Nvidia invests heavily in Mira Murati's Thinking Machines Lab, forging AI partnership

Nvidia’s commitment goes beyond a mere supply contract; the chipmaker has taken an undisclosed equity stake in Thinking Machines Lab, cementing a financial bond that aligns both firms’ long‑term ambitions (The‑Decoder). The partnership will grant the startup at least one gigawatt of Nvidia’s next‑generation Vera Rubin compute, a capacity Nvidia describes as “gigawatt‑scale” and slated for deployment early next year (CNBC; The Verge). By locking in this volume of hardware, Nvidia ensures a steady pipeline of frontier‑AI workloads that can showcase the performance of its latest GPUs, while Thinking Machines gains the compute horsepower needed to train large‑scale models without the capital outlay traditionally required for such infrastructure.

Thinking Machines Lab entered the market with a $2 billion seed round led by Andreessen Horowitz, valuing the startup at $12 billion, and Nvidia was already listed among the investors in that round (The‑Decoder). The new equity infusion deepens that relationship, positioning Nvidia not just as a supplier but as a strategic partner with a vested interest in the startup’s success. According to Reuters, the multi‑year deal will also see the two companies co‑develop training and deployment systems optimized for Nvidia hardware, a move that could accelerate the rollout of “frontier AI” models to enterprise customers and research institutions (Reuters).

The timing of the alliance is notable given recent personnel shifts at Thinking Machines. Co‑founders Barrett Zoph and Luke Metz, both former OpenAI researchers, have departed to rejoin OpenAI, leaving Mira Murati as the sole high‑profile founder (The‑Decoder). Murati, who left OpenAI in 2024, has positioned the startup as a “next‑generation” AI lab focused on building models that rival the capabilities of OpenAI’s own offerings. By securing Nvidia’s gigawatt‑scale compute, she aims to offset the talent loss and keep the lab’s development velocity on track.

From Nvidia’s perspective, the partnership dovetails with its broader strategy to embed its hardware deeper into the AI stack. The Vera Rubin systems represent the company’s push toward higher efficiency and lower latency for massive model training, and a guaranteed gigawatt of usage from a high‑profile customer provides a marquee reference for future sales. Nvidia’s CEO has previously emphasized the importance of “strategic, long‑term collaborations” to drive adoption of its next‑gen GPUs, and the Thinking Machines deal is the most concrete example of that mantra to date (The‑Decoder).

Looking ahead, Thinking Machines Lab is reportedly preparing another funding round, suggesting that the Nvidia stake may serve as a catalyst for further capital inflows (The‑Decoder). If the startup can translate the Vera Rubin compute into commercially viable models, it could challenge incumbent AI providers and expand the ecosystem of businesses that can access cutting‑edge AI without building their own supercomputing clusters. The partnership thus marks a pivotal moment where a leading chipmaker and a fast‑growing AI lab align their resources and incentives to shape the next wave of artificial‑intelligence innovation.