One gigawatt. That's the amount of computing power - enough to light up 750,000 homes - that Mira Murati's Thinking Machines Lab just locked in from Nvidia.
The multi-year deal announced today pairs the former OpenAI CTO's startup with Nvidia's next-generation Vera Rubin chip platform, with deployment starting early 2027. Nvidia is also making a "significant investment" in the company, though neither side disclosed how much.
For context, a gigawatt of AI compute is a threshold only the largest labs in the world have approached. This isn't a startup buying cloud credits to fine-tune someone else's model. Murati is signaling she intends to train frontier models from scratch, putting Thinking Machines in direct competition with her former employer OpenAI, plus Anthropic, Google DeepMind, and xAI.
From 30 to 120 in a Year
Thinking Machines Lab has grown from roughly 30 employees to 120 since Murati founded it after leaving OpenAI. The company says it's focused on building AI that's "understandable, customizable, and collaborative" - corporate language, but the compute commitment tells a clearer story. You don't buy a gigawatt to build chatbot wrappers.
Nvidia CEO Jensen Huang called AI "the most powerful knowledge discovery instrument in human history" in the announcement. Murati said the partnership "accelerates our capacity to build AI that people can shape."
The Compute Arms Race Continues
This deal fits a pattern. Nvidia has been making strategic investments in AI labs alongside compute supply agreements, effectively locking in its biggest customers while giving startups the hardware they need to compete with deep-pocketed incumbents. For Nvidia, it's a guaranteed revenue stream. For Thinking Machines, it's a credibility signal to future investors and recruits that the company is playing at the highest level.
The real question is what Thinking Machines actually ships. Having compute is necessary but not sufficient - OpenAI, Google, and Anthropic have spent years building training infrastructure, safety processes, and distribution channels. Murati has the technical pedigree and now the hardware. Turning that into products people use is the hard part.