Google and Thinking Machines Lab Strike a Major AI Infrastructure Deal
In one of the most significant AI infrastructure deals of 2026, Thinking Machines Lab — the startup founded by former OpenAI CTO Mira Murati — has inked a multi-billion-dollar agreement with Google Cloud, TechCrunch has exclusively reported. At the heart of the deal is access to Nvidia's next-generation GB300 chips, the latest in the company's Blackwell Ultra line of AI accelerators.
The deal deepens what appears to be a growing partnership between Google and one of the most closely watched AI ventures of the year. Murati left OpenAI in late 2024 after nearly seven years at the company, including time as its chief technology officer, to launch Thinking Machines Lab. Since then, the startup has been among the most anticipated entrants in the increasingly competitive frontier AI space.
What the GB300 Chips Bring to the Table
Nvidia's GB300 chips represent a significant leap in raw AI compute. Designed for large-scale training and inference workloads, the GB300 series offers substantially higher memory bandwidth and compute throughput compared to its predecessors — exactly what frontier AI labs need to train and run the most capable models.
For Thinking Machines Lab, securing access to this infrastructure via Google Cloud gives the startup a credible path to building and deploying competitive large-scale AI systems without having to own the hardware outright. It's a strategy increasingly common among well-funded AI labs that want to move fast without the capital burden of data centre ownership.
Google's AI Infrastructure Play
For Google, the arrangement is equally strategic. Google Cloud has been in an aggressive race with Microsoft Azure and Amazon Web Services to become the go-to platform for AI companies. Landing a high-profile partnership with Thinking Machines Lab — and by extension Mira Murati, one of the most recognized names in AI — is a meaningful signal to the broader industry.
Google has also been deepening its own Nvidia chip access, alongside its in-house Tensor Processing Units (TPUs), to ensure it can meet the surging demand for AI compute from both internal teams and external customers.
Why This Deal Matters
The scale of the agreement — described as multi-billion-dollar — puts Thinking Machines Lab in rare company. Most AI startups operate with far more constrained compute budgets. A deal of this size suggests Thinking Machines Lab has secured substantial backing and has ambitions to compete directly with the likes of Anthropic, OpenAI, and xAI on frontier model development.
It also reinforces a broader trend: the AI compute arms race is intensifying, and the companies that can lock in infrastructure at scale will have a meaningful structural advantage as model capabilities — and the costs to develop them — continue to climb.
For the AI industry, all eyes will now be on what Thinking Machines Lab actually builds with this compute. Murati has so far kept the company's technical roadmap close to the chest, making this infrastructure deal one of the clearest signals yet that the lab is gearing up for something substantial.
Source: TechCrunch — Original reporting
