Updated
Updated · CNBC · Sep 18
Anthropic, OpenAI Pursue 20-30 MW AI Data Center Deals for Faster Capacity
Updated
Updated · CNBC · Sep 18

Anthropic, OpenAI Pursue 20-30 MW AI Data Center Deals for Faster Capacity

3 articles · Updated · CNBC · Sep 18

Summary

  • Anthropic and OpenAI are seeking much smaller 20-30 megawatt data center deals, with talks spanning the Nordics, the U.K. and some U.S. capacity, sources told CNBC.
  • Those smaller allocations can bring usable compute online faster than multi-hundred-megawatt builds, especially for inference workloads that can run across separate chip clusters and sites.
  • The shift complements the labs' huge recent commitments: Anthropic struck a roughly $45 billion, 460 MW cloud deal with Nscale, while OpenAI has expanded Stargate beyond 10 GW and added 3 GW in Georgia and 8 GW in Ohio.
  • Speed is becoming more valuable as large projects face land, power and community constraints, and JLL expects inference to overtake training in data center capacity use in 2027.

Insights

Why are tech giants quietly pivoting from massive mega-campuses to smaller, scattered AI data centers worldwide?
Could the explosive demand for AI inference actually transform neighborhood data centers into grid-stabilizing assets?