Anthropic, the San Francisco-based artificial intelligence company behind the Claude family of models, has disclosed that its run-rate revenue has surpassed $30 billion, more than tripling from approximately $9 billion at the end of 2025, as it announced a major compute infrastructure deal with Broadcom and Google.
The revenue figure emerged alongside a Broadcom regulatory filing that revealed Anthropic will access approximately 3.5 gigawatts of next-generation Google Tensor Processing Unit (TPU) computing capacity via Broadcom starting in 2027, the largest infrastructure commitment Anthropic has made to date.
TPUs are custom chips designed by Google specifically to accelerate AI workloads, and the new capacity represents a sharp escalation from the 1 gigawatt Broadcom is already supplying Anthropic this year.
The filing also contains a notable risk disclosure, stating that consumption of the expanded AI compute capacity by Anthropic is dependent on its continued commercial success, and that the parties are in discussions with certain operational and financial partners in connection with the deployment.
The disclosure appears to have prompted Anthropic to move quickly to reassure markets, with the company using its own announcement to highlight the $30 billion run-rate figure and note that more than 1,000 business customers now each spend over $1 million on an annualised basis, more than double the figure reported as recently as February.
Broadcom's filing sets out two separate agreements with Google: a long-term deal to develop and supply custom TPUs for future generations of the search giant's AI chips, and a supply assurance agreement covering networking and other components for Google's AI data racks through to 2031.
The Anthropic arrangement is linked to both, with Broadcom acting as the intermediary between Google's TPU production capacity and Anthropic's consumption of it.
The deal is the third major escalation of a commercial relationship that Broadcom chief executive Hock Tan first hinted at on an earnings call in September 2025, when he disclosed that an unnamed customer had placed a $10 billion order for custom TPU racks before confirming the customer was Anthropic in December 2025, when a further $11 billion order had followed.
Analysts at Mizuho have estimated the Anthropic relationship could generate $21 billion in AI revenue for Broadcom in 2026 and $42 billion in 2027.
Broadcom shares rose around 3% in extended trading following the announcement.
Anthropic said the majority of the new capacity will be based in the United States, consistent with a pledge made in November 2025 to invest $50 billion in American AI computing infrastructure.
Related reading
- Mistral AI raises $830 million for Paris data center
- OpenAI publishes open-source teen safety tools for developers building AI apps
- Anthropic secures multi-gigawatt Google and Broadcom chip deal as annual revenue run rate hits $30bn
The company also stressed that it maintains a diversified hardware strategy, running and training its Claude models across Amazon Web Services Trainium chips, Google TPUs, and Nvidia GPUs, and that Amazon remains its primary cloud and training partner through the Project Rainier supercluster arrangement.
Claude is currently the only frontier AI model available across all three of the world's largest cloud platforms: AWS, Google Cloud, and Microsoft Azure.
The recap
- Broadcom to build custom TPUs for Google under long-term agreement
- Anthropic plans to access approximately 3.5 gigawatts of TPUs
- Deployment with Anthropic access beginning in 2027, discussions ongoing
