Nvidia GPU Rental Costs Jump 48 Percent in Two Months as AI Demand Outstrips Supply

The cost of renting an Nvidia Blackwell GPU has risen 48 percent in just two months, according to market data released in April 2026. The increase reflects demand for AI computing power that continues to outpace available supply.

Blackwell is Nvidia's latest generation of AI chips, designed for training and running large language models. Data centers and AI companies have been competing aggressively for access to these chips, driving up rental prices on cloud platforms that offer GPU access by the hour.

The supply crunch is partly structural. Most AI data centers are located in the United States, and nearly all leading AI chips are fabricated by TSMC in Taiwan. Advanced chips are then shipped back to Taiwan for final packaging and testing before being distributed globally. This supply chain has limited how quickly new chips can reach the market.

AI companies are confronting acute energy shortages as well. Data centers globally now draw 29.6 gigawatts of power, enough to run the entire state of New York at peak demand. Some AI companies have begun rationing computing power as a result.

Meta announced it is committing an additional $21 billion to CoreWeave for GPU capacity between 2027 and 2032, on top of an existing $14.2 billion commitment. The company projects its 2026 capital expenditures will reach $115 to $135 billion, nearly double 2025 levels.

Amazon's cloud unit reported that its AI revenue run rate exceeded $15 billion in the first quarter of 2026. The company's custom chip business, including Graviton and Trainium processors, has an annual revenue run rate above $20 billion.

Analysts say the GPU shortage is unlikely to ease significantly in the near term. New chip fabrication capacity takes years to build, and demand from AI companies, governments, and research institutions continues to grow.