Mira Murati’s Thinking Machines Lab strikes multi‑year deal with NVIDIA, reportedly including 1GW Vera Rubin deployment and major investment

The deal — scale and money

Former OpenAI CTO Mira Murati’s new AI start‑up Thinking Machines Lab announced a multi‑year strategic partnership with NVIDIA (英伟达) on March 10, it has been reported. Thinking Machines Lab — co‑founded by Murati to push frontier AI research and open models — will reportedly deploy at least a 1GW‑scale NVIDIA Vera Rubin system beginning early next year, and NVIDIA has made a “significant investment” in the company, according to Chinese outlets IT Home (IT之家) and Phoenix (ifeng/凤凰网). The companies say the partnership will expand training and service systems built on NVIDIA architecture to broaden access to advanced and open AI models.

Why it matters

How big is 1GW? Very large — the phrase signals hyperscale training infrastructure rather than a modest cluster. If realized, a 1GW Vera Rubin deployment positions Thinking Machines as a major hub for large‑model training and experimentation, and it underscores NVIDIA’s push to lock in customers for its highest‑end AI systems. For Western readers: Vera Rubin is NVIDIA’s reference architecture and system family for large‑scale model training and inference, designed around its latest AI accelerators. The partnership shows a leading AI lab and the dominant GPU provider aligning on both hardware and software ecosystems.

Geopolitics and export controls

There are broader geopolitical implications. Semiconductor exports and AI hardware have been a focus of U.S. export controls and trade policy in recent years. It has been reported that NVIDIA’s commercial moves are increasingly shaped by these constraints; cross‑border hardware sales and joint ventures now draw extra scrutiny. Reportedly securing on‑premise, high‑power systems and direct investment into a new AI lab may be NVIDIA’s way to deepen partnerships while navigating an increasingly fragmented global AI supply chain.

What’s next

Thinking Machines says the collaboration will accelerate “building AI that people can shape and own,” according to CEO Mira Murati. Details on timelines, exact investment size, and where the systems will be located remain sparse — and some claims are unverified. Will this partnership materially shift the open‑model landscape? Watch for more granular disclosures about deployment sites, regulatory approvals, and the lab’s early model releases.