Anthropic lands billion-dollar AI chip deals with Google, Broadcom

Anthropic has announced plans to invest hundreds of billions of dollars in chip deals with tech giant Google and Broadcom.

Under the deal, a US-based AI company will spend the money on Google’s chips and cloud services, aiming to secure critical computing resources to meet the company’s sugaring demand in the midst of the AI boom.

Taking to X, Anthropic announced partnership with Google and Broadcom “for multiple gigawatts of next-generation TPU capacity, coming online starting in 2027, to train and serve frontier Claude models.”

Besides using Google’s TPU, a competitor chip to Nvidia’s GPU, Anthropic will also get access to search giant’s cloud services.

Broadcom will deliver around 3.5GW of capacity on Google’s hardware. On the whole, the company under the deal will gain access to close to 5GW in new computing capacity over coming years.

In recent months, Anthropic executives are racing to secure resources related to computing power to meet surging demand for coding agents, helping to fund and train them.

“We are building the capacity necessary to serve the exponential growth we have seen in our customer base while also enabling Claude to define the frontier of AI development,” said Krishna Rao, Anthropic’s chief financial officer

The company also announced its intentions to keep supplying custom TPUs for Google as a part of a long-term agreement through 2031. The continuous supply of chips will help Google to boost sales of its in-house chips to compete with Nvidia.

Circular deals

Last year, OpenAI, the rival of Anthropic, also struck a deal with Broadcom, AMD, and Nvidia to expand AI tools and infrastructure. Google also invested billions into Anthropic, giving it a 14 per cent stake as of March last year.

In November, Anthropic committed to purchase $30 billion of additional capacity from Nvidia and Microsoft and spend $50 billion on new data centers in New York and Texas.