Anthropic expands partnership with Google and Broadcom for multiple gigawatts of next-generation compute

Anthropic has announced a significant expansion of its compute infrastructure through a new agreement with Google and Broadcom, securing multiple gigawatts of next-generation Tensor Processing Unit (TPU) capacity. This substantial increase in compute power is expected to become operational starting in 2027, with the primary goal of supporting Anthropic’s frontier Claude models and meeting escalating customer demand worldwide.

Scaling to Meet Unprecedented Demand

The partnership underscores Anthropic’s aggressive growth trajectory and its commitment to scaling its AI capabilities. According to Krishna Rao, Anthropic’s CFO, this agreement represents the company’s “most significant compute commitment to date” and is a continuation of its strategic approach to infrastructure development. The need for this expansion is driven by a rapid acceleration in demand from Claude customers throughout 2026.

Anthropic’s run-rate revenue has now surpassed $30 billion, marking a substantial increase from approximately $9 billion at the close of 2025. This growth is mirrored in its enterprise customer base; the number of business customers each spending over $1 million on an annualized basis has doubled to over 1,000 in less than two months, following the company’s Series G fundraising announcement in February.

The majority of this newly acquired compute capacity will be located in the United States, further expanding on Anthropic’s November 2025 commitment to invest $50 billion in strengthening American computing infrastructure. This move deepens Anthropic’s existing collaborations with Google Cloud, building upon an increased TPU capacity announced last October, and reinforces its relationship with Broadcom.

Strategic Implications for the AI Industry

This massive compute acquisition highlights a crucial trend in the AI industry: the increasing necessity for vast computational resources to develop and deploy cutting-edge AI models. “Multiple gigawatts” signifies an enormous scale of investment in specialized hardware, indicating a continued arms race among leading AI developers to secure the processing power required for advanced AI research and commercial applications.

Anthropic currently utilizes a diversified hardware strategy, training and running its Claude models on a mix of AI hardware platforms, including Amazon Web Services (AWS) Trainium, Google TPUs, and NVIDIA GPUs. This multi-platform approach allows Anthropic to optimize workloads by matching them to the most suitable chips, which the company states translates to enhanced performance and greater resilience for its customers. Despite the expanded partnership with Google and Broadcom, Amazon remains Anthropic’s primary cloud provider and training partner, particularly for Project Rainier. Claude also maintains its unique position as the only frontier AI model accessible to customers across all three of the world’s largest cloud platforms: Amazon Web Services (via Bedrock), Google Cloud (via Vertex AI), and Microsoft Azure (via Foundry).

The announcement solidifies the strategic importance of close alliances between AI developers, cloud providers, and chip manufacturers. As AI models grow in complexity and capability, the availability and efficiency of underlying compute infrastructure become paramount competitive differentiators.

What to Watch

The activation of this next-generation TPU capacity in 2027 will be a key indicator of Anthropic’s progress in developing more powerful Claude models and its ability to capture a larger share of the rapidly expanding enterprise AI market. Observers will be keen to see how this compute power translates into tangible advancements in AI safety and research, as well as the commercial deployment of new AI applications.

---