Anthropic expands Google and Broadcom deal for “multiple gigawatts” of TPU capacity, eyes U.S. footprint

April 6, 2026
System with various wires managing access to centralized resource of server in data center
Photo by Brett Sayles on Pexels

The agreement

It has been reported that Anthropic has signed a new agreement with Google and Broadcom for multiple gigawatts of next‑generation TPU capacity, with hardware expected to come online starting in 2027. The company says this is its largest compute commitment to date — a move Krishna Rao, Anthropic’s CFO, called a necessary step to keep pace with “unprecedented growth.” Big words. Bigger hardware.

Why it matters

It has been reported that Anthropic’s run‑rate revenue has surged past $30 billion, up from roughly $9 billion at the end of 2025, and that the number of enterprise customers spending more than $1 million annually has more than doubled to over 1,000 in a matter of weeks. Staggering growth. What does that feel like inside a data hall? Like trying to build a second city overnight. The bulk of the new capacity will be sited in the United States, reinforcing Anthropic’s November 2025 pledge to invest $50 billion in U.S. computing infrastructure — a clear nod to the geopolitics of AI scale.

Technical strategy and market posture

Anthropic trains and serves Claude across a mix of hardware — AWS Trainium, Google TPUs and NVIDIA GPUs — and it has been reported that Amazon remains the company’s primary cloud partner as work on Project Rainier continues. The company argues that hardware diversity lets it match workloads to the best chip, improving performance and resilience for customers with mission‑critical needs. It has been reported that Claude remains the only frontier AI model available to customers across all three major clouds: AWS (Bedrock), Google Cloud (Vertex AI) and Microsoft Azure (Foundry). That multi‑cloud play is a selling point in a market where lock‑in can be painful.

The bigger picture

This deal is another data‑center chapter in the ongoing arms race for AI compute capacity. For customers it promises scale and choice; for competitors, it’s a signal that infrastructure — not just model code — is the new battleground. Expect more pacts like this one as firms scramble to secure the raw compute that powers next‑generation models.

Sources: anthropic