Anthorpic partners with Broadcom and Google to obtain 3.5 gigawatts of TPU computing power

Artificial intelligence company Anthropic announced that it has signed a new round of cooperation agreements with Google and Broadcom to roll out next-generation TPU (Tensor Processing Unit) compute power at the gigawatts level. Anthropic expects to gain access to approximately 3.5 gigawatts of TPU compute resources starting in 2027, becoming the core infrastructure for the future development of its Claude models.

Claude is also one of the few frontier models deployed across all three major cloud platforms: AWS (Bedrock), Google Cloud (Vertex AI), and Microsoft Azure (Foundry).

Anthorpic’s annualized revenue has already surpassed $30 billion

Judging from operational data, Anthropic’s growth rate has entered an acceleration phase. The company disclosed that in 2026 its annualized revenue surpassed $30 billion, a significant jump from roughly $9 billion at the end of 2025; in addition, the number of enterprise customers with annual spending exceeding $1 million has increased from 500 in February to more than 1,000 within just two months, showing that enterprise adoption of generative AI is spreading rapidly.

Most of this新增 compute capacity will be deployed within the United States, further expanding Anthropic’s $5 billion U.S. compute investment plan proposed in November 2025. Anthropic emphasized that it adopts a multi-chip strategy, running in parallel on Amazon Web Services Trainium, Google TPU, and NVIDIA GPU to improve performance and system resilience.

Currently, Claude is also one of the few frontier models deployed across all three major cloud platforms: AWS (Bedrock), Google Cloud (Vertex AI), and Microsoft Azure (Foundry).

WSJ: Anthropic can access 3.5 gigawatts of TPU compute power

The WSJ further revealed hardware and supply-chain details behind this cooperation. According to the report, Broadcom will be responsible for developing and supplying custom TPU chips for Google, and will provide network equipment and components related to data center racks. The cooperation term extends to 2031 and is part of a longer-term supply assurance agreement.

Under this arrangement, Anthropic expects to gain access to approximately 3.5 gigawatts of TPU compute resources starting in 2027. Broadcom also pointed out that this allocation of compute capacity still depends on Anthropic’s future business performance, indicating a tightly linked relationship between its compute supply and revenue growth.

The WSJ also noted that Google is actively pushing to expand the TPU ecosystem, hoping that through external supply and collaboration models, it can transform its in-house AI chips from internal tools into market-level products, further competing with GPU-dominated AI infrastructure. Against this backdrop, Anthropic has become a key customer—driving TPU shipments and strengthening Google’s voice in the AI compute market.

This article about Anthropic and Broadcom, Google collaborating to obtain 3.5 gigawatts of TPU compute power first appeared on Chain News ABMedia.