Anthropic’s 3.5 GW Google TPU Commitment and CoreWeave Pact to Boost Cloud Revenue
Anthropic signed a multiyear agreement with CoreWeave to access production-scale computing for AI model training, with a phased rollout and options for expansion. Anthropic also plans to run workloads on 3.5 gigawatts of Google’s Broadcom-made Tensor Processing Units, potentially boosting Google Cloud’s chip utilization and revenue.
1. Anthropic and CoreWeave Multiyear Agreement
Anthropic signed a multiyear agreement with CoreWeave to provide production-scale computing capabilities for its AI model development, featuring a phased rollout and options for future capacity expansion. Financial terms and total gigawatt coverage remain undisclosed.
2. Commitment to Google’s Broadcom-Made TPUs
Under a separate commitment, Anthropic will run workloads on 3.5 gigawatts of Google’s Broadcom-made Tensor Processing Units, leveraging Google Cloud’s infrastructure to power its AI services and potentially boosting TPU utilization rates.
3. Rising Competition for AI Chip Capacity
The deal underscores the intensifying scramble among AI firms—including OpenAI’s custom chip projects and Meta’s own AI processors—to secure semiconductor capacity, a trend that could benefit cloud providers like Google through increased demand for rental or hosted chip services.