Thinking Machines Nears $1B Round at $40B

Analysis

Accel is reportedly in talks to lead a $1 billion round for Thinking Machines, the AI lab founded by former OpenAI CTO Mira Murati, at a valuation of at least $40 billion, The Information first reported, with TechCrunch adding that existing investor Nvidia has also discussed participating in the round.

From Record Seed to a Markdown Ask

• Seed round (early 2025) -- $2 billion raised at a $12 billion valuation, among the largest seed financings ever, led by Andreessen Horowitz with Nvidia, GV, Lightspeed and Conviction Partners participating.
• Late-2025 target -- $50 billion valuation the company was reportedly seeking in funding conversations before this round.
• New round (reported) -- $40 billion valuation, more than 3x the seed mark in roughly 18 months, but below the late-2025 ask.

That markup velocity sits in the same range as AfterQuery's 10x jump and Wonderful's doubling that Pulse has scrutinized elsewhere this year. What's different here is the direction of travel on the ask itself: the number now on the table is actually lower than what the company was seeking less than a year ago -- a rare data point in a market where growth-stage AI valuations have mostly moved one direction.

What Thinking Machines Actually Sells

The company's product is Tinker, a platform for adapting and fine-tuning models that charges usage-based compute fees rather than a flat subscription -- customers pay for the compute they consume adapting models to their own data, a pricing model closer to a cloud provider than a traditional software company. In July 2026, Thinking Machines launched Inkling, an open-weight model that lets customers customize AI systems on proprietary data, Pulse covered at the time as part of a broader small-open-model strategy the company has pursued. Annual revenue run rate is reported above $100 million, which puts this round at roughly 400 times revenue -- an extreme multiple even by 2026 AI-lab standards, though one investors are clearly willing to underwrite given Murati's OpenAI pedigree and the team she assembled around her.

The Departures Worth Noting

Thinking Machines has also lost talent back to OpenAI: co-founders Lilian Weng and Luke Metz both returned to their former employer, a notable reversal for a lab whose entire pitch rested on assembling top-tier ex-OpenAI researchers under Murati's leadership. Losing two co-founders back to the company you were founded to compete with is the kind of detail that would sink a normal seed-stage story; it hasn't stopped this round from happening, which says something about how much investor conviction is riding on Murati specifically rather than the full original team.

The Competitive Field

Thinking Machines competes for mindshare and capital against Safe Superintelligence (Ilya Sutskever's lab), Anthropic, and the open-weight ecosystem increasingly organized around Hugging Face, which Nvidia just agreed to acquire for $12.9 billion. Where Thinking Machines differentiates is Tinker's usage-based model-adaptation angle -- selling infrastructure for customizing models rather than competing head-on to build the single best frontier model, a strategy that sidesteps direct benchmark competition with OpenAI's Astra or Anthropic's Mythos 5 in favor of owning a specific workflow layer.

What's Unverified

Neither Thinking Machines nor Accel has confirmed the round, and "in talks" rounds routinely change size or valuation before signing -- the $50 billion-to-$40 billion shift already shows this specific number has moved once. No customer count, retention data or gross margin has been disclosed alongside the reported figures, the same disclosure gap Pulse has flagged on comparable AI markup stories this year. Whether the round closes at $40 billion, lower, or not at all is the thing to track before treating this valuation as settled.