Vessl AI raised the hourly price of Nvidia H100 graphics processing units (GPUs) on its VESSL Cloud service by 24.7% this month, to $2.98 from $2.39. The increase follows similar moves by U.S. neocloud providers and highlights a broader shortage in AI computing capacity, where even previous-generation GPUs are becoming more expensive rather than declining in price. Ornn data shows H100 rental rates rose 21.9% in a month to $3.28, or about 70% of the rate for Nvidia's higher-end H200. Vessl AI operates about 7,000 GPUs, mostly through leases, and plans to reach 10,000 by year-end. Companies have struggled to secure capacity and price quotes for South Korean government AI transformation projects, while universities and research institutions are reducing research activity or seeking additional funding. Growing inference demand from AI agents and enterprise services, along with limited data-center power, cooling and available capacity, is making access to computing resources an increasingly important competitive factor.