Exclusive: Google expands Thinking Machines Lab relationship with new multi-billion dollar partnership

Former OpenAI founder Mira Murati’s Thinking Machines Lab has signed a new multibillion-dollar deal to expand use of Google Cloud’s AI tools, including machines powered by Nvidia’s latest GPUs, TechCrunch has learned.

The deal is worth a single-digit billion, according to a source familiar with the matter, and includes access to Google’s latest AI machine built on top of Nvidia’s new GB300 chips, along with infrastructure services to support model training and deployment.

Google has been actively involved with AI developers as it seeks to wrap together its cloud offerings with other services such as storage, the Kubernetes engine, and Spanner, which were hosted. Earlier this month, Anthropic signed a contract and Google and Broadcom for several gigawatts of tensor processing unit (TPUs) power (these are AI chips developed by Google for machine learning).

But the competition is fierce. This week, Anthropic signed a new agreement with Amazon to secure up to 5 gigawatts of power to train and deploy Claude.

Earlier this year, Thinking Machines partnered with Nvidia in a deal that included funding from the chipmaker. But this is the first time the lab has partnered with a cloud provider. The deal isn’t exclusive, so Thinking Machines could use multiple cloud providers over time, but it’s a sign that Google wants to shut down its fast-growing labs.

Murati left his job as chief technology officer of OpenAI and founded Thinking Machines in February 2025. The company, which soon raised $2 billion in annual revenue. The cost of $ 12 billionhas remained largely secretive, but launched its first drug in October. Recorded Tinkeris a tool that enables the creation of custom AI models.

Wednesday’s deal provided insight into what Thinking Machines is up to. In a press release, Google noted that it could help expand the core services, which Tinker’s infrastructure relies on. Reinforcement learning is a training method that has underpinned recent breakthroughs in labs, including DeepMind and OpenAI, and the growing Google Cloud partnership shows the cost-effectiveness of the project.

Thinking Machines is among the first Google Cloud customers to get its GB300-powered machines, which offer a 2X improvement in training and serving speed compared to previous generation GPUs, on Google.

“Google Cloud gave us the speed and reliability we needed,” Myle Ott, founding researcher at Thinking Machines, said in a statement.