Anthropic and OpenAI are pursuing smaller AI data center deals in the 20-30 megawatt range, adding to the multi-gigawatt commitments they already have in place, sources told CNBC. The move comes as the race to secure computing infrastructure for AI workloads intensifies.
Both companies inked enormous infrastructure deals over the past year — Anthropic committed roughly $45 billion to a cloud partnership with Nscale for around 460 MW in West Virginia, while OpenAI has surpassed its original 10 GW commitment to the Stargate project and added further capacity targets of 3 GW in Georgia and 8 GW in Ohio.
Anthropic has explored smaller-scale deals across the U.K. and the Nordics, and OpenAI has sought similar opportunities in the Nordics. Talks between both firms also covered potential U.S. deployments at that scale, sources said.
Smaller deals offer what Jabez Tan, head of research at Structure Research, called "speed to usable capacity" — securing a few megawatts at an existing powered site can be more practical than waiting for a much larger block in one location. For workloads that can operate across separate sites, a collection of smaller deployments can add up to substantial capacity.
The trend reflects a broader shift from model training to inference, the day-to-day process of serving AI models to users. Training requires massive chip clusters working closely together, while many inference workloads can serve separate requests across multiple smaller clusters, opening up more possible locations.
JLL expects the proportion of total data center capacity used for inference to overtake training by 2027. Inference accounted for 9% of global data center workloads in 2025 compared to 14% for training. By 2030, inference is projected to use 37% of that capacity versus 13% for training.
The shift is already reshaping the sector. Nvidia announced a collaboration in February to study smaller-scale data centers designed for distributed inference. Crusoe, the U.S. neocloud that built a large Texas data center complex used by OpenAI, is investing in smaller facilities that can be deployed faster and at lower cost than massive builds, the Wall Street Journal reported. Crusoe did not respond to a request for comment.
Crusoe also announced a $3.9 billion funding round on Thursday at a $30.9 billion post-money valuation, underscoring the surging interest in the neocloud space.
OpenAI said it is building a diversified compute portfolio to meet growing global demand. "Different workloads need different infrastructure, so we have conversations with a range of partners and assess opportunities based on our requirements, performance, reliability, timing and cost," a spokesperson told CNBC. "We don't comment on specific commercial discussions." Anthropic declined to comment.
Large data center projects in the U.S. and Europe face increasing pushback from communities, while Europe specifically contends with limited land and power availability. The sector is under pressure across much of the continent.
