On January 22, 2026, The New York Times published a long reconstruction of the turmoil inside Thinking Machines Lab. The company, founded in February 2025 by former OpenAI CTO Mira Murati, had raised $2 billion within a year at a $12 billion valuation. Now every co-founder besides Murati is gone, and a new round targeted at a $50 billion valuation has stalled. This is more than Silicon Valley gossip — it is a firsthand slice of what the AI talent war and valuation discipline actually look like in 2026.
The drama is best told as a timeline: three early principals confronted Murati over control, the CTO was fired two days later, and OpenAI announced it had scooped up all of them 58 minutes after Murati’s public post.
The January 12 Showdown
On January 12, researcher Sam Schoenholz set a meeting with Murati, joined by CTO Barret Zoph and co-founder Luke Metz — all four were OpenAI alumni. Per the NYT, the three were unhappy with the company’s direction: products were lagging rivals, and the fundraise at a $50 billion valuation was going nowhere. They demanded that Zoph be given direct authority over technical direction, and threatened to leave otherwise. Murati’s reply was blunt: you are already the CTO — do your job.
The Firing, and OpenAI’s Immediate Rehiring
On January 14, Murati fired Zoph — the NYT reports he had been talking with Sam Altman about a return since October. OpenAI barely hesitated. Fifty-eight minutes after Murati posted on X that the company had “parted ways” with Zoph, OpenAI’s CEO of applications Fidji Simo announced that Zoph, Metz, and Schoenholz were all coming back. Within a company of roughly 100 people, about nine more have left or received OpenAI offers, and Meta has made offers to some staff worth hundreds of millions of dollars. Murati promoted Soumith Chintala, co-creator of PyTorch, to CTO. On January 21, OpenAI named Zoph head of its business selling AI services to enterprises.
A Late Product and a Strained $50 Billion Round
Thinking Machines closed its $2 billion seed round in July 2025 — led by Andreessen Horowitz, with Accel, Nvidia, AMD, and Jane Street participating — at a $12 billion valuation, before it had shipped anything. Its first product only arrived in October: a service that helps developers fine-tune models for specific tasks, competing directly with similar offerings from Google, Amazon, and Microsoft. Since then, the round aimed at a $50 billion valuation has struggled, with some investors privately questioning whether the company can survive as an independent. The contrast with the same week’s news that Mistral’s CEO expects revenue above €1 billion in 2026 is stark: valuations without a revenue story are getting harder to sell.
Meta’s Overtures and the Anthropic Thaw
The NYT also reconstructed the earlier fault lines. Last summer, Meta executives made contact about a possible acquisition; Zoph argued for selling, while Murati judged the approach unserious and insisted on independence. Co-founder Andrew Tulloch also pushed for a sale and left for Meta in October. Meanwhile, Murati has grown closer to Anthropic CEO Dario Amodei — the two communicate regularly, and she has long been friends with his wife. Secret meetings, collapsed deal talks, a power struggle, and a firing: before the company could prove its product, it proved its governance risk.
Lessons From the Talent War
Venky Ganesan, a partner at Menlo Ventures, gave the NYT an Anna Karenina framing: happy companies are all alike; every unhappy company is unhappy in its own way. For anyone building an AI team now, three lessons stand out. First, the opportunity cost of top researchers is a nine-figure package at a rival lab — equity narratives do not retain people who already want to go back to OpenAI. Second, the “elite team plus unannounced product” model for ten-billion-dollar valuations is breaking down; investors now look at revenue and shipping cadence. Third, who ultimately decides technical direction and whether to sell are governance questions that must be settled on day one — not discovered in a showdown meeting.
Sources
- The A.I. Start-Up Soap Opera Riveting Silicon Valley — The New York Times
- Mira Murati’s startup is losing two of its co-founders to OpenAI — TechCrunch
- Mira Murati on X
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.