Detailed Analysis
Anthropic has signed a major infrastructure agreement with Google and Broadcom securing multiple gigawatts of next-generation TPU capacity, with deployment expected to begin in 2027. The deal represents the company's largest compute commitment to date and is designed to support its rapidly scaling frontier Claude models amid surging enterprise demand. CFO Krishna Rao framed the partnership as part of a disciplined, long-term infrastructure strategy — one oriented around matching unprecedented customer growth with sufficient computational backbone to sustain continued model development at the frontier. The majority of the new capacity will be sited in the United States, extending Anthropic's November 2025 pledge to invest $50 billion in American computing infrastructure.
The financial metrics disclosed alongside the announcement underscore the urgency behind the compute expansion. Anthropic's annualized revenue run-rate has surpassed $30 billion as of April 2026, a striking acceleration from approximately $9 billion at the close of 2025. Perhaps more telling is the enterprise adoption signal: the number of business customers spending over $1 million annually has doubled — from over 500 to more than 1,000 — in less than two months since the company's Series G fundraising announcement in February. That pace of high-value customer acquisition reflects not merely broader AI market growth, but a specific deepening of Claude's penetration into mission-critical enterprise workflows that require substantial, sustained compute consumption.
The partnership deepens an already layered relationship with Google Cloud, building on increased TPU capacity announced in October 2025, while simultaneously expanding Anthropic's collaboration with Broadcom, a leading designer of custom AI accelerator chips. Critically, the new agreement does not replace but rather augments Anthropic's existing multi-platform compute strategy. Amazon Web Services remains Anthropic's primary cloud and training partner — with Project Rainier as a central collaboration — while Google Cloud and Microsoft Azure round out a distribution footprint that makes Claude the only frontier AI model available across all three of the world's largest cloud platforms: AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. This hardware and cloud diversification allows Anthropic to route workloads to the chips best optimized for specific tasks, improving both performance and system resilience.
The deal fits within a broader industry trend in which leading AI labs are making increasingly massive, long-horizon infrastructure commitments to stay competitive at the frontier of model capability. As training runs for next-generation models grow exponentially more compute-intensive, securing guaranteed access to gigawatt-scale capacity years in advance has become a strategic necessity rather than an operational detail. Anthropic's approach — spreading compute across proprietary TPUs, AWS Trainium, and NVIDIA GPUs from multiple vendors — also reflects a deliberate hedge against supply chain concentration risks that have constrained competitors. By locking in multi-gigawatt TPU commitments through a combined Google-Broadcom agreement while preserving its AWS partnership, Anthropic is positioning itself for a sustained multi-year scaling trajectory that few organizations outside the largest hyperscalers have historically attempted.
Read original article →