Anthropic expands Google and Broadcom chip deal as AI compute demand surges

Anthropic said on April 6, 2026, that it has signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, a move that underscores how aggressively leading artificial intelligence companies are still competing for computing power. The company said the capacity is expected to come online starting in 2027.

The announcement adds to a string of large infrastructure commitments across the AI sector, where access to chips, cloud capacity and power has become a central constraint on model development. Anthropic framed the deal as part of its effort to scale its customer base and continue building Claude.

Anthropic says capacity will begin coming online in 2027

In a statement published on April 6, Anthropic said the agreement covers multiple gigawatts of TPU capacity and expands its use of Google Cloud and Google-built chips supplied through Broadcom. The company did not disclose a total dollar value for the arrangement in its announcement.

Anthropic said the new capacity is intended to support its foundation models, agents and enterprise applications. The company also said the deal reflects its broader infrastructure strategy as demand for Claude grows.

Google and Broadcom deepen an existing AI supply chain

The agreement ties together three companies that already sit at different points in the AI stack: model development, cloud infrastructure and chip design and manufacturing. Anthropic said the arrangement builds on its existing relationship with Google Cloud, while Broadcom said it will help supply the TPU capacity that Anthropic plans to use.

Reuters reported that the deal gives Anthropic access to about 3.5 gigawatts of Google TPU capacity starting in 2027, a scale that illustrates how much compute leading AI developers are seeking to secure for future systems. Anthropic did not give that figure in its own announcement.

Compute remains a defining bottleneck in AI

The latest agreement comes as major AI companies continue to announce large spending commitments on chips and data centers. That spending has become a defining feature of the industry, with companies racing to lock in enough hardware to train and run increasingly demanding models.

Anthropic has been among the most active companies in that race. The company said the new partnership is meant to support its growth as it serves more customers and develops more advanced systems.

What to Watch

Investors and competitors will be watching whether Anthropic gives more detail on the timing, scale or economics of the expanded TPU supply. The broader question is whether the industry’s chip and power buildout can keep pace with demand, or whether compute shortages continue to shape which AI companies can scale fastest.