Google expands partnership with Thinking Machines Lab in new multibillion-dollar deal

Google has expanded its partnership with Thinking Machines Lab, the AI startup founded by former OpenAI executive Mira Murati, through a new multibillion-dollar agreement focused on advanced cloud infrastructure and AI computing systems.

According to a source familiar with the arrangement, the deal is valued in the single-digit billions. It provides Thinking Machines Lab with access to Google’s latest AI infrastructure, including systems powered by Nvidia’s newest GB300 chips. The agreement also includes broader cloud services to support both training and deployment of large-scale AI models.

Google has been actively pursuing partnerships with leading AI developers to strengthen its position in the global AI infrastructure race. The company is increasingly bundling its cloud offerings with a wider stack of services, including storage systems, Kubernetes-based orchestration tools, and its Spanner distributed database platform.

Earlier this month, Anthropic signed a major infrastructure agreement with Google and Broadcom to secure multiple gigawatts of tensor processing unit (TPU) capacity. These TPUs are Google’s custom-designed chips built specifically for machine learning and large-scale AI workloads.

At the same time, competition in the AI infrastructure space continues to intensify. Just this week, Anthropic also announced a separate agreement with Amazon that provides access to up to 5 gigawatts of compute capacity for training and running its Claude AI models.

Thinking Machines Lab had earlier partnered with Nvidia in a deal that included investment from the chipmaker. However, this new agreement with Google marks the startup’s first major cloud infrastructure partnership. The deal is non-exclusive, allowing Thinking Machines Lab to work with multiple cloud providers in the future potentially. Still, it also signals Google’s strategy of securing early relationships with fast-growing frontier AI labs.

Mira Murati launched Thinking Machines Lab in February 2025 after leaving her role as chief technology officer at OpenAI. The company quickly raised a $2 billion seed round at a $12 billion valuation and has maintained a relatively secretive profile since its inception. Its first product, Tinker, launched in October and is designed to automate the creation of custom frontier AI models.

The latest Google Cloud agreement also sheds light on the startup’s technical direction. In a press statement, Google said its infrastructure will support reinforcement learning workloads, which form a core part of Tinker’s architecture. Reinforcement learning has played a key role in recent AI advancements from research labs such as DeepMind and OpenAI. Still, it remains extremely compute-intensive and requires large-scale infrastructure investments.

Thinking Machines Lab is among the first customers to gain access to Google’s new GB300-powered systems, which the company says deliver up to twice the training and inference performance compared to previous-generation GPUs.

“Google Cloud got us running at record speed with the reliability we demand,” said Myle Ott, a founding researcher at Thinking Machines Lab, in a statement.