Google deepens Thinking Machines Lab ties with a new multi-billion-dollar agreement that expands the startup’s access to Google Cloud AI infrastructure and advanced Nvidia-powered systems.
According to TechCrunch, the new arrangement with Mira Murati’s startup is valued in the single-digit billions and includes infrastructure for model training, deployment and systems built on Nvidia’s latest GB300 chips.
The agreement marks a significant move in the growing competition among major cloud providers to secure fast-rising AI labs.
Google Deepens Thinking Machines Lab Ties Through Cloud Expansion
Thinking Machines Lab was founded by former OpenAI executive Mira Murati in February 2025.
The startup later raised a $2 billion seed round at a $12 billion valuation and has largely operated in stealth mode. It launched its first product, called Tinker, in October.
Google’s new deal is the first cloud services agreement signed by the company, according to the report.
While the partnership is not exclusive and Thinking Machines may work with multiple providers over time, the deal suggests Google is moving early to build long-term relationships with frontier AI developers.
What the Google Cloud AI Deal Includes
The agreement reportedly gives Thinking Machines access to Google’s latest AI systems powered by Nvidia GB300 chips, along with infrastructure services that support training and deployment workloads.
Google said its GB300-powered systems offer a two-times improvement in training and serving speed compared with previous-generation GPUs.
Myle Ott, a founding researcher at Thinking Machines, said in a statement that Google Cloud helped the company get running at record speed with the reliability it required.
Reinforcement Learning at the Center
The announcement also provided clues about the startup’s technical direction.
In its press release, Google said it can support Thinking Machines Lab’s reinforcement learning workloads, which are central to Tinker’s architecture.
Reinforcement learning has been behind recent advances at labs such as DeepMind and OpenAI, but it also requires significant computing resources.
That helps explain the scale of the new agreement.
Competition for AI Labs Is Intensifying
Google has been actively signing cloud deals with AI developers as it bundles infrastructure with services such as storage, Kubernetes Engine and Spanner.
Earlier this month, Anthropic signed an agreement with Google and Broadcom for multiple gigawatts of TPU capacity.
This week, Anthropic also signed a separate agreement with Amazon for up to 5 gigawatts of capacity to train and deploy Claude.
As Google deepens Thinking Machines Lab ties, the broader cloud race for elite AI customers is clearly accelerating in 2026.
Don’t miss out on our latest news—follow us for the latest AI news, breakthroughs