Claude Secures Multi-Gigawatt TPU Deal with Google and Broadcom for 2027.

Anthropic has announced a major infrastructure agreement with Google and Broadcom to secure multiple gigawatts of next-generation Tensor Processing Unit capacity, scheduled to come online starting in 2027. The deal comes as demand for Claude accelerates rapidly, with the company's run-rate revenue climbing from $9 billion at the end of 2025 to over $30 billion.

The agreement represents a significant expansion of compute resources dedicated to training and serving Claude's frontier AI models. TPUs, Google's custom-designed chips optimized for machine learning workloads, have become critical infrastructure for companies building large language models. Securing gigawatt-scale capacity positions Anthropic to handle continued growth as businesses increasingly adopt AI assistants for content creation, coding, and analysis.

Anthropic's revenue trajectory reflects the broader surge in enterprise AI adoption. The company's revenue more than tripled in a matter of months, growing from $9 billion to surpass $30 billion on a run-rate basis. This acceleration mirrors patterns seen across the AI industry as companies move from experimentation to production deployment.

The timing of the TPU agreement suggests Anthropic is planning for sustained growth beyond current demand levels. By securing capacity for 2027, the company is betting that enterprise and consumer demand for advanced AI models will continue expanding, requiring substantially more computational power than what's available today.

Competitors including ChatGPT and Gemini have similarly invested heavily in compute infrastructure, though few deals of this magnitude have been publicly disclosed. The multi-gigawatt scale indicates preparation for models significantly larger and more capable than current versions.

The vast majority of the new TPU capacity will be located in the United States, adding to Anthropic's November commitment to invest $50 billion in American computing infrastructure. This geographic concentration aligns with growing policy emphasis on domestic AI capabilities and data sovereignty concerns among enterprise customers.

US-based infrastructure offers potential advantages for companies serving American businesses, including reduced latency, simpler compliance with data residency requirements, and alignment with government priorities around critical technology. The decision also positions Anthropic favorably as policymakers debate AI infrastructure investments and national competitiveness.

For builders and small businesses using Claude through APIs or interfaces, the expanded capacity should translate to more consistent availability and potentially faster response times as the infrastructure scales. Following their recent technical advancements, the additional compute resources will support continued model improvements without compromising service quality during peak demand.

The partnership with Broadcom, a major semiconductor and infrastructure company, suggests the deal involves not just access to existing TPU designs but potentially custom configurations optimized for Anthropic's specific workloads. Broadcom's expertise in networking and data center infrastructure will be critical for deploying gigawatt-scale computing efficiently.

As AI companies race to secure the computational resources needed for next-generation models, infrastructure agreements like this one are becoming as strategically important as the model architectures themselves. The 2027 timeline gives Anthropic a multi-year runway to continue developing Claude while competitors face potential capacity constraints.