Thinking Machines Lab has struck a long-term strategic partnership with Nvidia that gives the startup both fresh backing and access to an enormous future supply of AI compute, marking one of the biggest infrastructure moves yet by a young artificial intelligence company. Nvidia and Thinking Machines said the agreement includes a “significant investment” from Nvidia and a plan for Thinking Machines to deploy at least one gigawatt of Nvidia’s next-generation Vera Rubin systems.
The deal is a major step for the San Francisco-based startup founded by former OpenAI chief technology officer Mira Murati. On its website, Thinking Machines describes itself as an AI research and product company focused on giving people access to tools and knowledge that can be adapted to their own needs. Nvidia said the companies will also work together to optimize Thinking Machines’ products and models for Nvidia’s hardware and software stack.
Murati framed the partnership as central to the company’s ability to build and scale its systems. In statements released by the companies, she said Nvidia’s technology underpins much of the AI field and that the new arrangement will expand Thinking Machines’ capacity to develop AI products. Nvidia chief executive Jensen Huang, in the same announcement, said Thinking Machines had assembled a strong team to push frontier AI forward.
The scale of the compute commitment is what makes the agreement stand out. Axios reported that a gigawatt of AI compute is a threshold typically associated only with the largest frontier labs, suggesting Thinking Machines is aiming to compete at the top end of model development rather than remain a smaller application-layer company. That interpretation is reinforced by the companies’ description of a multiyear partnership built around Rubin systems, which are part of Nvidia’s next wave of AI infrastructure.
The agreement also builds on a startup that had already attracted unusual attention. Wired reported in 2025 that Thinking Machines raised a record $2 billion seed round at a $12 billion valuation, with backing from Andreessen Horowitz, Nvidia, Accel, Cisco, and AMD. That financing round immediately made the company one of the most closely watched entrants in the AI startup race.
Murati has been central to that attention from the beginning. Wired identified her as the company’s chief executive and tied the startup’s founding team to a group of former OpenAI researchers and leaders. That founding roster included Barret Zoph and Luke Metz, both of whom had previously worked at OpenAI and were named as co-founders of Thinking Machines.
But the company has also gone through visible personnel changes. In January, TechCrunch reported that Murati announced the departure of Barret Zoph, who had been the company’s co-founder and CTO. The same coverage identified Luke Metz as another co-founder and noted that he had also previously worked at OpenAI. Axios later reported that Zoph and Metz had rejoined OpenAI, making the Nvidia deal arrive against a backdrop of leadership and talent turnover.
Even with those departures, Thinking Machines appears to be expanding rapidly. Axios reported that the company has grown from roughly 30 employees a year ago to about 120 today. The startup has publicly released one product, Tinker, but has said relatively little about the full scope or timeline of its broader model plans, which makes the Nvidia partnership one of the clearest signals so far about how aggressively it intends to scale.
The Nvidia deal also highlights the increasingly tight relationship between chip supply, capital, and competitive positioning in the AI market. Nvidia is not just selling hardware to fast-growing model companies. It is also investing in them, which gives companies like Thinking Machines both capital and a pathway to scarce compute at a moment when access to training infrastructure can determine which labs stay in the top tier. Wired’s earlier reporting already showed Nvidia among the startup’s investors, and the new partnership deepens that connection.
With Murati leading the company, Zoph and Metz still part of its early story, and Nvidia now committing both capital and Rubin-scale compute, Thinking Machines has moved further into the front rank of AI startups trying to challenge OpenAI and Anthropic. The partnership does not answer every question about what the company will ship next, but it does make one thing clearer: Thinking Machines is building for scale, and it now has one of the industry’s most important suppliers helping it do it.
Read our coverage of OpenAI’s potential NATO deal for more on how AI companies are extending their reach into defense, infrastructure, and strategic government work.
