Fidelity: Fidelity participated in the 2015 financing and is likely one of the larger crossover beneficiaries, though exact proceeds depend on current ownership.

• Gigafund: Gigafund has concentrated Musk ecosystem exposure, including SpaceX. The IPO would likely be one of its defining fund-level return drivers.

• DFJ / Draper Fisher Jurvetson-related investors: DFJ was an early SpaceX backer. Even a smaller retained position could represent a very large multiple because of the early entry price.

• Tesla: Tesla reportedly invested approximately $2B in SpaceX for nearly 19M shares. The current value depends on IPO price, final share count, and whether Tesla continues to hold the position.

• Employees and executives: Reported individual holdings include Luke Nosek at approximately $6B, Gwynne Shotwell at approximately $2.3B, Ira Ehrenpreis at approximately $250M, and Randy Glein at approximately $50M.

The broader venture takeaway is that SpaceX could become one of the largest fund-returning outcomes ever. The key distinction will be paper value at IPO versus realized DPI after lockups and share sales.

Bottom Line

The SpaceX S-1 shows a company already operating at public-company scale. Q1 2026 revenue of $4.694B implies almost $19B of annualized revenue, while Starlink contributed about 69% of total revenue and generated $1.19B of operating profit.

The IPO should be analyzed as a consolidated infrastructure platform. Starlink is the financial center, launch provides the deployment and cost advantage, government and defense provide durability, and Starship represents long-term upside.

At a $1.75T valuation, the market would be underwriting SpaceX at approximately 93x annualized Q1 revenue. The investor question is whether SpaceX can turn this combination of assets into durable margins, free cash flow, and segment-level growth that supports one of the highest-profile public-market valuations in history.

Big Desk Energy | Subscribe | Big Desk Energy

startup insights, stories, and vibes sent to your inbox every Tuesday

Big Desk Energy • Big Desk Energy

Value Add VC | 45+ Free VC Tools Hub

45+ free venture capital tools for fund managers and investors. Fund benchmarking, unicorn tracking, IPO dashboards, AI valuations, SpaceX IPO, and more. Built by Trace Cohen.

Value Add VC • Trace Cohen

Attio is the AI CRM for high-growth teams.

Connect your email, calls, product data and more, and Attio instantly builds your CRM with enriched data and complete context. Whether you’re running product-led growth or enterprise sales, Attio adapts to your unique GTM motion.

Then Ask Attio to plan your next move.

Run deep web research on prospects. Update your pipeline as you work. Find customers and draft outreach emails. Powered by Universal Context, Attio's intelligence layer, Attio searches, updates, and creates across your data to accelerate your workflow.

Ask more from your CRM.

Ask Attio

I also made a curated list of recommended products you can find here https://valueaddvc.com/tools

I analyzed 70 U.S. IPOs from 2021–2025 across valuation size, sector, and vintage year to understand what actually happened after companies went public.

The issue is not that companies cannot go public. The issue is that both private and public markets spent years pricing companies as if growth, liquidity, and multiple expansion would continue indefinitely.

2021 was the clearest example of this distortion.

397 U.S. IPOs raised over $142B, the largest IPO year ever by proceeds. Companies entered the public markets during a period of zero interest rates, massive liquidity, meme-stock psychology, retail speculation, and a broader environment where future narratives mattered more than present cash flows.

Then rates moved from near 0% to over 5%, and long-duration growth assets repriced across the market.

The broader IPO market reflects that shift clearly. Most 2021 IPOs still trade below their IPO price today, and many are down 70–90% from peak valuations.

In the cohort I analyzed:
• 81% of 2021 IPOs are still underwater
• 11 lost more than half their value
• several declined 70–90%+

Examples include $BIRD, $BMBL, $TASK, $SG, and $LZ.

Many of these were not fundamentally bad businesses. They were simply priced as if execution risk no longer existed.

The most interesting finding from the dataset was what I call the “Valuation Smile.”

The best IPO outcomes did not come from the largest companies. The strongest-performing cohort was companies that went public between roughly $1B–$2B in valuation.

That bucket had:
• a 50% win rate
• a 571% average return
• zero companies that lost more than 50%

Companies like $CRDO, $ATAT, and $KGS materially outperformed after listing because they were large enough to prove durability, but still small enough to compound meaningfully in public markets.

Meanwhile, the $10–20B IPO range was the weakest-performing category in the dataset:
• only a 20% win rate
• negative average returns
• several of the sharpest post-IPO drawdowns

Examples include $TASK, $DNUT, and $DOCS.

Historically, IPOs helped fund future growth. Increasingly, modern IPOs function more as late-stage liquidity transitions after private markets have already priced in years of future success.

The public market used to fund upside. Today it often inherits fully-priced upside.

That changes the entire psychology of the IPO.

Once a company IPOs above roughly $40–50B, the math becomes extremely difficult. A company entering the market at those valuations already needs extraordinary scale to generate venture-like public returns.

The data reflects this clearly.

The $40B+ IPO bucket had:
• a 17% win rate
• a -50% median return
• half the companies losing more than 50% of their value

Examples include $RIVN, $DIDI, and $VG, while $ARM was one of the few major exceptions.

The sector divergence was equally important.

Semiconductors and AI infrastructure significantly outperformed, including $CRDO, $ALAB, and $ARM. Meanwhile many SaaS IPOs remained underwater, including $BRZE, $TOST, $KVYO, and $TASK.

In 2021, high-growth SaaS companies often traded at 20–40x revenue. Today many still-growing SaaS businesses trade closer to 4–8x revenue despite continuing to grow.

The market did not stop rewarding growth. It became more selective about which types of growth deserved premium multiples.

Infrastructure scarcity, semiconductors, compute demand, and AI leverage replaced the old “growth at any price” framework that dominated the ZIRP era.

Analyst expectations evolved as well. During the 2021 cycle, markets tolerated heavy cash burn, slowing margins, and distant profitability because revenue growth alone justified premium valuations.

Today the market demands operating leverage, margin durability, AI positioning, and clearer paths to cash flow generation. One earnings miss or guidance cut can erase years of momentum because expectations are materially less forgiving.

This is why I think the broader conversation around IPOs misses the point.

The IPO market itself is functioning.

Companies are still going public. Capital still exists. Liquidity still exists.

What changed was the market’s tolerance for unrealistic expectations.

Retail investors, analysts, crossover funds, growth-stage private markets, and media narratives collectively created an environment where companies were expected to enter public markets already fully optimized.

That is not how long-term compounding typically works.

The best public companies historically had room to grow after listing. Many modern IPOs arrive carrying aggressive private-market marks, ambitious analyst models, and years of future success already embedded into the valuation.

Then the public market simply reprices reality.

Best Tech, Marketing & Sales Tools for Startups & VCs | Value Add VC

Curated list of the best tools for startups, VCs, and founders

Big Desk Energy | Subscribe

startup insights, stories, and vibes sent to your inbox every Tuesday

200+ AI Side Hustles to Start Right Now

AI isn't just changing business—it's creating entirely new income opportunities. The Hustle's guide features 200+ ways to make money with AI, from beginner-friendly gigs to advanced ventures. Each comes with realistic income projections and resource requirements. Join 1.5M professionals getting daily insights on emerging tech and business opportunities.

Get the guide

The complete startup and VC tech stack — tools I actually use, recommend, and trust. From global HR to sales outreach to privacy protection.

Best Tech, Marketing & Sales Tools for Startups & VCs | Value Add VC

Curated list of the best tools for startups, VCs, and founders

The conversation around enterprise AI is still anchored in the wrong layer. Most of the focus sits on models, agents, and technical capability, when the actual constraint is much more structural: whether an organization can absorb and operationalize these systems inside how work really happens.

The assumption embedded in most “agent-first” thinking is that workflows exist in a clean, defined form, ready to be translated into software. In practice, that assumption breaks immediately once you step inside a large enterprise. What appears as a workflow at the executive level is closer to a distributed system of decisions across teams, tools, and informal coordination layers. The logic of the business is not encoded in systems; it is embedded in people, habits, and a long tail of exceptions.

That makes the first phase of enterprise AI not engineering, but reconstruction. Before anything can be automated, the organization itself has to be made legible. That requires mapping how work actually happens, resolving conflicting system definitions, and surfacing dependencies that are often invisible until you try to formalize them. What looks linear decomposes into a network of handoffs, ownership gaps, and inconsistent data flows.

Even after that work is done, progress is constrained by alignment. Enterprise AI rarely maps to a single owner or budget. The inefficiency is experienced in one part of the business, the enabling systems are owned elsewhere, and the mandate for change sits in a third function. Moving forward becomes a coordination problem across incentives, not a straightforward build decision.

This is where most external narratives break. They assume that once the system is built, value follows. In reality, the largest failure point is adoption. When AI systems formalize previously informal work, they introduce visibility, measurement, and standardization. That shift benefits the organization at a system level, but it often changes the experience of the individual doing the work in ways that are not immediately rewarded. The result is predictable: partial adoption, workarounds, and a slow reversion to legacy processes.

This is why agent engineering, on its own, consistently underdelivers in enterprise environments. Agents assume structured inputs, stable data, and well-defined processes. Most enterprises do not meet those conditions. When agents are introduced, they tend to expose fragmentation rather than resolve it.

What is emerging now across the market is a correction to this misunderstanding.

The recent moves by @AnthropicAI and @OpenAI point in the same direction: enterprise AI is not a pure software business. It is a deployment and transformation business.

Anthropic’s joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs is explicitly structured as a services layer to help companies integrate AI into operations, particularly across private equity portfolios.

In parallel, OpenAI is pursuing the same model through its “Deployment Company,” a large-scale joint venture backed by private equity firms to embed its technology inside enterprises.

Both are converging on a model that looks much closer to consulting than SaaS: embedding teams, reshaping workflows, and only then layering in AI systems.

This is not incidental. It reflects where the actual bottleneck sits. Model quality is improving rapidly and becoming more accessible. The harder problem is translating messy, unstructured organizational reality into something that can be executed by software.

That is exactly the layer @PalantirTech built around. Its forward-deployed model was based on the idea that software alone does not create value inside complex institutions. Value comes from making the organization legible, aligning systems and incentives, and iterating in close proximity to users until the system reflects reality.

What is happening now is less a new paradigm and more a return to that insight, driven by AI.

The competitive advantage in enterprise AI will not come from building better standalone agents. It will come from owning the translation layer between how organizations actually operate and how software requires them to operate. That means understanding work at a granular level, reducing fragmentation across systems, aligning incentives across stakeholders, and only then introducing automation.

Until that sequence is respected, enterprise AI will continue to look more promising in concept than in execution, not because the models are insufficient, but because the environments they are being deployed into are.

—

The complete startup and VC tech stack — tools I actually use, recommend, and trust. From global HR to sales outreach to privacy protection.

Best Tech, Marketing & Sales Tools for Startups & VCs | Value Add VC

Curated list of the best tools for startups, VCs, and founders

Trace Cohen