Nvidia GPU Rental Rates Rise, With H200 Up Nearly 49% in a Month

Rental prices climbed for all four tracked Nvidia GPU models in the mid-November snapshot, raising hourly costs for AI-compute buyers.

The H200 rate rose nearly 49% from the preceding month to about $7.75 per GPU-hour. B200 reached about $9.84 per GPU-hour after a 32% increase. H100 was priced at $3.25 per GPU-hour, up 8.3%, while A100 reached $1.06, a rise of 11.58%.

B200 rentals cost roughly nine times as much per hour as A100 rentals. H200 rentals cost more than twice as much as H100 rentals. Among the four models in the index, B200 remained the most expensive to rent, while A100 had the lowest rate.

Ornn’s index tracks rental transactions completed by neocloud providers, which are cloud companies that rent GPU capacity for AI workloads. It gives more weight to transactions with higher volume and settles each day at midnight ET. The resulting rates reflect completed rentals rather than prices posted in listings.

Ornn also operates a marketplace for GPU capacity and uses the index as a pricing reference there. Customers can rent spot capacity on a pay-as-you-go basis or reserve access for a set period. The index’s hourly rates measure the cost of rented computing time, not the price of buying a physical GPU.

The mid-November reading adds a new monthly snapshot to earlier coverage of rental prices across listed GPU tiers. Because the index is based on completed transactions, its readings can differ from advertised rental rates.