Anthropic Expands Broadcom Deal for 3.5GW TPU Compute

Anthropic announced on April 6, 2026, an expanded partnership providing access to approximately 3.5 gigawatts of next-generation Google Tensor Processing Unit capacity via Broadcom, beginning in 2027. This follows 1 gigawatt supplied in 2026 and aligns with prior multibillion-dollar deals, including a October 2025 agreement for 1 million TPUs worth tens of billions. Broadcom serves as the intermediary, supplying custom TPUs and networking components to Google through 2031, while Anthropic's usage depends on sustained commercial performance. The majority of this capacity will be located in the United States, extending Anthropic's November 2025 pledge to invest $50 billion in domestic AI infrastructure with sites in Texas and New York.

The expansion addresses surging demand from Anthropic's customer base, where revenue run rate exceeded $30 billion by April 2026, up from $9 billion at the end of 2025. Enterprise adoption doubled to over 1,000 business customers spending more than $1 million annually since the February 2026 Series G funding round, which raised $30 billion at a $380 billion valuation led by Coatue Management and GIC. This growth necessitates scaled inference and training compute, complementing foundational partnerships with Amazon Web Services using Trainium chips and Nvidia GPUs. Broadcom anticipates $21 billion in AI revenue from Anthropic in 2026, rising to $42 billion in 2027, stemming from prior $10 billion and $11 billion orders disclosed in late 2025.

Strategically, the deal positions Anthropic to maintain leadership in AI development, matching workloads to optimal hardware while prioritizing U.S.-based expansion. It underscores Broadcom's rising role in AI infrastructure, evolving from a chip supplier to a key enabler for models like Claude and Google's Gemini. The commitment reflects disciplined scaling to support exponential customer growth and frontier AI advancements, amid competitive dynamics where Anthropic captures over 73% of new corporate AI spending.