Anthropic To Use 1 Million Google TPUs In Deal Worth “Tens Of Billions Of Dollars”

Google appears to have unlocked the hardware angle to its AI strategy.

Anthropic has said that it will use up to one million Google TPUs (Tensor Processing Units) in a new deal. These TPUs are Google’s version of GPUs, and Google largely had used them internally so far. Anthropic said the deal is worth “tens of billions of dollars”, and will bring more than a gigawatt of capacity online for the company in 2026.

“Anthropic’s choice to significantly expand its usage of TPUs reflects the strong price-performance and efficiency its teams have seen with TPUs for several years,” said Thomas Kurian, CEO at Google Cloud. “We are continuing to innovate and drive further efficiencies and increased capacity of our TPUs, building on our already mature AI accelerator portfolio, including our seventh generation TPU, Ironwood,” he added.

“Anthropic and Google have a longstanding partnership and this latest expansion will help us continue to grow the compute we need to define the frontier of AI,” said Krishna Rao, CFO of Anthropic. “This expanded capacity ensures we can meet our exponentially growing demand while keeping our models at the cutting edge of the industry,” he added.

Anthropic will end up using three different kinds of AI chips for delivering its AI models to users. Among its arsenal are NVIDIA’s GPUs, Google’s TPUs, and Amazon’s Trainium chips. Interestingly, both Google and Amazon are significant investors in Anthropic. The use of all three kinds of chips gives Anthropic diversity and bargaining power among its chip suppliers.

But this might be the most significant TPU deal thus far for Google. Google does not sell its TPUs as standalone hardware externally to customers or other cloud providers, but instead, offers TPUs only as a service through its Google Cloud Platform (GCP). Apart from Anthropic, Apple and former OpenAI Chief Scientist Ilya Sutskever’s new startup, SSI also use TPUs. The biggest user for Google’s TPUs is Google itself, which relies on these chips not only for its AI specific operations, but to also run machine learning operations for functions like Google search and maps.

And TPU deals such as the one with Anthropic might be a big value unlock for Google. With its GPUs, NVIDIA has become the world’s most valuable company as tech companies have scrambled to acquire GPUs to power their AI models. Google has had TPUs for a while, but has largely chosen to use them for internal purposes. But if Google can strike such TPU deals, and perhaps one day also sell TPUs, it could create a massive new vertical for the company which appears to be finally hitting its stride in the AI space.

Posted in AI