NVIDIA Makes Major Investment in Thinking Machines Lab, Mira Murati's AI Startup

Mira Murati is making a big move. The former OpenAI CTO has secured a multi-year strategic partnership with NVIDIA, including an investment described as "significant" by both companies and a chip supply contract valued at tens of billions of dollars according to the Financial Times.

Thinking Machines Lab: From 0 to $2 Billion in One Year

Founded in February 2025, Thinking Machines Lab aims to create AI systems that are "more broadly understood, customizable, and generally capable." In barely a year of existence, the startup has raised over $2 billion from top-tier investors: Andreessen Horowitz, Accel, NVIDIA, and even AMD's venture arm, NVIDIA's direct competitor in the chip market.

The trajectory is meteoric. The company went from about thirty employees at its founding to roughly 120 team members today, recruiting more talent from major AI labs than it loses.

One Gigawatt of Vera Rubin: A Frontier-Scale Signal

The heart of the deal: Thinking Machines Lab will deploy at least one gigawatt of NVIDIA's next-generation Vera Rubin systems to train its models and run its products. Deployment will begin in 2027.

One gigawatt of computing power — that's a threshold only the world's largest AI labs have approached. This figure sends a clear message: Mira Murati doesn't plan to settle for building tools on top of other people's models. She wants to compete at frontier model scale, against OpenAI, Google DeepMind, and Anthropic.

The deal also includes a technical collaboration: Thinking Machines Lab products will be specifically optimized for NVIDIA chips, strengthening the Santa Clara giant's hardware-software ecosystem.

Products Already in Production

Thinking Machines Lab isn't at the PowerPoint stage. The startup launched Tinker last year, a cloud service that lets developers create fine-tuned, customized versions of open-source large language models.

Recent job listings also reveal the company is working on:

• AI models optimized for audio processing
• Visual reasoning models
• Custom implementations of fundamental Transformer architecture components, notably the attention module

Internal Turbulence to Watch

The picture isn't without shadows. Co-founder Andrew Tulloch left the startup in October 2025 to join Meta. More concerning: in early 2026, three additional co-founders — Barret Zoph, Luke Metz, and Sam Schoenholz — departed to return to... OpenAI.

These departures raise legitimate questions about the founding team's stability, even though overall headcount growth suggests Murati is attracting enough talent to compensate.

Additionally, last year Murati turned down an acquisition offer from Mark Zuckerberg (Meta), confirming her determination to maintain control of her project.

What This Changes for the AI Ecosystem

This deal illustrates a fundamental trend: NVIDIA is no longer content to just sell GPUs. The giant takes strategic stakes in the most promising AI labs, locking in both customers for its next chip generations and influence over research direction.

For the French and European AI ecosystem, it's also a warning signal: the amounts at stake — tens of billions for a single supply contract — show the growing gap between resources deployed in the United States and those available in Europe. Even Mistral AI, the French champion, operates on a very different scale. Mistral is nonetheless targeting €1 billion in revenue by end of 2026.

The NVIDIA-Thinking Machines Lab partnership confirms that the frontier AI race is also — and perhaps above all — an infrastructure race. Whether Mira Murati can turn these billions into lasting technological advantage against rivals like OpenAI, who hold a considerable head start, remains to be seen.