Both AI labs are pursuing small-scale data center deals across the UK, Nordics, and US to bypass the lead times of larger campuses. Anthropic and OpenAI are seeking blocks of 20 to 30 MW to facilitate inference, where AI workloads spread across multiple sites instead of relying on a single tightly connected training facility. This strategy complements massive builds such as OpenAI's multi-GW Stargate projects and Anthropic's 460 MW agreement with Nscale for a project valued at roughly $45 billion in West Virginia.
Anthropic aims for 5 GW of available compute by the end of 2026 and roughly 10 GW by the end of 2027, matching OpenAI's infrastructure trajectory. The company is utilizing a $1.25 billion monthly payment to SpaceX through May 2029, a 5 GW partnership with Amazon, and a potential $10 billion compute lease from Meta over two years. JLL expects inference to overtake training as a share of global data center capacity in 2027.
Key sources
- SOURCE@cnbc“Anthropic and OpenAI hunt for smaller data center deals, sources tell CNBC, in race to deploy AI capacity”x.com
- SUPPORT@wallstengine“JLL expects inference to overtake training as a share of data center capacity in 2027 and reach 37% of global workloads by 2030”x.com
- SUPPORT@choblin29“Smaller deployments at sites that already have power can come online faster than another giant campus”x.com
- SUPPORT@wallstengine“agreed to pay SpaceX about $1.25B per month for compute through May 2029”x.com
- SUPPORT@stockmktnewz“Anthropic expects to end 2027 with ~10 GW of compute”x.com
- SUPPORT@financialjuice“Anthropic, OpenAI seek compute capacity deals of 20-30 MW”x.com
- SUPPORT@financialjuice“Anthropic, OpenAI mull small UK, Nordics data center deals”x.com
- SOURCEmarketbrief.now