Google Inks Multi-Billion Deal with Thinking Machines Lab
Thinking Machines Lab, the startup from ex-OpenAI leader Mira Murati, has closed a multi-billion-dollar pact with Google Cloud to boost its AI infrastructure. The agreement covers access to systems using Nvidia's newest GPUs. A person close to the discussions put the value in the single-digit billions.
The setup gives Thinking Machines entry to Google's cutting-edge AI tools based on Nvidia's GB300 chips. It also provides services for training models and putting them into production.
Expanding Google Cloud's AI Reach
Google keeps making cloud contracts with AI builders. The company wants to combine its cloud platform with extras like storage, a Kubernetes engine, and Spanner database. Just weeks ago, Anthropic cut a deal with Google and Broadcom for several gigawatts of tensor processing unit capacity. TPUs serve as Google's own AI chips for machine learning tasks.
Rivalry heats up fast. Anthropic locked in another contract with Amazon this week for as much as 5 gigawatts to train and run its Claude model.
Thinking Machines tied up with Nvidia earlier this year. That included cash from the chip giant. Now comes its initial link with a cloud operator. The arrangement stays non-exclusive, so the lab can tap other clouds later. Still, it shows Google chasing quick-scaling AI outfits from the start.
Thinking Machines' Quick Rise
Mira Murati stepped away from her role as OpenAI's top technologist and started Thinking Machines in February 2025. The firm pulled in a $2 billion seed investment soon after, hitting a $12 billion valuation. It stayed under wraps for months. Then in October, it rolled out Tinker, software that handles building tailored frontier AI models.
Murati's move came amid shifts at OpenAI, where she helped shape products like ChatGPT. Her new venture focuses on advanced AI development tools. The lab joins a wave of post-OpenAI startups pushing model creation boundaries.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Google Cloud's announcement shed light on the lab's work. It highlighted support for reinforcement learning jobs, key to Tinker's design. This method drives big advances at places like DeepMind and OpenAI. Such efforts demand huge computing power, matching the deal's size.
Tech Specs and Early Access
Thinking Machines counts as one of the initial Google Cloud users for GB300-based setups. Google says these deliver twice the training and serving performance of last-gen GPUs.
"Google Cloud got us running at record speed with the reliability we demand," said Myle Ott, a founding researcher at Thinking Machines, in a statement.
Nvidia's GB300 chips represent the latest in high-performance computing for AI. They build on prior lines to handle massive workloads. Google pairs them with its infrastructure for top efficiency.
The partnership fits Google's push in AI cloud services. It launched initiatives like Google Next to showcase these tools. Thinking Machines benefits from early positioning in this space.
OpenAI background adds context: Murati joined in 2018, rising to CTO by 2022. She oversaw technical strategy during rapid growth. Her exit in late 2024 fueled speculation on new AI labs. Thinking Machines emerged as a direct player in frontier models.
Competition overview: Amazon, Google, and Microsoft dominate cloud AI capacity. Deals like Anthropic's split across providers reduce risk. Startups seek scale without building data centers.
This deal underscores cloud giants' scramble for AI leaders. Thinking Machines gains firepower for Tinker expansions.