Anthropic made money last quarter. It's the first AI lab to do it.

Hi AI Futurists,

For three years the standard line on frontier AI was that nobody makes money on it, and the numbers backed that up. Anthropic just filed a quarter that breaks the pattern, eight weeks before it goes public. Let's take a look.

Our agenda.

Best, Lex Sokolin

P.S. Hit reply with suggestions or feedback!

Sign up for a free trial of Attio, the agentic CRM.

Ask Attio to build a daily workflow that surfaces the deals that need your attention today, like anything with a stage change, a recent reply, or a new signal in the last 24 hours.

Review your pipeline in Claude, synced live from Attio via MCP.

That's it.

Try Attio Now

🏗️ Nvidia is guaranteeing $105 billion of OpenAI's rent in Ohio (Data Center Knowledge). OpenAI signed a 20-year lease for 8 gigawatts on an old uranium enrichment site, and its chip supplier cosigned. July reports had the guarantee at $250 billion.

⚡ Etched doubled its valuation to $21 billion in a month (Unite.AI). Jane Street led the $700 million round after buying one of the racks and running it in its own datacenter. Best due diligence there is.

✈️ Google paid $10 million for a dead airline's inbox (The Register). Spirit's liquidation handed over 100 million emails, 500 million Teams messages, and 30 million recorded service calls. Mercor bid $7.5 million and lost.

🧒 OpenAI launched a separate ChatGPT for teenagers (ABC News). Anyone the system reads as 13 to 17 gets routed into it, with homework answers replaced by guiding questions. Roughly four years after teenagers started using the old one.

⚖️ Twenty-nine states took Meta to trial over teen harm (NPR). Opening statements ran Tuesday in Oakland federal court, with California, Colorado, Kentucky, and New Jersey arguing for the group. Six to seven weeks of internal documents ahead.

🚕 Pony.ai has more than 4,000 robotaxis contracted outside China (Automotive World). Uber alone accounts for over 2,000 of them across Europe, with the rest spread across Southeast Asia, the Middle East, and Luxembourg. Permits are the bottleneck now, not the cars.

🎬 ByteDance signed an AI copyright pact with Hollywood (NBC News). The deal with the Motion Picture Association covers Seedance and Seedream across TikTok, CapCut, and Dreamina. In February the same trade group was sending cease-and-desist letters.

📈 Samsung raised advanced chipmaking prices by up to 15% (The Next Web). Its 4nm and 5nm lines are full because TSMC's are full, and Chinese customers took the steepest increases. A business that has lost money since 2022 just found pricing power.

🇨🇳 Nvidia's H200s are trickling into China at 13% of the legal limit (TECHi). ByteDance and Tencent took roughly 10,000 each against a licensed ceiling of 75,000 apiece. Washington set the cap, and Beijing is the one keeping the orders small.

🎧 Apple's camera AirPods showed up in shipping code, then slipped to 2027 (9to5Mac). A demo video buried in a macOS release candidate shows the B790 buds feeding Siri a live view of the room. Finished enough to film, not finished enough to sell.

Anthropic booked $11.5 billion of revenue in the second quarter and came out of it with positive adjusted operating income, which no frontier AI lab had managed before. Revenue was $4.73 billion in the first quarter and $787 million in the same quarter last year, a fourteen-fold jump in twelve months that puts first-half booked revenue at $16.2 billion. The annualized run rate reached $65 billion by the end of July, up from $9 billion at the close of 2025.

Where that money comes from explains why Anthropic got here and OpenAI has not. Roughly 80% of its revenue is API and enterprise business, the kind that arrives on a contract and carries reported gross margins above 80%, with Claude Code alone at an $8 billion run rate by May. The losses were never coming from the product, they were coming from training the models, and the top line finally outgrew that bill. OpenAI is running the harder version of the same experiment, a much bigger consumer base against a $40 billion run rate and a projected $14 billion loss this year. Read the fine print, though. These figures are preliminary and unaudited, and Anthropic told investors in May that profitability may not hold through year end as data center commitments come due.

The practical read for the rest of us is that the cheapest era of enterprise AI pricing is behind us. Anthropic lists in October at a price shareholders have floated near $2 trillion, which would be the largest market debut on record, and a public company acquires an earnings call and a share price that answers to margin. Anyone whose product economics depend on a model provider staying generous should price the next contract accordingly. Three years of arguing about whether any of this can pay for itself just lost the best evidence for no, and the answer turned out to be dull: sell to businesses, keep your margins, wait.

Astute: B2B creator marketing that runs itself

For marketers who know creator partnerships work and cannot spend the whole week chasing them.

Taku: borrow someone else's AI setup and make it yours

For the non-engineer who keeps seeing great AI workflows online with no practical way to run them.

Vois: studio-quality voice on your own laptop

For anyone turning articles, decks, or courses into audio and tired of watching a meter run.

This 121-page research report provides the first comprehensive taxonomy of public companies, private ventures, and tokenized protocols building the infrastructure for autonomous AI systems. Compiled by Lex Sokolin, former Chief Economist at ConsenSys, fintech strategist at Autonomous Research, and current Managing Partner at Generative Ventures, this report delivers institutional-grade analysis of 100+ companies across 14 critical infrastructure layers. Learn more here.

That’s all for today, folks!