Anthropic has announced its largest infrastructure deal to date: a partnership with Google and Broadcom that will deliver 3.5 gigawatts of next-generation Tensor Processing Unit (TPU) compute capacity starting in 2027. The agreement, disclosed on April 6, deepens Anthropic's existing presence on Google Cloud and signals a dramatic scaling of the company's ambitions.
A $30 Billion Run Rate and Surging Demand
The deal arrives as Anthropic's business is accelerating at a pace that few predicted. The company's annualized revenue run rate has reached $30 billion, more than tripling from roughly $9 billion at the end of 2025 — a trajectory that builds on the $30 billion Series G round at a $380 billion valuation closed in February. More than 1,000 enterprise customers now spend over $1 million annually on Claude — a figure that doubled in less than two months from the 500-plus reported in February.
Anthropic CFO Krishna Rao called the partnership the company's "most significant compute commitment to date," describing it as "a continuation of our disciplined approach to scaling infrastructure."
What the Deal Includes
Under a parallel five-year agreement, Broadcom will continue to develop and supply future generations of Google's custom TPUs through 2031. Broadcom has been Google's TPU design partner since 2016 and currently produces the seventh-generation Ironwood chips. The deal also covers networking components and rack-server hardware.
For Anthropic specifically, the 3.5-gigawatt allocation represents an enormous expansion of available compute. The vast majority of the new infrastructure will be sited in the United States, consistent with Anthropic's November 2025 pledge of $50 billion in American AI infrastructure investment.
Infrastructure Diversity as Strategy
Notably, Anthropic is not going all-in on any single chip vendor. The company trains Claude across AWS Trainium, Google TPUs, and NVIDIA GPUs, making the model available on Amazon Web Services (Bedrock), Google Cloud (Vertex AI), and Microsoft Azure (Foundry). This multi-cloud, multi-chip strategy gives Anthropic negotiating leverage and resilience against supply-chain disruptions. The scaled compute will also feed research efforts like the Frontier Red Team's recent work using Claude Opus to surface 22 CVEs in Firefox.
Market Reaction
Investors responded positively to the announcement. Broadcom shares climbed over 6% to $333.60, while Alphabet gained roughly 2%. Broadcom CEO Hock Tan projected that AI chip revenue alone could be "significantly in excess of $100 billion" by 2027, up from $8.4 billion in the company's most recent quarter.
What It Means for the AI Industry
The deal underscores a structural shift in how frontier AI companies secure compute. Rather than relying solely on NVIDIA GPUs, leading labs are locking in multi-year commitments with custom silicon providers. For Anthropic, the partnership ensures it has the raw capacity to train and serve increasingly powerful versions of Claude as demand from enterprise customers continues to surge.
With run-rate revenue tripling in roughly four months and compute commitments measured in gigawatts, Anthropic is positioning itself not just as an AI research lab but as a full-scale infrastructure player — one that now rivals the spending profiles of the hyperscalers themselves.
