NVIDIA announced a “significant” investment and a massive multi-year partnership with Thinking Machines Lab, the high-profile AI startup founded by former OpenAI CTO Mira Murati.
The deal effectively secures Thinking Machines as a flagship customer for NVIDIA’s upcoming Vera Rubin chip platform, while providing Murati’s team with the “firepower” to compete with industry giants like Google, Anthropic, and OpenAI.
The Gigawatt-Scale Deal
The most striking part of the announcement is the sheer scale of the hardware commitment.
- 1-Gigawatt Capacity: Thinking Machines has committed to deploying at least one gigawatt (1GW) of NVIDIA’s next-generation Vera Rubin systems.
- Estimated Value: While the specific dollar amount remains undisclosed, industry experts estimate that a 1GW build-out—roughly the power required for 750,000 homes—costs approximately $50 billion.
- Deployment Timeline: Integration is targeted to begin in early 2027, making Thinking Machines one of the first labs to leverage the Vera Rubin architecture.
Strategic Objectives
This is more than a simple sales agreement; it is a co-design partnership aimed at defining the next era of “customizable” AI.
| Objective | Detail |
| System Co-Design | The companies will collaborate on designing advanced training and serving systems optimized specifically for NVIDIA architectures. |
| Frontier Research | The hardware will power Thinking Machines’ pursuit of “frontier models” that are designed to be more understandable and collaborative with humans. |
| Democratization | A shared goal to broaden access to frontier AI and open-source models for enterprises and the scientific community. |
| Software Synergy | Thinking Machines will utilize these architectures to enhance its “Tinker” API, which allows developers to fine-tune open-source models. |
The “Vera Rubin” Platform
Thinking Machines will be the first to utilize the full “rack-scale” Rubin stack:
- Rubin GPUs: Features 336 billion transistors and optimized inference capabilities.
- Vera CPUs: High-performance ARM-based processors designed to work in tandem with the GPUs.
- NVLink 6: The latest high-speed interconnect technology that treats an entire rack as a single, massive accelerator.
Why This Matters
For NVIDIA, this investment continues its successful strategy of backing the startups that become its biggest customers. For Mira Murati, the deal provides a guaranteed supply of the world’s most sought-after chips during a global shortage, ensuring Thinking Machines isn’t left in the cold as AI demand continues to skyrocket.
“NVIDIA’s technology is the foundation on which the entire field is built. This partnership accelerates our capacity to build AI that people can shape and make their own.”
— Mira Murati, CEO of Thinking Machines Lab
