Thinking Machines Lab, a startup founded by former OpenAI executive Mira Murati, has secured a new multi-billion-dollar agreement to significantly expand its utilization of Google Cloud's advanced AI infrastructure. This exclusive deal, valued in the single-digit billions, according to TechCrunch, includes access to Google’s cutting-edge AI systems, which are powered by Nvidia’s latest GB300 chips, alongside comprehensive infrastructure services vital for model training and deployment. This marks Thinking Machines Lab's first significant partnership with a cloud services provider, following an earlier investment and partnership with Nvidia.

Google Cloud is actively pursuing a strategy of forming numerous cloud agreements with leading AI developers. Its objective is to integrate its powerful AI computing offerings with other essential cloud services, such as data storage, its Kubernetes engine, and the Spanner database product. This initiative aims to solidify Google's position in the rapidly evolving AI ecosystem by securing early partnerships with fast-growing frontier labs. The market for AI cloud services is intensely competitive, with other major players also vying for developer partnerships.

For instance, Anthropic recently entered into agreements with both Google and Broadcom to secure multiple gigawatts of tensor processing unit (TPU) capacity, Google’s custom-designed AI chips optimized for machine learning workloads. In a testament to the fierce competition, Anthropic also signed a separate agreement with Amazon this week to gain up to 5 gigawatts of capacity for training and deploying its Claude AI models. These multiple agreements highlight the critical need for vast computational resources in advanced AI development.

Mira Murati, who previously served as OpenAI's chief technologist, founded Thinking Machines Lab in February 2025. The company quickly gained traction, raising a $2 billion seed round at an impressive $12 billion valuation shortly after its inception. Despite its high valuation and prominent founder, the company has maintained a high level of secrecy until recently. In October, Thinking Machines launched its inaugural product, dubbed Tinker, a specialized tool designed to automate the creation of custom frontier AI models.

Wednesday’s deal with Google Cloud provides crucial insights into the innovative work being undertaken at Thinking Machines Lab. Google’s press release specifically mentioned its capability to support the startup’s reinforcement learning workloads, a training approach that forms the architectural backbone of Tinker. Reinforcement learning has been instrumental in recent groundbreaking achievements at leading AI research labs, including DeepMind and OpenAI, underscoring its computational intensity. The extensive scale of the Google Cloud deal reflects the immense computational demands associated with such advanced AI research.

Thinking Machines Lab is among the first cohort of Google Cloud customers to gain access to its GB300-powered systems. Google reports that these systems offer a remarkable 2X improvement in training and serving speed compared to prior-generation GPUs. Myle Ott, a founding researcher at Thinking Machines, affirmed the success of the collaboration, stating, “Google Cloud got us running at record speed with the reliability we demand.” While the deal is not exclusive, allowing Thinking Machines to potentially leverage multiple cloud providers in the future, it unmistakably signals Google’s strategic focus on locking in burgeoning frontier AI labs early in their development cycle.