Nvidia GPU rental costs and market value trends

Rental costs for Nvidia H200 GPUs have increased by approximately 160% since the beginning of 2026, driven by demand that continues to outpace supply. This surge outpaces other models in Nvidia’s lineup, including the B200, which saw a 120% increase, and the H100, which rose by about 35%. Even the older A100 chip has seen price increases compared to January.

On-demand median rental prices for the H200 currently range between $4.27 and $4.94 per GPU-hour across tracked providers. Hyperscalers such as AWS and Azure command higher rates, averaging between $7.91 and over $10.60 per hour.

In contrast to rising rental prices, the market value of high-end GPUs faces a significant depreciation curve. Data suggests that these chips lose roughly 50% of their market value within 18 to 24 months of release. As the initial massive wave of H100 deployments reaches this 18-month cliff, hundreds of thousands of units are hitting a depreciation wall each quarter. This trend may benefit inference clouds and AI consumers by lowering the cost of renting compute capacity.