Anthropic Secures Next-Gen TPU Capacity From Google and Broadcom as Revenue Hits $30 Billion

Anthropic has expanded its partnerships with Google and Broadcom to secure multiple gigawatts of next-generation Tensor Processing Unit (TPU) compute capacity, starting in 2027, as part of a massive infrastructure push to fuel its surging AI operations.^[3][1][2] The AI company, known for its Claude models, announced this deal alongside news that its revenue run rate has topped $30 billion, more than tripling from $9 billion at the end of 2025, according to Bloomberg.[Source 2] This comes on the heels of a $21 billion commitment to Broadcom for nearly 1 million Google TPU v7p units, set for delivery by late 2026, as revealed during Broadcom's fiscal 2025 earnings call.^[1]

The agreement builds on Anthropic's October 2025 partnership with Google Cloud, which promised access to up to one million TPUs worth tens of billions of dollars and over a gigawatt of capacity by 2026.^[4] Now, Broadcom is stepping in as a key supplier, delivering fully assembled Ironwood Racks—complete rack-level AI systems—directly to Anthropic's data centers, bypassing its traditional role as a mere component provider to Google.^[1] Broadcom confirmed long-term pacts through 2031 for custom TPUs and networking components, enabling Anthropic to tap into about 3.5 gigawatts of TPU-based compute beginning in 2027, as part of broader multi-gigawatt commitments.^[2]

Anthropic's CFO Krishna Rao described the move as the company's "most significant compute commitment to date," aimed at powering frontier Claude models amid "exponential growth" in customer demand.^[3] The vast majority of this new capacity will be located in the United States, expanding on a November 2025 pledge to invest $50 billion in American computing infrastructure.^[3] This positions Anthropic to compete in the intensifying AI race, where massive compute resources are essential for training and deploying advanced models.

For Broadcom, the deal—split into a $10 billion order from Q3 fiscal 2025 and an additional $11 billion in Q4—highlights its pivot toward selling high-performance, cost-efficient custom ASICs as alternatives to Nvidia's GPUs, following a similar pact with OpenAI.^[1] It also ties into Broadcom's expanded collaboration with Google for AI racks through 2031, ensuring supply chain stability for next-gen infrastructure.^[2] Google Cloud CEO Thomas Kurian previously noted Anthropic's preference for TPUs due to their strong price-performance.^[4]

The partnerships matter deeply in the AI industry, where access to specialized chips like TPUs determines who can scale models fastest amid skyrocketing demand. Anthropic, affecting enterprises worldwide through its Claude Partner Network launch, gains a durable pipeline to serve growing users while bolstering U.S. tech leadership.^[3] Next steps include rolling out the initial 1 gigawatt by late 2026, with the full multi-gigawatt expansion ramping up from 2027, potentially transforming Anthropic's capacity to over 4.5 gigawatts combined.^[1][2][3]