Thinking Machines Lab has secured a multiyear partnership with Nvidia that includes a significant investment and a commitment to deploy at least 1 gigawatt of the chipmaker’s Vera Rubin compute platform beginning early 2027.
The deal will support frontier model training and platforms designed to deliver customisable artificial intelligence at scale. Nvidia will also co-design advanced training and serving systems optimised for its own architectures as part of the agreement.
What the 1-Gigawatt Commitment Means
A single gigawatt of compute power is equivalent to the electricity needed to power a mid-size city of nearly one million homes. The scale of the deployment signals a substantial infrastructure commitment from Nvidia toward Thinking Machines Lab’s operations.
Thinking Machines Lab Leadership and Background
Thinking Machines Lab is led by Mira Murati, who previously served as Chief Technology Officer at OpenAI. Murati left OpenAI two years ago and became CEO of Thinking Machines Lab in 2025. The company focuses on building AI systems that are customisable and shaped by human needs.
Expanding Access to AI Models
The partnership also targets broader access to both frontier and open-source AI models across industry, academia, and scientific communities. Co-designing systems optimised for Nvidia architectures forms a core component of this objective.
“AI is the most powerful knowledge discovery instrument in human history.”
Jensen Huang, CEO, Nvidia
“The partnership accelerates our mission to build AI that is customisable, collaborative and shaped by human needs.”
Mira Murati, CEO, Thinking Machines Lab
Nvidia’s Investment Strategy in AI Startups
Nvidia’s investment in Thinking Machines Lab follows a pattern of circular deals that have become common in the AI industry. Chipmakers and model developers increasingly exchange capital, compute access, and technical collaboration in structured multiyear agreements.
Moreover, the Vera Rubin platform at the center of this deal represents Nvidia’s next compute architecture, which the company is positioning for large-scale AI training workloads. Deployments of this scale are expected to begin in early 2027, according to the terms of the partnership.
Consequently, the agreement places Thinking Machines Lab among a small group of AI companies with confirmed access to Nvidia’s most advanced compute infrastructure at gigawatt scale. The startup has not disclosed the financial terms of Nvidia’s investment.