What happened
Microsoft and Meta committed over $122.2 billion to neocloud providers CoreWeave and Nebius for AI compute capacity, with total potential commitments exceeding $145 billion including OpenAI and Anthropic deals. These commitments, representing approximately 90% of AWS's trailing twelve-month revenue, fund neoclouds' rapid infrastructure buildout, including 3.5 GWs of contracted power capacity each. Nvidia also invested $2 billion in both CoreWeave and Nebius, who consistently deploy the latest GPUs, such as CoreWeave being first with Nvidia GB200 and Vera Rubin systems.
Why it matters
Hyperscalers gain rapid access to frontier GPU generations and optimised compute utilisation, shifting significant capital expenditure off their balance sheets. This strategy allows platform engineers to accelerate AI development timelines by deploying high-density GPU infrastructure within months, a critical advantage over multi-year hyperscale data centre builds. Procurement teams face new financial models where long-term capacity deals with neoclouds, despite their smaller current revenues, secure essential compute resources, reflecting the broader trend of soaring AI compute costs.
