Data Industry Antitrust

Billion-Dollar Bets on Uncertain Ground: AI Valuations Face a Reality Check

Mira Murati's new lab, Thinking Machines, is reportedly nearing a $5 billion funding round led by Accel at a $40 billion valuation — roughly one year after Murati departed OpenAI, where she served as chief technology officer, and without any publicly released product. The valuation is a founder bet: investors are purchasing Murati's judgment, network, and credibility from years at the center of some of the most consequential AI decisions of the past several years. Accel, whose early bets include Facebook, Slack, and Dropbox, anchoring the round lends the thesis institutional credibility.

In contrast, Anthropic is reportedly pursuing an initial public offering at a $1.5 trillion valuation, and investors are balking. That figure would place Anthropic alongside Apple, Microsoft, and Nvidia — companies generating hundreds of billions in annual revenue. Anthropic's revenue, while growing rapidly, is nowhere near that scale. The bull case requires believing that AI will restructure the global economy and that Anthropic will capture a substantial share of the resulting value — a claim requiring many things to go right simultaneously.

An outside study finding that Anthropic's AI is improving twice as fast as competitors makes that bull case more legible, since a steeper improvement curve over eighteen months would produce a widening capability differential that competitors could not easily close. But investor skepticism about the $1.5 trillion figure has deepened in the days since the Democratic sweep, as a Congress that may impose compliance costs, liability frameworks, or operational restrictions shifts the risk calculus further against the bulls.

The ChatGPT lawsuit filed by James Gates adds legal uncertainty to the valuation environment across the sector. Gates alleges that ChatGPT provided his wife with advice on using oleander as a poison and concealing evidence; criminal charges against the wife were dropped in April, but the civil suit is proceeding. The case raises unresolved questions about AI company liability when a model provides harmful information a user then acts on — territory where Section 230, designed for platforms hosting user-generated content rather than systems generating novel responses, may not apply cleanly. If the suit advances far enough to establish legal principles, it could create precedent reshaping what AI companies must prevent their models from discussing, with consequences for every valuation in the sector.