Anthropic has signed a multi-year agreement to secure 3.5 gigawatts of Google Cloud TPU capacity, with Broadcom supplying the networking ASICs enabling high-speed data transfers. The partnership leverages Broadcom’s advanced switching technology to connect Anthropic’s AI models directly to Google’s TPU pods. This deal represents one of the largest TPU commitments to date by a single AI developer. It positions Anthropic to scale model training and inference workloads rapidly.

Broadcom reported annual revenue of $30 billion, driven largely by robust demand for high-performance networking components in AI infrastructure. The chipmaker’s switches play a critical role in optimizing data throughput between processors and memory in large AI clusters. Broadcom’s financial performance underscores the escalating need for specialized hardware as AI workloads grow. This strengthens its position as a key supplier in the hyperscale AI market.

For Google, the agreement locks in long-term TPU utilization, boosting the stability of Cloud AI revenue. Committing 3.5 GW of capacity enhances forecasting accuracy and capital allocation for TPU production. Increased utilization may accelerate additional TPU pod deployments and justify capacity expansions. Overall, this partnership signals growing enterprise adoption of Google’s AI compute services.