Mira Murati's Thinking Machines Lab has signed a multibillion-dollar agreement with Google Cloud, expanding a relationship that began in 2025 into one of the largest AI infrastructure deals of the year. The deal, valued in the single-digit billions, gives the former OpenAI CTO's startup access to Google's latest A4X Max virtual machines powered by Nvidia's new GB300 chips, which Google claims deliver a 2x speedup over prior GPU generations.

The compute will primarily power the training and deployment of Thinking Machines' models, including its recently launched Tinker product — a tool that automates the creation of custom frontier AI models. Google's Jupiter network will handle the weight transfers required by the startup's reinforcement-learning workloads. "Google Cloud got us running at record speed with the reliability we demand," said Myle Ott, a founding researcher at Thinking Machines Lab.

The agreement is non-exclusive, leaving Murati's team free to use other cloud providers. Founded in February 2025 after Murati left OpenAI, Thinking Machines raised a $2 billion seed round at a $12 billion valuation and has emerged as one of the most closely watched AI startups. The deal intensifies competition between hyperscalers, as Google courts a frontier lab that could otherwise have landed with Microsoft-aligned OpenAI or Amazon-backed Anthropic.