Thinking Machines strikes major infrastructure deal with NVIDIA
Thinking Machines, the AI company led by former OpenAI executive Mira Murati, announced a partnership with NVIDIA to deploy a massive 1-gigawatt AI infrastructure stack built around Vera Rubin systems for frontier model training. The company also disclosed an NVIDIA investment, though financial terms were not provided.
Frontier training at gigawatt scale
The collaboration centers on building and operating high-density compute designed for training next-generation models. A 1GW deployment places the effort among the most ambitious AI compute buildouts disclosed to date, reflecting intensifying competition among labs and platform providers to secure reliable access to advanced accelerators, networking, and power.
NVIDIA’s Vera Rubin platform is expected to underpin the infrastructure, positioning the partnership as both a technology and capacity play. While the companies did not specify timelines, locations, or the number of systems to be deployed, the scale suggests multi-site development and extensive supporting infrastructure, including cooling and grid interconnects.
Investment to scale custom AI systems
In addition to the infrastructure partnership, NVIDIA will provide an undisclosed investment in Thinking Machines. The funding is intended to help the company scale custom AI systems, an area where demand is rising as enterprises seek specialized models, optimized training pipelines, and integrated deployment stacks rather than one-size-fits-all offerings.
The announcement underscores a broader trend: leading AI developers are increasingly pairing capital with long-term compute arrangements to reduce supply uncertainty and accelerate product roadmaps. For Thinking Machines, the deal signals an intent to compete at the highest tier of model development by securing both cutting-edge hardware and the financial backing needed to operationalize it.
Related
