OpenAI and Anthropic are actively seeking smaller data center facilities with capacities ranging from 20 to 30 MW. These sites are intended to provide immediate compute power while larger, gigawatt-scale infrastructure projects remain under construction.
Anthropic is reportedly exploring potential deals in the UK and Nordic countries. OpenAI is also investigating smaller deployments in the Nordics, and both companies have discussed similar opportunities within the United States to meet the rising demand for model inference.
The companies continue to pursue massive infrastructure projects alongside these smaller sites. Anthropic previously signed an agreement with Nscale for approximately 460 MW of capacity in West Virginia, valued at roughly $45 billion. OpenAI has committed to large-scale projects, including a 10 GW target for its Stargate infrastructure, plus additional 3 GW and 8 GW developments in Georgia and Ohio.
Smaller data centers offer a way to bypass challenges associated with large-scale construction, such as local opposition. In the second quarter of 2026, local resistance in the U.S. blocked 45 data center projects with a combined value of $68 billion due to concerns over land use, noise, and resource consumption.
Industry experts note that securing a few megawatts at existing sites is often more practical than waiting for large blocks of power. By distributing workloads across multiple smaller clusters, AI companies can effectively scale their operations while avoiding the multi-year development timelines required for mega-structures.
