Mira Murati’s Startup Lands Huge Nvidia Chip Deal

For the past few months, the easiest story to tell about Mira Murati’s Thinking Machines Lab was the messy one. A new deal with Nvidia, however, paints a different picture of the company’s future plans.

A cofounder left for Meta in November 2025. Then, on Jan. 14, 2026, cofounders Barret Zoph and Luke Metz left to rejoin OpenAI, along with researcher Sam Schoenholz, in a high-profile talent shake-up reported by WIRED. For a young frontier lab whose biggest asset is basically “some of the smartest people in AI all agreed to work here,” that’s not a small bruise. That’s the kind of thing that makes investors sweat, rivals smirk, and every group chat in San Francisco start typing at once.

And yet, today’s Nvidia partnership announcement changes the tone of that story in a major way.

Not because it proves the drama didn’t matter. It did. But because it suggests Thinking Machines has managed to do the one thing that matters most in frontier AI after losing talent: keep moving fast enough that the market still treats you like a serious contender.

The headline is almost cartoonishly big. Thinking Machines says it has signed a multi-year partnership with Nvidia to deploy at least one gigawatt of next-generation Vera Rubin systems, with deployment targeted for early 2027. Nvidia is also making what both companies describe as a significant investment.

Translation: this is not “nice startup, good luck out there.” This is Jensen Huang putting real chips, real capital, and real reputational weight behind Murati’s lab.

That matters because in 2026, Compute is credibility.

We’re long past the phase where frontier AI competition is mainly about who has the best demo or the prettiest benchmark chart. The game now looks a lot more like the compute land grab we’ve been writing about: whoever secures chips, power, datacenter capacity, and the right systems talent gets a seat at the adults’ table. Everyone else is doing a very expensive cosplay of a Frontier Lab.

Thinking Machines just bought itself a much more convincing chair.

Why this deal landed differently

The timing is the story. If this deal had landed before the departures, it would’ve read like a straightforward “Murati startup gets major infrastructure partner” announcement. Landing it after months of visible turbulence makes it something else: a rebuttal. It tells the market that, whatever internal damage those exits caused, Nvidia still believes this company is worth betting on at scale.

That doesn’t automatically mean Thinking Machines is thriving internally. We should be careful not to confuse a massive partnership with a clean bill of health. Big compute commitments don’t ship products, train culture, or replace founding researchers who walk out the door. They do, however, make one thing clear: Thinking Machines has not been relegated to the “interesting but fragile” bucket.

In fact, Axios reported today that the company has grown from around 30 employees a year ago to roughly 120 now, with more hires than departures, according to a source close to the company. If that holds, the more accurate read here is not “Thinking Machines survived a collapse.” It’s “Thinking Machines took a punch, then kept hiring, kept building, and just locked in one of the biggest compute signals in the industry.”

That also fits the product breadcrumbs we’ve seen so far. Thinking Machines hasn’t fully shown its hand on frontier models yet, but it has been steadily building out Tinker, its training platform for researchers and developers. The company’s pitch has consistently centered on customizable AI, open models, and tools that let people shape systems rather than just consume them. The Nvidia announcement reinforces that vision: bigger infrastructure not just for raw model training, but for serving platforms users can adapt to their own needs.

Infrastructure can cover a lot of turbulence

So yes, this looks like them pushing through.

But there’s a second, slightly less flattering interpretation, too: Thinking Machines may be proving that in the current market, infrastructure can paper over a lot. In AI’s talent war, elite researchers still matter enormously, as we saw when OpenAI leaned harder into compensation and recruiting tactics, such as dropping the vesting cliff. But the labs that win are increasingly the ones that can replace individual departures with institutional momentum. If Nvidia is willing to back you with future Rubin-scale capacity, the market may forgive a lot of organizational chaos.

That is both impressive and a little unnerving.

Because what this really signals is that the frontier race is becoming less about whether a startup can avoid stumbles and more about whether it can stay financed, supplied, and believable through them. Murati’s lab appears to have cleared that bar, at least for now. Losing big names in January could have turned Thinking Machines into a cautionary tale about how hard it is to peel talent away from OpenAI and keep them. Instead, by March 10, 2026, the company has managed to force a different question:

What if the departures were a setback, but not a derailment?

That’s the real significance of this deal. It doesn’t prove Thinking Machines has solved the hard part. It proves the company still gets to attempt the hard part at frontier scale.

And in this market, that alone is a very big deal.

Editor’s note: This content originally ran in the newsletter of our sister publication, The Neuron. To read more from The Neuron, sign up for its newsletter here.

The post Mira Murati’s Startup Lands Huge Nvidia Chip Deal appeared first on eWEEK.