Anthropic's Google Cloud deal includes 1 million TPUs and 1 GW of capacity in 2026

In a significant move for the artificial intelligence landscape, Anthropic and Google have officially unveiled a major cloud collaboration. This partnership grants Anthropic access to a staggering one million of Google’s custom-designed Tensor Processing Units (TPUs), marking the largest TPU agreement to date, with a deal worth tens of billions of dollars. Set to bring over a gigawatt of AI compute capacity online by 2026, industry experts estimate that establishing a 1-gigawatt data center could cost around $50 billion, with approximately $35 billion typically devoted to chip expenses. While other competitors, such as OpenAI, may boast even larger projections, Anthropic’s approach is more understated, focusing on strategic execution rather than flashy announcements. Founded by former researchers from OpenAI, Anthropic has embraced a methodical and efficient growth strategy, concentrating on the enterprise market. A cornerstone of its infrastructure is a multi-cloud architecture, enabling the Claude family of language models to operate across Google’s TPUs, Amazon’s custom Trainium chips, and Nvidia’s GPUs. This diversified setup allows Anthropic to optimize workloads for training, inference, and research. Anthropic’s CFO, Krishna Rao, emphasized the importance of this partnership, stating, "Anthropic and Google have a longstanding partnership, and this latest expansion will help us continue to grow the compute we need to define the frontier of AI." The flexibility of spreading workloads across various vendors enhances cost-effectiveness and performance. Google Cloud CEO Thomas Kurian noted the positive impact of TPUs on Anthropic’s operations, highlighting the strong price-performance ratio that has been evident over the years. As demand for compute resources rises, Anthropic's revenue is on a sharp upward trajectory, nearing an impressive $7 billion annually. The Claude model now supports more than 300,000 businesses, representing a remarkable 300-fold increase in two years. Despite Google’s growing role, Amazon remains a key partner, having invested $8 billion in Anthropic, compared to Google’s $3 billion stake. Amazon’s infrastructure, particularly its custom-built supercomputer known as Project Rainier, is critical for Claude’s operations, providing cost-efficient performance through the use of Trainium chips. Wall Street analysts are already noting the impact of Anthropic’s success on AWS growth, with expectations of continued contributions to revenue in the coming years. Meanwhile, Anthropic’s resilient multi-cloud strategy proved its worth during a recent AWS outage, ensuring Claude remained unaffected. Importantly, Anthropic retains control over its model weights, pricing, and customer data, maintaining a neutral position among cloud providers. This stance may be crucial as competition among major cloud services intensifies.