Mira Murati's Thinking Machines Lab signs a chip supply deal with Nvidia worth tens of billions of dollars, planning to deploy 1GW+ of next-gen Vera Rubin chips
Multiyear partnership includes a further ‘significant’ investment from Nvidia into one-year-old AI start-up
Financial Times
Context & Ripple Effects
Thinking Machines Lab moved from a $2 billion seed round that included Nvidia to a multiyear supply relationship in which Nvidia is also making an additional investment. The arrangement turns an early financial tie into a major infrastructure commitment for the young company.
The reported 1GW-plus Vera Rubin deployment places TML among the AI builders securing capacity well ahead of use. Its later Google Cloud agreement for Nvidia-based systems indicates that the company is assembling compute access through more than one channel rather than relying on a single delivery model.
First-order effects
Thinking Machines Lab gains a committed path to more than 1GW of next-generation Nvidia capacity, while Nvidia gains a large multiyear customer commitment and a deeper financial stake in the startup.
The deal ties TML's near-term infrastructure plans to Nvidia's Vera Rubin rollout, making execution of that platform consequential for TML's planned compute buildout.
Second-order effects
A commitment of this scale can tighten the pool of next-generation capacity available to other AI developers, increasing the incentive for rivals to reserve supply early or use cloud capacity and alternative hardware strategies.
Nvidia's supplier-and-investor role gives it more influence over TML's infrastructure choices; TML's cloud deal shows how startups may combine direct procurement with cloud-based access to meet large compute needs.
Third-order effects
If similar arrangements persist, frontier AI development will increasingly be shaped by long-duration capacity contracts and capital relationships, not just by access to chips bought on demand.
The pattern concentrates leverage with firms able to finance, manufacture, and operate large-scale compute, while pushing smaller developers toward cloud intermediaries or more heterogeneous compute stacks.
The trend: AI startups are increasingly converting venture backing into long-term, utility-scale compute commitments with the chip suppliers and cloud providers that control scarce infrastructure.
Thinking Machines Lab and NVIDIA just announced a massive partnership to build a 1-gigawatt AI supercomputing cluster using the upcoming Vera Rubin chips. This project focuses on training the next generation of giant AI models while making them easier for regular people and [imag…
Grateful to Jensen and @nvidia team for their support. Together, we're working to deploy at least 1GW of Vera Rubin systems, bringing adaptable collaborative AI to everyone. https://thinkingmachines.ai/ ... [image]
We have entered a world where compute has never been so quickly converted to value... (if you can get your hands on it) World class partnership. The future is beautiful. Congrats @miramurati and @thinkymachines.
a thriving ecosystem is good. genuinely. it's nice to see people like jerry, mira, ilya, and yann trying to build with a fraction of the budget of openai and anthropic. but i don't really see the vision. they lack the five things that matter most: compute research taste
Excited to partner with NVIDIA. bringing up 1GW or more of compute starting with Vera Rubin, co-designing systems and architectures together, and more. NVIDIA has also made a significant investment in @thinkymachines