The cost of renting Nvidia graphics cards for AI tasks has increased significantly since the start of 2026. According to data from Ornn Data cited by Andreessen Horowitz, rental prices for certain accelerators have surged by as much as 143 percent.
The Nvidia H200, introduced in 2024, saw the highest price increase at 143 percent since January 1, 2026. Rental rates for the B200 rose by 111 percent, while the RTX PRO 6000 and H100 saw increases of 73 percent and 54 percent, respectively. Even the older A100 model experienced a 9 percent price hike.
Market trends were volatile throughout the year. While rental costs for most accelerators dropped in the first quarter, with H100 and B200 prices falling by approximately 23 percent below January levels, a recovery began in April and May. The H200, in particular, showed sharp price fluctuations throughout the spring and into October.
The rise in AI service token consumption during the spring of 2026 is considered a potential factor for the increased rental costs. However, current statistics do not isolate the impact of specific sectors like corporate AI systems or autonomous agents, nor do they account for data center capacity or power supply constraints.
Older hardware remains in demand due to the ongoing AI boom. The Tesla V100, released nearly a decade ago, is still utilized for running mid-sized language models. Specifically, 32GB versions of the V100 are now being sold with a premium of approximately $200, reflecting the broader pressure on both new and secondary computing markets.
