Rental costs for Nvidia H200 chips have surged by approximately 160% between January 1, 2026, and mid-October 2026. This trend reflects a persistent gap between the high demand for AI computing power and available supply.
The H200 features 141 GB of HBM3e memory and 4.8 TB/s of memory bandwidth, offering 1.76 times the memory capacity of the H100. Despite its age, the six-year-old A100 model has also seen rental price increases since the beginning of the year.
Median on-demand rental prices for the H200 currently range from $4.27 to $4.94 per GPU-hour. Spot marketplaces offer lower rates between $1.47 and $3.88, while major cloud platforms like AWS and Azure charge between $7.91 and over $10.60 per hour.
Individual providers are adjusting their pricing structures. Nebius increased its on-demand H200 rate by 20%, moving from $4.50 to $5.40 per GPU-hour effective October 1, 2026. The company also raised H100 rental rates by 17% during the same period.
