Healthcare in the United States is awkward to describe. It is an essential service that everyone consumes at some point in their lives but that most don’t fully or directly pay for. The ultimate cost of a procedure isn’t known at the time of its provision and instead floats in a liminal space until after insurers have had a chance to review it.

That resolution process is no simple thing either. Every diagnosis and procedure is translated into one or more of 10s of thousands of codes, each of which needs to be backed by the proper documentation. Miscodings happen all the time by accident while intentional codings are subject to dispute. Insurers might deny claims initially or ask for more information or pay less than expected (~18% of submitted claims are initially denied, according to Waystar’s management). Providers might then appeal the denial or correct the claim.

Further complicating matters, the same procedure carries different prices depending on who’s paying. Medicare sets its rates by formula while each private insurer negotiates its own contracted rates. A hospital might therefore have dozens of different prices for the exact same CT scan. And even a single patient encounter can involve multiple payers, including a primary insurer, a secondary insurer, the patient itself (deductible and copay), and sometimes a government program as well. Each payer has different coverage policies – for some injuries, for instance, an insurer might require an X-ray before approving a more expensive MRI scan – and those policies sometimes change without warning. In its recent Analyst Day, management pointed out how for the same category of denied claims, one national insurer denied the entire claim about half the time while another denied individual claim lines ~90% of the time.

In short, healthcare billing is immensely complex. A claim can contain hundreds of data elements, its resolution turning on any number of contingencies, like whether the patient is covered, whether prior authorization is required, whether the care is medically necessary and properly coded, how it’s priced under the relevant contracts, etc. It is little wonder then that the administrative costs of delivering care amounts to ~15% to 25% of US healthcare spending.

The chain of steps that carry a procedure from claim to cash is called revenue cycle management (RCM). Anywhere between 10% to 20% of this activity is outsourced to firms like R1 RCM and Ensemble. The other 80% to 90% remains under the control of healthcare providers. In either case, software from Waystar and numerous other vendors is used to manage all or parts of the process.

Subscribe now

Read more

What are scuttleslops?

[scuttleslops] AI Data Center Economics, ADI, INTU, IDXX

InPractise Interview with Former Global Executive, AWS Data Center Business at Amazon; August 13, 2026

The former AWS executive says current AI data-center projects are among “the most challenging things I’ve been involved in” from a return-on-investment standpoint because tens of billions of dollars go out upfront for land, power and accelerators while the revenue ramp comes later. His current base case for hyperscaler AI infrastructure is roughly a 15%–20% internal rate of return, improving into the 20s for well-utilized Trainium deployments and potentially reaching 35%–40% by 2028 if inference scales as expected.

On capital intensity, he estimates a fully loaded one-gigawatt AI campus at roughly $40–$45 billion today: about $12–$15 billion for the site, shell, electrical and cooling infrastructure, another $6–$8 billion for outside-the-fence power and transmission, and roughly $25–$30 billion for servers, accelerators, networking and storage (based on AWS Trainium). AWS historically built ahead of demand partly because management believed that if it did not lock up scarce land and power, somebody else would. The downside is that the industry is currently paying extraordinary premiums for power — he estimates roughly 30%–40% above normalized levels in many deals — because an expensive megawatt is still preferable to billions of dollars of GPUs sitting idle and depreciating.

AWS’s custom silicon is supposed to change that return equation. Trainium 2 is only modestly cheaper than an equivalent NVIDIA deployment because AWS receives substantial NVIDIA discounts, but the expert expects Trainium 3 to improve cost per floating-point operation much more materially. His argument is not merely that AWS saves money on each chip; it can potentially push three to four times more useful compute through the same constrained megawatt, which raises the revenue-generating capacity of land and power that have already been secured. In his words, Trainium 3 “changes what you get for your $40 billion.”

Subscribe now

Read more

Adyen, H1 ‘26 Earnings Call, August 13, 2026

Growth is still being driven less by new logos than by existing customers steadily handing it more of their payments business. Roughly two-thirds of first-half revenue growth came from merchants that joined before 2025. Merchants typically give Adyen less than 20% of their payment volume in years three through seven, but more than 40% after year 12. Customers such as Uber, Microsoft, Spotify and Google are still expanding after more than a decade, implying the addressable opportunity inside an existing account can be several times larger than the initial contract.

Adyen increasingly wants to own the infrastructure surrounding the transaction, not simply the card-processing step. Payments remain the anchor, but management now describes Adyen as a “financial operating system” spanning loyalty, billing and money movement. Talon.One brings promotions and loyalty into the same platform as Adyen’s online and in-store transaction data, giving retailers a better shot at recognizing the same shopper across channels and targeting offers accordingly. Orb is strategically different: it adds metering and usage-based billing, which management thinks is increasingly essential for software and artificial-intelligence companies whose customers pay according to consumption rather than a fixed monthly subscription. Management insists this does not mark the beginning of an acquisition roll-up strategy: core technology will generally still be built internally, while adjacent products will be evaluated on a build-versus-buy basis.

Van der Does repeatedly argued that combining payments, loyalty and billing data creates a richer picture of the customer and therefore a better foundation for personalization, pricing and conversion products. This is also how management is thinking about the threat from AI agents: merchants worry that consumers will increasingly shop through ChatGPT-like agents rather than visiting a merchant’s app or website directly, weakening the merchant’s direct relationship with the customer. Adyen Agentic is supposed to act like an interoperability layer, letting a merchant connect once to Adyen rather than separately integrating inventory, pricing and payments with every emerging AI-agent protocol.

Adyen compared this explicitly with Adyen’s traditional role connecting merchants to many payment methods: Adyen would publish inventory into multiple agent ecosystems and then route the resulting transaction back into the merchant’s existing fraud, payment and operational systems. The vision is therefore not some proprietary AI model; it is Adyen’s position as the neutral translation and transaction layer sitting between a very large merchant and an increasingly fragmented set of commerce protocols.

So, competitively, management’s claim has broadened noticeably from “we get better authorization rates” to “we can simplify more of a global merchant’s infrastructure than anyone else.” Van der Does was unusually explicit: “there’s no other company that can do everything that Adyen can do,” arguing that breadth increasingly makes Adyen the logical consolidation platform for large merchants. That is a more defensible position than competing purely on payment acceptance rates because the switching cost rises as Adyen touches more workflows, jurisdictions and products. Adyen also continues to distinguish itself by refusing to own the consumer relationship. It has no current plan to build a branded consumer wallet, which management says appeals to large merchants that view their shoppers as their own asset

Subscribe now

Read more

Zoetis, Q2 ‘26 Earnings Call, August 06, 2026

Zoetis’s second quarter exposed a sharp deterioration in its U.S. companion-animal business rather than a broad collapse across the company. Organic revenue fell 1%, with U.S. revenue down 7% and companion-animal revenue down 6%, while livestock grew 11% and international revenue rose 6%. Management blamed two things happening at once: veterinary visits have been declining for several years as prices have risen faster than general inflation, and competitors are now using heavier discounts and rebates to take patients from Zoetis rather than expanding the underlying markets.

The competitive problem is most obvious in dermatology, the franchise Zoetis largely created with Apoquel and Cytopoint. Global dermatology revenue fell 16%, U.S. dermatology fell 18%, and Zoetis’s U.S. in-clinic share dropped roughly 10 percentage points year over year to 86%. Management said that historically new entrants helped grow these therapeutic categories, but “this is not what we are seeing today”; in a stagnant market, rivals are instead competing through “discounting, rebates, cross-portfolio bundling and other incentives” to steal existing patients.

Zoetis is responding by cutting the effective price customers pay without permanently lowering published list prices. The distinction is important: the company might give a clinic a rebate for making Apoquel its preferred product, bundle several Zoetis drugs together at a discount, or give a pet owner a coupon at checkout. Kristin Peck said these “gross-to-net” investments are deliberately targeted and temporary because once a list price is cut, “it’s very hard to pull back.”

That strategy protects volume but materially worsens near-term economics. Zoetis had previously expected positive price realization, but management now thinks full-year net pricing could be roughly flat to down 1%, and potentially down 2% toward the low end of guidance. The company is effectively saying that preserving a dog on Apoquel or Simparica today is worth accepting a lower realized price if the alternative is allowing a discounted competitor to establish the patient relationship.

Affordability is a separate problem from competition and may prove harder to fix. U.S. flea, tick and heartworm visits were down roughly 7%, while dermatology and osteoarthritis pain visits declined more than 2%, and online and retail channels are no longer growing quickly enough to offset weaker clinic traffic. Management explicitly said dermatology’s weakness is “mostly a competitive situation,” whereas parasiticides are being hurt more by the broader decline in preventative-care spending.

Global franchise sales of Simparica were flat, and U.S. sales declined 6%, but Simparica Trio still held about 21% of the U.S. oral-parasiticide market, almost twice its nearest competitor, while its share among puppies was approximately 28%. That puppy position matters because newly acquired patients can remain on the product for years, but Zoetis is now spending more on rebates and consumer promotions to prevent competitors from disrupting that lifetime relationship.

Subscribe now

Read more

Booking Holdings, Q2 ‘26 Earnings Call, August 04, 2026

Room nights grew 5%, gross bookings rose 9%, revenue increased 8% and adjusted EBITDA grew 9%, all above guidance. Adj. EPS grew 15% y/y, faster than adjusted EBITDA growth, helped by a 6% reduction in average share count. The Middle East conflict hurt international flights through higher ticket prices and lower capacity, but travelers substituted toward domestic and regional trips rather than abandoning travel altogether.

The Connected Trip strategy continues to show real traction. Transactions containing more than one travel product—such as a hotel plus a flight or rental car—grew at a low-double-digit rate, more than twice as fast as Booking.com’s overall transactions. These connected bookings still represent only a low-double-digit percentage of transactions, so there is substantial room to increase the number of customers buying more than a hotel room.

Payments are the infrastructure that makes this possible. Merchant gross bookings—where Booking handles the customer’s payment rather than merely passing the reservation to the hotel—reached 73% of gross bookings, up four percentage points. Fogel called payments “the glue that brings it all together” because it lets Booking combine hotels, flights and other services into one transaction and manage refunds or service problems centrally.

The Genius loyalty program is another increasingly important defense against Google, Expedia and direct hotel bookings. Level 2 and Level 3 Genius members represented more than 30% of active customers but generated a high-50% share of room nights, meaning the most loyal customers book much more frequently than the average user. They also plan further in advance and are more likely to book directly through Booking rather than arriving through a paid advertisement. This lowers Booking’s dependence on external traffic sources and gives hotels a reason to participate because Genius members bring repeat demand.

The United States remains a large share-gain opportunity because Booking is still weaker there than in Europe. U.S. room nights grew at a high-single-digit rate, driven by domestic demand, and management said progress has come from the unglamorous work of adding inventory, fixing localized product gaps, improving service and marketing more effectively. Alternative accommodation room nights, 37% of Booking’s total, remain a particular weakness, growing only 4%, in part due to the Middle East conflict.

Google’s search-page changes—including AI-generated overviews—are reducing the amount of free traffic Booking receives from ordinary search results, but the damage is currently small. Search-engine optimization, or SEO, was never a large source of bookings, and Booking’s direct channel still accounts for roughly the mid-60% range of transactions. App usage is helping: the share of bookings made through the app has risen from the low-50% range two years ago to the high-50% range today. Management therefore sees no evidence that customers displaced from Google’s organic results are abandoning Booking in large numbers.

On LLMs: “If we look at the traffic we're currently receiving from large language models, so this is not what we are receiving directly in terms of traffic, but from large language models, both on a paid and an unpaid basis, that is still significantly below 1% of our room nights. That hasn't moved so much recently, so no material change over the last few months or quarters. I'm not saying that will never change in the future. This, of course, at some point, might go in a different trajectory. At this moment, it's still very minimal and is not really moving so much. Going a little deeper, the organic traffic, what we are receiving from the LLMs, we believe the vast majority is coming to us according to third-party data sources.”

Subscribe now

Read more

Universal Music Group, Q2 ‘26 Earnings Call, July 30, 2026

Universal Music Group grew second-quarter revenue 13.3% to €3.3 billion, but the headline was flattered by the newly acquired Downtown business; excluding Downtown, revenue rose 6.4%. Adjusted EBITDA increased only 1.5% to €674 million, and excluding Downtown it was flat, so revenue growth did not translate into much incremental profit.

The main reason was mix: faster growth came from lower-margin physical sales, audiovisual, live-related revenue, and independent-artist distribution, while merchandising lost money and corporate legal and professional expenses increased. Recorded Music subscription revenue rose 6.7% excluding Downtown, with roughly 3.5 percentage points of growth coming from better economics in new streaming agreements. That was partly offset by a 1.5-point market-share headwind caused by a lighter release schedule earlier in the year, although management said share improved as the quarter progressed.

Over the past decade, UMG has averaged nearly 14 of the top 20 artists on the International Federation of the Phonographic Industry’s annual global ranking, or about 70% of the most commercially successful acts. That matters because a hit does more than generate streams from one album: it revives the artist’s older catalog, raises merchandise and touring demand, creates documentary and brand opportunities, and attracts other artists who want the same support. Grainge described this as a cycle in which “every new release by a successful artist lifts their entire body of work,” while a newly signed artist creates the valuable catalog of the future.

“Streaming 2.0” is UMG’s attempt to extract more money from streaming without simply relying on subscriber growth. The new agreements with Spotify, YouTube, Amazon, TikTok, Pandora, and others include better pricing as well as protections against fraud, royalty dilution, and large volumes of low-value artificial-intelligence content (in the first quarter of 2026, Universal said these wholesale increases added about three percentage points to recorded-music subscription growth). Grainge’s blunt formulation was that “fans don’t want AI slop,” and UMG does not want fraudulent or machine-generated tracks siphoning money from legitimate artists.

The practical strategy is to move listeners from free services to paid subscriptions and eventually into more expensive tiers that offer additional access, features, or artist experiences. India is the clearest example of UMG trying to force a poorly monetized market to mature. India generated roughly one trillion music streams in 2025, but management estimates that only 7% to 10% of individual streaming users pay for subscriptions. Beginning in late August, major UMG releases will be available only to paying subscribers for their first 72 hours before appearing on ad-supported tiers.

The idea is simple: fans who want immediate access to a major release must pay, giving streaming platforms a reason to market subscriptions and artists a better payout. UMG used a similar playbook in China, where it worked with local platforms to put more music behind paywalls; China subsequently became the world’s fourth-largest recorded-music market and grew more than 20% in 2025. The risk is that windowing may irritate users or push them toward piracy, but management argues that its China experience shows scarce premium content can accelerate paid conversion rather than reduce listening.

The Downtown acquisition expands UMG beyond traditional record-label economics into services for independent labels and artists. Combined with Virgin Music, Downtown gives UMG distribution, royalty administration, marketing, and self-service tools for artists who do not want a conventional record deal. This is strategically attractive because successful independent artists can later become larger service clients, licensing partners, or full label relationships, but the near-term economics are weak: Downtown contributed €202 million of revenue and only €10 million of EBITDA during the quarter. Its margin was roughly 5%, far below UMG’s core Recorded Music margin

On artificial intelligence, UMG’s position is not to block the technology but to insist that it be licensed, controlled, and optional for artists. The Spotify partnership would allow fans to create authorized covers and remixes only from artists who opt in, with UMG seeking a share of the resulting revenue rather than allowing unlicensed models to copy its catalog. Management says it has already secured opt-in agreements from a meaningful number of artists, though the service will need recognizable names and a broad enough catalog to be useful. Consumer research cited by UMG found that about 30% of respondents were interested in AI remixing or personalizing music, while interest reached 80% among surveyed 16- to 19-year-olds.

Subscribe now

Read more

Builders FirstSource, Q2 ‘26 Earnings Call, July 30, 2026

Builders FirstSource described the second quarter as a test of its ability to manage through a weak housing cycle without sacrificing its longer-term competitive position. Net sales declined approximately 9% to $3.9 billion, adjusted EBITDA fell 35% to $329 million, and adjusted EBITDA margin contracted 350 basis points to 8.5%. Organic sales declined 8% in single-family construction, 10% in multifamily, and 2% in repair and remodeling as elevated mortgage rates, inflation, and weak consumer confidence continued to pressure housing demand. Management reduced its full-year outlook and now expects single-family starts to decline nearly 7%, multifamily starts to fall 4%, and repair and remodeling activity to decline 1%.

The company believes it maintained market share despite aggressive pricing by competitors trying to fill excess capacity. CEO Peter Jackson said some rivals had made “aggressive moves, sometimes too aggressive,” and later regretted the business they accepted, while Builders FirstSource remained willing to walk away from unprofitable work. Competitive pricing appears to be stabilizing, but Jackson emphasized that the market remains “a dogfight” and that margins will remain under pressure until volumes recover.

Subscribe now

Read more

Old Dominion Freight Line, Q2 ‘26 Earnings Call, July 29, 2026

Old Dominion Freight Line reported a strong second quarter, with revenue rising 10.4%, earnings per share increasing 32.3%, and the operating ratio improving 450 basis points to 70.1%. The quarter marked a return to revenue growth despite less-than-truckload tonnage remaining 4.1% below the prior year, as a 15.2% increase in revenue per hundredweight more than offset lower volume. Excluding fuel surcharges, yield increased 5.5%, reflecting Old Dominion’s continued refusal to chase freight that does not meet its account-level return requirements.

Management presented the results as validation of a strategy maintained throughout the freight recession: preserve service quality, continue investing, and wait for competitors’ capacity and financial constraints to create market-share opportunities. Superior service supports pricing power: management’s strategy is to price each account according to its specific cost and profitability while generating enough yield to offset inflation and reinvest in the network.

Old Dominion has no meaningful constraints in drivers, tractors, trailers, or real estate and still has more than 35% excess service-center capacity. By contrast, management has heard that some competitors are struggling to complete pickups at month- and quarter-end, temporarily pushing freight toward Old Dominion. The company believes that when industry growth meaningfully accelerates, constrained competitors will surrender disproportionate share to Old Dominion, as occurred in previous upcycles when its tonnage growth exceeded peers by 800 to 1,000 basis points.

Demand is improving, although management still characterizes the recovery as being in the “early innings.” July tonnage remained approximately 1% below the prior year, but its sequential decline was much better than normal seasonality, and revenue per day was tracking 7.5% to 8% higher. Management also expects some freight to migrate from truckload to less-than-truckload as truckload rates strengthen, though it has not yet seen the large weight-per-shipment changes that would indicate a major inflection.

Operating leverage is embedded in the network. Old Dominion handled a 4% sequential tonnage increase with essentially the same workforce. Management believes third-quarter incremental margins can remain in the 45% to 50% range, and direct operating costs are already 200 to 250 basis points better than in the comparable 2022 quarter, when operating ratio was at 69.5%, despite materially lower shipment volumes. Overhead costs remain elevated as a percentage of revenue because the network is underutilized, leaving an estimated 300 to 400 basis points of potential leverage as revenue returns. Satterfield said the path to a sub-70% annual operating ratio is “pretty easy to map out” and suggested that the company’s eventual next objective could again be 500 basis points lower.

In response to a point about a competitor considering autonomous trucks for some line-haul operations, CFO Adam Satterfield points out that Old Dominion gets unusually high utilization from each tractor because the same truck often performs two different jobs: during the day, a driver uses it for pickup and delivery, or P&D—collecting freight from customers and delivering freight around a local market. At night, another driver uses that same tractor for linehaul—pulling trailers between Old Dominion terminals over longer highway routes. Autonomous trucking is currently much more applicable to the predictable highway portion than to local pickup and delivery. So an autonomous tractor might be useful at night between, say, Charlotte and Atlanta, but it may not be capable of handling the daytime work of navigating city streets.

That creates two unattractive choices for Old Dominion. It could dedicate the autonomous tractor exclusively to nighttime linehaul and buy another conventional tractor for daytime pickup and delivery, meaning it now owns two tractors to perform work previously handled by one. Alternatively, it could pay an autonomous-technology provider a per-mile or recurring fee on a tractor that only uses the technology during part of the day, leaving an expensive system idle during its daytime P&D work. So even if autonomy eliminates a linehaul driver, the savings may be offset by higher technology fees, the need for additional tractors, and lower overall equipment utilization

“But to me, it’s something too that, I don’t know, it’s hard to imagine a world where you’ve got 80,000-pound trucks driving up and down the highways without someone sitting in the cab to deal with those one-off scenarios, to deal with cargo theft that is already a problem when you’ve got a driver in the cab. So there’s a lot of other incremental challenges that would present and would have to be dealt with from a regulatory standpoint before this goes worldwide. And obviously, it’s been dealt with and utilized in certain lanes and so forth, but the scale to be nationwide, still have some reservations about.”

Related scuttleblurb post: [ODFL, SAIA, XPO] the LTL carriers are going to be ok (1/5/26)

Intel Corp., Q2 ‘26 Earnings Call; July 23, 2026

Intel delivered another better-than-expected quarter, marking its seventh consecutive period of exceeding guidance, with management arguing that improving execution is finally beginning to translate into tangible financial results. Second-quarter revenue reached $16.1 billion, $1.8 billion above the midpoint of guidance, while non-GAAP gross margin of 41.8% exceeded guidance by roughly 280 basis points and earnings per share came in at $0.42 versus a $0.20 outlook. CEO Lip-Bu Tan said Intel is experiencing “the strongest revenue growth in more than 15 years,” but the more important message was that product demand continues to exceed the company’s ability to supply it.

The Data Center and AI segment was the strongest part of the quarter, with revenue rising 59% year over year to $6.3 billion and operating margin reaching 40%. Xeon 6 is one of the fastest product ramps in Intel’s history, and Granite Rapids demand is particularly constrained because customer reception has exceeded expectations. Management now expects industry server CPU units to grow at a strong double-digit rate in both 2026 and 2027, with momentum extending into 2028, and believes increasing core counts should also support higher average selling prices.

Subscribe now

Read more

InPractise Interview with a Former Senior Director at Synopsys: EDA Front-End vs Back-End Revenue Dynamics; May 26, 2026

The expert’s central view is that Synopsys and Cadence remain protected by unusually high switching costs, deep foundry relationships and the sheer technical difficulty of advanced chip design. As he explained, “TSMC will not spend time with a startup company,” which makes it extremely difficult for new entrants to compete at leading-edge nodes. The back-end tools also directly determine power, performance and area, or PPA, the three metrics that often decide whether a chip is commercially competitive.

The practical switching costs are substantial even when customers technically have alternatives. Engineers are trained for years on one company’s flow, and back-end teams build proprietary scripts and design techniques around those tools. Moving from Synopsys to Cadence can slow a team for months, so switching is generally driven by senior management negotiating price rather than engineers preferring a different product. Large customers often mix vendors in selected areas, especially design-for-test or independent sign-off, but they usually keep one primary flow for most of the chip. The expert summarized the customer mindset simply: engineers “are in their comfort zone, so why change?” This stickiness allows Synopsys and Cadence to sign large, multi-year, enterprise-style agreements, although powerful customers such as NVIDIA can negotiate very large discounts.

Artificial intelligence looks more like an opportunity than a disruption risk for the leading EDA vendors. Generative AI may automate portions of front-end RTL coding and verification, but those areas are less differentiated, easier for startups to enter and apparently contribute a smaller share of industry revenue. The hardest back-end work still depends heavily on experienced engineers who understand physical effects, process variability and complex trade-offs.

Synopsys’s DSO.ai is a more meaningful commercial opportunity because it runs many design experiments automatically and can help optimize PPA much faster than a conventional team. He does not yet see a dramatic AI-driven revenue inflection, however, describing adoption as “an iterative process” rather than a step-function change. Large semiconductor companies are growing more powerful and negotiating harder, which may offset some of the additional license consumption created by AI agents. Synopsys’s current monetization model is nevertheless attractive because DSO.ai consumes many conventional back-end licenses, effectively turning automation into additional tool demand.

Subscribe now

Read more

Copart Investor Call; July 6, 2026

Copart’s special call was essentially a “Jay Adair is back” message, with the incoming CEO trying to reset the investor narrative around growth, competition, and execution. Adair emphasized that his return is not temporary: “my intent is to lead the company for the next 10-plus years”.

Adair outlined three growth pillars: international expansion, domestic whole-car expansion, and technology services for customers. He repeatedly pushed back against what he sees as a competitor-driven narrative that Copart’s customer relationships have weakened, saying, “I believe our customer relationships are stronger than ever,” while also saying he does not intend to “put my playbook for winning out there publicly.” Adair acknowledged “there is a unit loss” tied to an account that was lost, while also suggesting Copart partly chose not to do certain business. His counterargument is that this does not indicate a broad deterioration in customer relationships and that Copart can win business back over time.

Copart says its global buyer network spans more than 160 countries, and Adair argued that international buyers, crossover buyers, and finance buyers are critical to the record auction returns Copart generates. He framed this liquidity as self-reinforcing: sellers send Copart more vehicles because Copart produces better returns, and those volumes deepen buyer liquidity further.

Total loss frequency remains a secular tailwind, with management noting it reached roughly 23.6%, up nearly five points over four years, helped by rising repair costs and strong auction returns that make total-loss decisions more attractive to insurers. At the same time, near-term insurance volume pressure is being attributed to a cyclical consumer retrenchment, not a structural break: inflation pushed combined ratios out of balance, insurance rates rose, and consumers shifted toward higher deductibles or liability-only coverage. Adair was explicit that “we believe the consumer retrenchment is cyclical, not structural,” and expects coverage levels to normalize as premiums stabilize.

International is being reaccelerated after a period of slower expansion, especially as Copart worked through the German market model. Adair said the company has now “figured that out,” is profitable there, and will be “firing that machine back up again” across Europe and other markets. Domestic whole-car is another priority, but Adair was guarded on specifics, saying the goal is for the business to look “very different in 3 to 4 quarters.” The opportunity seems to be taking Copart’s buyer liquidity and applying it more aggressively to noninsurance sellers such as dealers, banks, and finance companies.

Subscribe now

Read more

The Circuit: Apple’s Siri AI, Nebius Inflection, China Trip Takeaways; June 15, 2026

The most important Apple insight is that the speakers view Siri’s opportunity not as competing directly with Gemini or ChatGPT for broad web search, but as turning AI into a productized, system-level feature. Bajarin says Apple’s strength is “integration at the systems level,” where the user can ask practical questions like finding a report in messages or pulling flight information from email. Goldberg frames this as “making a product out of AI,” arguing that most consumers do not care whether something is labeled AI; they care whether the device can do something useful that it could not do before. Their shared point is that Apple’s AI moat, if it exists, will come from owning the device, the operating system hooks, the private personal index, and the user experience rather than having the world’s best general-purpose model.

A related strategic point is Apple’s attempt to clarify how much of the new Siri is “Apple” versus Google/Gemini. Bajarin says Apple went out of its way to emphasize that its on-device models, broad world knowledge, and contextual Siri work are not simply white-labeled Gemini, even if the cloud model appears to use some base Gemini technology that Apple fine-tuned or adapted. Goldberg’s interpretation is that this is mostly brand protection: Google is not perceived as privacy-friendly, while Apple’s AI pitch depends on trust. The speakers also highlight Apple’s use of Nvidia and Intel for confidential compute, which Bajarin views as notable because Apple historically preferred to control silicon and had long-running bad blood with Nvidia. The implication is that Apple may use Nvidia infrastructure for now, but Goldberg expects Apple eventually to build its own inference accelerator if the scale and economics justify it.

The Nebius discussion centers on whether NeoClouds are real long-term cloud businesses or merely temporary arbitrage vehicles around GPU scarcity and power access. Bajarin comes away more positive on Nebius, saying it looks “closest-ish to a hyperscaler” among the NeoClouds because it has more of a software stack than peers and a deeper technical bench. Goldberg adds that the core question for all NeoClouds is whether they are “real companies” or just “landlords or short-term arbitrage in electricity and GPU availability,” and he says Nebius and CoreWeave are among the few that look capable of building sustainable businesses. Nebius’ Yandex heritage matters here: Goldberg notes that Yandex was a legitimate tier-two hyperscaler with strong engineering talent, giving Nebius more credibility than a pure GPU leasing startup.

The growth opportunity for Nebius is not just training clusters but broader AI infrastructure, including software workloads, dedicated CPU racks for inference, and eventually physical AI/robotics infrastructure. Bajarin says he pressed Nebius on whether it was hearing demand for CPUs, and management indicated it is preparing to scale both Nvidia Vera and Intel CPU racks. He also highlights Nebius’ claim that it can design more efficient racks than standard OEM machines through better networking, wiring, power handling, and latency optimization. This is important competitively because, in his view, AI infrastructure companies need to prove they are technical operators, not merely financiers of GPUs and power contracts. The risk, however, is still that most enterprise customers will remain multi-cloud, with some workloads staying on hyperscalers and only certain AI workloads going to NeoClouds.

Goldberg’s China trip produces a blunt takeaway: U.S. export-control policy, especially the Entity List, is not working as intended. He says the Entity List once created real fear among Chinese companies, but now “nobody’s afraid of it in the same way” because companies such as YMTC have survived and, in his telling, are now “thriving.” The most striking example is a Chinese company on the Entity List that still sources roughly 90% of its critical components from major U.S. technology suppliers, which leads Goldberg to ask what the Entity List is actually accomplishing if U.S. companies can still sell into restricted entities. He also notes that enforcement capacity is weak, claiming there are only five BIS inspectors in China and that they are highly constrained. The risk for U.S. policy is that restrictions may accelerate domestic substitution in China without actually stopping the flow of key components.

InPractise Interview with Former Group CFO, Nebius Group; June 9, 2026

The speaker frames Nebius less as a hastily assembled AI-infrastructure play and more as the surviving AI core of the old Yandex: after the forced Russia disentanglement, the company kept Nebius, Toloka, AVRide, TripleTen, and a large remaining stake in ClickHouse, while selling the Russian and gray-zone assets. The strategic origin story matters because management wants investors to see Nebius not as a GPU reseller, but as an engineering-heavy company that inherited a culture of building search, cloud, fintech, self-driving, and AI systems in-house.

They contrast that background with CoreWeave’s crypto-mining heritage, arguing that Nebius has a more “full-stack” builder mentality and therefore more control over costs, supply chain, deadlines, and customer experience. One of the most important early strategic choices was the clean break from Russia: the company hired a Big Four firm to verify that it had “no Russian customers, no Russian suppliers, nothing,” and Nvidia apparently required that proof before investing.

Subscribe now

Read more

The Jason & Scot Show: The Non-Agentic Episode; June 18, 2026

The biggest story in the episode is TikTok Shop. Jason calls it “the most important story in commerce in the U.S. this year,” noting that TikTok has roughly 100 million daily U.S. users spending an average of 54 minutes per day on the platform. TikTok Shop reportedly sold $9 billion in its first year, $15 billion in its second year, and is on track for roughly $23 billion to $24 billion in the US in 2026, making it one of the fastest-growing retail platforms in history. Scot points out that this is roughly 48% year-over-year growth, far above the broader e-commerce growth rate.

TikTok Shop’s threat is not only the GMV it directly captures, but the demand it creates elsewhere. Jason argues that for every dollar TikTok collects on-platform, it may generate another $4 of purchases that happen later on Amazon, Walmart, or in stores. He says the old retail discovery model was “SIS,” or “saw in store,” where shoppers discovered new products on physical shelves, but TikTok is replacing that discovery layer. That is a major strategic risk for retailers and brands that still think the store shelf or the search box is the main point of demand creation.

TikTok also changes the marketing playbook. Jason argues that traditional paid ads and celebrity influencer marketing do not work the same way on TikTok; brands need sampling, seeding, affiliate incentives, and many small creators rather than a few huge celebrities. He says the path to success is sending free product to a large number of very small creators, hoping content propagates up from tier-four creators to tier-one creators. This is a new kind of performance funnel, and brands that treat TikTok as just another ad channel may miss the point.

Subscribe now

Read more