NVIDIA H100 SXM 80 GB rental prices

What these prices include

Prices are public list prices in US dollars, read once a day from each provider. The per-GPU figure is the instance price divided by the number of GPUs in the instance; it does not include storage, network transfer, taxes or negotiated discounts.

Azure prices are Linux pay-as-you-go figures from the retail prices API, per region. Spot prices are the current spot rate, which Azure can change at any time. Reserved prices divide the reservation total by the hours in the term (8,760 per year) and assume the reservation is fully used.

RunPod prices are per GPU on two tiers: Secure Cloud runs in RunPod-managed data centers, Community Cloud on vetted third-party hosts. RunPod does not price by region.

Lambda prices are per GPU for 1x, 2x, 4x and 8x instances; the committed rows are 1-Click Cluster prices for 16 GPUs or more on a two-week to one-year term.

Read the full method, including how price changes are checked before they are published.

Questions people ask

How much does it cost to rent an H100 SXM per hour?

On 2026-10-23 the lowest on-demand price was $2.69 per GPU-hour at RunPod (Community Cloud); on-demand listings ranged up to $23.33 per GPU-hour across 3 providers.

Which provider has the cheapest H100 SXM?

RunPod had the lowest on-demand price on 2026-10-23 at $2.69 per GPU-hour (H100 SXM (Community Cloud), Community Cloud). Prices are checked daily; the table above shows every provider and region.

What is the difference between on-demand, spot and reserved H100 SXM prices?

On-demand is the hourly list price with no commitment. Spot (or preemptible) capacity is sold at a discount but can be reclaimed by the provider at short notice. Reserved prices require a one- or three-year commitment and are shown here as the reservation total divided by the hours in the term.