NVIDIA Partners with Mira Murati's Thinking Machines Lab in Historic 1-Gigawatt AI Compute Deal

On March 10, 2026, NVIDIA announced a multi-year strategic partnership with Thinking Machines Lab, the AI startup founded by former OpenAI Chief Technology Officer Mira Murati. Under the agreement, NVIDIA will make a significant investment in Thinking Machines Lab and supply at least one gigawatt of next-generation Vera Rubin computing systems. This collaboration signals a major shift in the AI compute landscape.

A New Chapter for the Former OpenAI CTO

Mira Murati has become one of the most recognizable figures in the AI industry. During her tenure at OpenAI, she led the development of revolutionary products including ChatGPT and DALL-E, serving as Chief Technology Officer for several years. In 2025, Murati departed from OpenAI to establish her own AI venture, Thinking Machines Lab.

According to the company's website, Thinking Machines Lab's mission is to "make AI systems that are more widely understood, customizable and generally capable." Unlike the prevailing "black box" AI approaches, Thinking Machines Lab emphasizes interpretability and user control. Recent job postings reveal the company is developing AI models optimized for audio processing and visual reasoning, exploring custom implementations of Transformer architecture.

1 Gigawatt: An Unprecedented Commitment

The centerpiece of this partnership is a compute supply agreement of staggering scale. Thinking Machines Lab has committed to deploying at least one gigawatt of NVIDIA's Vera Rubin systems beginning in early 2027. What does one gigawatt of compute represent? This equals the combined computational power of hundreds of thousands of high-performance GPUs—a threshold previously approached only by the largest AI labs.

NVIDIA CEO Jensen Huang stated: "What Thinking Machines Lab is building aligns perfectly with our vision—making AI a tool that everyone can shape and make their own." This partnership accelerates Thinking Machines Lab's capacity to build frontier AI models while further cementing NVIDIA's dominance in AI infrastructure.

Vera Rubin: NVIDIA's Next-Gen AI Superengine

The Vera Rubin systems being supplied represent NVIDIA's next-generation AI computing platform. Announced in late 2025 and entering production in mid-2026, Rubin integrates HBM4 memory and 3nm process technology, specifically designed for inference workloads and Agentic AI. The platform includes the R100 GPU and a new "Vera" CPU, capable of handling complex AI tasks requiring reasoning and action.

Notably, NVIDIA is hosting the GTC 2026 developer conference from March 16-19. Market observers expect Huang to reveal production specifications for the Vera Rubin NVL72 rack during his keynote. The timing of this partnership announcement, just before GTC, is seen as a deliberate showcase of NVIDIA's ecosystem dominance.

AI Compute Race Enters New Phase

This deal marks a significant milestone in the AI compute race. With Anthropic, OpenAI, and Google all ramping up infrastructure investments, AI has become the hottest sector in technology. The one-gigawatt commitment represents not just scale, but signals that AI models are hurtling toward larger parameters and more modalities.

For Thinking Machines Lab, NVIDIA's full support means building compute infrastructure capable of competing with OpenAI and Google DeepMind within a short timeframe. For NVIDIA, investing in Murati's company further expands its ecosystem influence, transforming a potential competitor into a close partner.

Reference: NVIDIA Blog、Axios、TechCrunch、CNBC