Anthropic inks big Google compute deal
Reporting says Anthropic signed a substantial compute agreement with Google, and that Alphabet also has a long‑term arrangement with Broadcom to develop custom TPUs and networking through 2031. Coverage frames these deals as part of a shift in where large AI models run their training and inference workloads. (timesofindia.indiatimes.com) (fool.com) (insidermonkey.com)
Anthropic said on April 6 that it signed its biggest compute commitment yet with Google and Broadcom for new artificial intelligence capacity starting in 2027. (anthropic.com) The company said the deal covers multiple gigawatts of Google Tensor Processing Unit capacity, with most of the new buildout planned in the United States. Anthropic also said its annualized revenue run rate has passed $30 billion, up from about $9 billion at the end of 2025. (anthropic.com) Google said the expansion will be delivered through Google Cloud services and Google-built Tensor Processing Units supplied through Broadcom. Anthropic said thousands of customers already use Claude through Google Cloud, including Coinbase, Cursor, Palo Alto Networks, Replit, and Shopify. (googlecloudpresscorner.com) Broadcom disclosed the other half of the arrangement in an April 6 filing with the Securities and Exchange Commission. The company said it has a long-term agreement to develop and supply future generations of Google Tensor Processing Units and a separate supply assurance agreement for networking and other components used in Google’s next-generation artificial intelligence racks through 2031. (sec.gov) Broadcom’s filing also said Anthropic will access about 3.5 gigawatts of Tensor Processing Unit-based compute beginning in 2027 through Broadcom, as part of Anthropic’s broader multi-gigawatt commitment. Broadcom said the actual level of consumption will depend on Anthropic’s commercial performance and talks with additional operating and financing partners. (sec.gov) For readers outside the chip industry, this is a deal about where large language models get their electricity, chips, and networking. Training and serving models like Claude requires giant clusters of processors tied together with fast links, and Google is offering Anthropic a stack that includes its own chips, cloud services, and rack hardware. (anthropic.com) (sec.gov) Anthropic had already been leaning into that approach. In October 2025, it said it would expand its use of Google Cloud technologies, including up to 1 million Tensor Processing Units, in a package worth tens of billions of dollars that was expected to bring well over 1 gigawatt of capacity online in 2026. (anthropic.com) At the same time, Anthropic has not abandoned other suppliers. The company said this month that it trains and runs Claude on Amazon Web Services Trainium chips, Google Tensor Processing Units, and Nvidia graphics processing units, and that Amazon remains its primary cloud provider and training partner through Project Rainier. (anthropic.com 1) (anthropic.com 2) That makes the April 2026 announcement less a clean break than a bigger wager on custom chips built inside cloud platforms. The immediate next step is physical: Anthropic, Google, and Broadcom now have to bring that new capacity online in 2027 and prove they can turn signed agreements into working clusters. (anthropic.com) (sec.gov)
