Anthropic has signed an agreement with Google and Broadcom for approximately 3.5 gigawatts of computing capacity built on Google's tensor processing units, with the capacity expected to come online starting in 2027, the company said Monday.

The deal expands an existing relationship. Broadcom CEO Hock Tan said during an earnings call last month that Broadcom was providing one gigawatt of Google TPU compute for Anthropic this year, according to CNBC. Broadcom helps Google design its TPU chips. In a securities filing Monday, Broadcom said Google and Broadcom have also entered a long-term supply assurance agreement running through 2031, according to Bloomberg.

Anthropic also said Monday that its revenue run rate has now crossed $30 billion on an annualized basis β€” more than three times the roughly $9 billion figure it recorded at the ende of 2025. Enterprise traction has also accelerated: The number of clients committing at least $1 million a year has surpassed 1,000, a threshold Anthropic said is twice what it was reporting around the time of its Series G announcement in February.

"We are making our most significant compute commitment to date to keep pace with our unprecedented growth," Anthropic CFO Krishna Rao said in a statement.

Anthropic said the majority of the new infrastructure will be built on U.S. soil, framing the commitment as a continuation of a pledge made last year to direct $50 billion toward domestic computing capacity. According to Monday's securities filing, Broadcom flagged that Anthropic's ability to draw on the additional compute hinges on its ongoing commercial performance, and noted that discussions with outside operational and financial partners are underway to support the rollout.

Anthropic trains and runs its Claude models across multiple hardware platforms β€” including AWS Trainium, Google TPUs, and Nvidia GPUs β€” and describes Amazon Web Services as its primary cloud and training partner. Claude is available on AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry.

The revenue growth comes as Anthropic navigates a legal dispute with the Pentagon, which designated the company a supply-chain risk after a standoff over AI safety guardrails. Anthropic has warned the label could cost it billions in lost revenue. Still, the company's annualized revenue has more than tripled in recent months, driven in part by demand for its Claude Code developer tools and broader enterprise adoption.

Broadcom stock gained as much as 3.6% in after-hours trading following the news. No financial terms were attached to the agreement. A post-earnings research note from Mizuho analysts put Broadcom's prospective AI revenue from Anthropic at $21 billion for 2026 and $42 billion for 2027, according to CNBC.