Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has entered a multi-billion-dollar AI infrastructure agreement with Google Cloud. The deal includes the use of NVIDIA's latest GB300 chips for computing systems, alongside model training and deployment services. This marks the company's first partnership with a cloud service provider, having previously collaborated with NVIDIA without utilizing third-party platforms. The agreement is non-exclusive, allowing future partnerships with other cloud providers. Google Cloud's GB300 system, adopted by Thinking Machines Lab, reportedly offers double the training and inference speed of previous-generation GPUs. The partnership will support reinforcement learning workloads, crucial for Thinking Machines Lab's first product, Tinker, which autonomously generates customized models. Founded in February 2025, the company secured a $2 billion seed round at a $12 billion valuation. Google has been actively expanding its AI collaborations, recently signing significant agreements with Anthropic and Amazon.