Thinking Machines Lab: $50B for a Company That Has Not Shipped Anything
February 21, 2026 by Asif Waliuddin
Mira Murati left OpenAI in November 2024. She founded Thinking Machines Lab. By July 2025: $2B raised at $12B valuation. By February 2026: reportedly in talks at $50-60B.
That is a 4-5x valuation jump in seven months. From stealth. With no public product, no demos, and no published research.
For comparison: Anthropic is valued at approximately $60B. Anthropic has Claude, enterprise contracts, HIPAA compliance, a healthcare vertical, and millions of active users. They have shipped real products to real customers generating real revenue.
Thinking Machines Lab is approaching the same valuation without having shipped any of that.
Here's what most people are missing:
-- This is not an anomaly. It is the market. xAI is at $230B. Skild AI tripled to $14B in one round. The AI funding environment is pricing narrative and pedigree at parity with product and revenue. Murati's credibility as OpenAI's CTO during the GPT-4 era is being valued the same as Claude's entire product suite.
-- The implicit bet is that talent and capital will reliably produce frontier capability. That bet might be correct. Murati is not a random founder -- she ran technical operations at the most influential AI lab in the world. But "probably will build something great" being worth the same as "already built something great" tells you where the risk pricing is in this market.
-- The gap between AI startup valuations and AI startup evidence is now large enough to call structural. It is not that any individual company is overvalued (maybe they are, maybe they are not). It is that the entire category is pricing in outcomes that have not occurred, at levels that assume near-certainty of success.
There is a word for when the entire market assigns near-certain probability to uncertain outcomes. I will let you fill it in.
The question is not whether Murati will build something impressive. She probably will. The question is whether a stealth-mode company with zero public evidence of product-market fit should be valued identically to a company that has already demonstrated it. If the answer is yes, then valuation has decoupled from evidence entirely, and we should stop pretending otherwise.
Follow for more AI Hype vs Reality takes.