BlockBeats News, August 13th. Analyst qinbafrank stated that five major trends in the AI Cloud Computing Power (CSP) sector were distilled from the latest financial reports of CoreWeave and Nebius. Demand continues to significantly outstrip recent deliverable supply—by the end of the second quarter, CoreWeave had operationalized power of approximately 1.5GW, contracted power of about 3.7GW, a backlog of orders of around $104 billion, not including the additional commitment of over $25 billion at the beginning of the third quarter; Nebius plans to deploy over 1GW annually starting in 2027, with a mid-term target of 5GW in contracted capacity. The industry bottleneck has shifted from "having customers" to "when electricity can be connected, delivering GPUs, and beginning billing."

Pricing power has indeed strengthened, but it is not a uniform increase: CoreWeave raised prices for various SKUs by around 25% in July; Nebius's older generation GPU prices rose by over 30% compared to the first quarter, with the average annualized revenue from new contracts in the second quarter exceeding $20 million per MW, and at the beginning of the third quarter, short-term emergency capacity even reached $40-50 million per MW. The most significant price increases are for short-term capacity, new-generation GPUs, large-scale clusters, and production-level inference, rather than traditional low-priority, long-term locked-in bare power.

Project-level ROI has been calculable, but company-level ROIC has not yet been proven. Nebius has disclosed a clearer project payback period for the first time. CoreWeave covers GPU investment with five-year contracts and asset-level financing but depreciation and interest still consume most of the operating profit. Both companies are upgrading to an "AI Infrastructure Operating System," shifting their focus from renting GPUs by the hour to providing an integrated platform for training, inference, storage, networking, model deployment, monitoring, governance, Agent operating environments, and cross-cloud operations.

There is a differentiation in the third-party capital introduction path: Nebius's light-asset model is closer to "capital partners investing to build an AI factory, Nebius outputting operating systems and operational capabilities"; CoreWeave Omni is closer to "deploying a complete cloud to customers' own data centers and GPUs." The direction is the same, but Nebius currently emphasizes reducing capital investment, while CoreWeave emphasizes hybrid cloud and sovereign AI delivery.