Weekly Intel - 2026-05-31

A lot happened this week in AI and tech. Capital keeps concentrating into a smaller number of bets, enforcement is getting real in Europe, and a couple of industry leaders are publicly admitting they got the AI jobs story wrong.

AI Industry Moves

Anthropic raises $65B in Series H at a $965B post-money valuation Anthropic is now worth nearly a trillion dollars as a private company, more than OpenAI on paper, and it got there in roughly three years. The $65B raise comes with $47B in annual revenue, up from roughly $10B a year ago. Most of that growth is coming from Claude Code and developer adoption. The AI companies pulling ahead got embedded in engineering workflows early, where switching costs build fast — model performance is barely the differentiator anymore. If you’re integrating AI into your operations right now, it’s worth thinking about what leverage you’ll have over your vendor in two years. You probably won’t have much.

OpenRouter raises $113M Series B OpenRouter pulled in $113M with a backer list that looks like a map of the modern data stack: CapitalG, NVentures, ServiceNow, MongoDB, Snowflake, Databricks, a16z, Menlo. Weekly token volume jumped 5x in six months to 25 trillion. The bet is on the routing layer, the piece that sits between developers and 400+ AI models and decides where requests actually go. As the models themselves start to look more similar in capability, the companies controlling that routing and reliability layer may end up with more long-term leverage than anyone expects.

Tech Industry

Dropbox CEO Drew Houston to step down Drew Houston is handing the reins to product chief Ashraf Alkarmi after 19 years as founder-CEO, with a co-CEO overlap period before Houston moves to executive chairman. Houston built something genuinely durable: a profitable cloud storage business while competing against Google and Apple, which is harder than it sounds. The open question is whether Alkarmi actually moves the product or whether Dropbox stays a solid, boring utility most people barely think about. I’ve noticed that founder-to-operator transitions in tech either start a second act or quietly signal the founder saw the ceiling first. Hard to tell which this is yet.

Uber president says AI spending is getting ‘harder to justify’ Uber burned through its entire annual AI budget four months into 2026. Its president Andrew Macdonald is now saying publicly what plenty of other leaders are probably thinking privately: token consumption keeps climbing but nobody can draw a clear line to useful features actually shipping. R&D hit $3.4 billion in 2025, up 9% year over year, and the best Macdonald could offer was “maybe implicitly” more is getting done. This is the first major consumer tech company to publicly frame AI investment as a justification problem rather than a capacity or pace problem. Whether that triggers a wider reset in how companies budget for AI tooling, or stays a one-off while everyone else keeps spending — I genuinely don’t know. But Uber being the first to say it publicly matters.

Accenture to acquire Ookla Accenture is buying Ookla, best known for Speedtest. The interesting part isn’t the brand, it’s the data underneath it. Ookla captures over 1,000 attributes per network test, and Accenture wants that connectivity intelligence as a foundation for AI-driven services across banking, utilities, retail, and government. I hadn’t expected network observability to move this fast as a strategic asset. As AI workloads scale, knowing exactly how the network performs at the device and application level starts to matter in ways it didn’t before. I’d expect Google and AWS to be thinking about the same play.

Legal & Regulation

EU fines Temu €200M for allowing sale of illegal products The European Commission hit Temu with a €200M fine after mystery shopping found widespread product safety failures: chargers that couldn’t pass basic electrical checks, baby toys with chemicals above legal limits. Temu has until late August to present a remediation plan. Brussels has been moving from warnings to actual enforcement faster than most platforms anticipated, and if your business depends on third-party marketplace distribution in Europe, the compliance expectation just moved. Amazon, Shein, and AliExpress are probably paying close attention.

Economy & Labor

Sam Altman and Dario Amodei are walking back AI jobs apocalypse predictions Altman now says he was “pretty wrong” about near-term displacement of white-collar roles. Amodei, who once predicted 50% job losses, is suggesting AI may actually expand the scope of human work. I’m not surprised by the reversal. The people building these systems have consistently overestimated how fast disruption arrives and underestimated how slowly organizations absorb new technology. What I’m more interested in is the second-order effect: a lot of companies used those original predictions to justify hiring freezes and restructuring. Whether those decisions get revisited now that the thesis has shifted is a separate question, and I don’t see anyone talking about it yet.

AI & Software

Claude Opus 4.8 The benchmark improvements on coding and reasoning are incremental, but the packaging around them is more interesting: user-controlled effort levels, a “dynamic workflows” mode in Claude Code for large-scale problems, and a 3x price cut on fast mode. Anthropic is moving from selling model intelligence toward selling configurable AI labor, where you can tune cost, speed, and depth per task. That’s the kind of control enterprises actually need to integrate AI into real workflows rather than just run experiments. I’ll be curious how fast the other foundation model providers follow.

YouTube to automatically label AI-generated videos Starting this month, YouTube will automatically detect and label AI-generated photorealistic content rather than relying on creators to self-disclose. Content made with YouTube’s own AI tools or carrying C2PA metadata gets permanently tagged. Creators can contest incorrect labels. Platform-level AI detection is becoming infrastructure, not just a stated policy. For any business producing video at scale: your AI-generated content will get flagged eventually. Whether your audience cares is the more interesting question, and I don’t think most producers have actually tested that yet.