The Story

Thinking Machines Labs (TML), the AI startup founded by former OpenAI CTO Mira Murati, has announced a multiyear partnership with Nvidia to secure large-scale compute infrastructure for its future AI systems.

According to the official announcement from Thinking Machines Labs, the deal will provide the company with access to at least one gigawatt of Nvidia compute capacity, enabling the startup to train large-scale frontier AI models.

Key details of the partnership

• Gigawatt-scale compute: The agreement will deploy at least one gigawatt of Nvidia infrastructure dedicated to training next-generation AI systems.
• Next-generation hardware: The compute cluster will include Nvidia’s upcoming Vera Rubin architecture.
• Deployment timeline: The infrastructure is expected to come online in phases with initial deployment targeted around 2027.

Background on Thinking Machines Labs

Mira Murati founded Thinking Machines Labs after leaving OpenAI, where she served as Chief Technology Officer and briefly as interim CEO. The startup previously raised around $2 billion in funding at an estimated $10 billion valuation.

The company currently offers a product called Tinker, a fine-tuning API aimed at enterprise users. However, the scale of the Nvidia partnership suggests a broader ambition to develop the company’s own frontier AI models.

Why it matters

Access to large-scale compute has become one of the defining advantages in the race to build advanced AI systems. Infrastructure commitments measured in gigawatts signal a level of investment typically associated with the largest AI labs.

The Nvidia partnership positions Thinking Machines Labs among the emerging startups attempting to compete with established AI leaders by securing the computing power required to train next-generation models.