Citigroup points out that the Neocloud industry has moved beyond the "bare metal" stage. Platforms like Hosted.ai and Hydra Host use GPU virtualization to match fragmented computing power supply and demand, increasing utilization from the typical 20-40% of dedicated architectures to 60-90%, and compressing the billing revenue cycle from months to hours. This is a structural benefit to the return on existing and new GPU capacity, potentially unlocking the near-term ROI for Neocloud and inference service providers. Previously, the market primarily priced based on computing power scarcity and capital expenditure scale, while pricing based on actual computing power utilization efficiency was rather crude—with B300 now appearing in secondary channels and chip supply constraints beginning to ease, the focus of competition is shifting from "how many cards to acquire" to "how much revenue each card generates."
