GPU cloud pricing
NVIDIA H100 price per hour: cheapest cloud
20 clouds rent out the H100. Voltage Park is cheapest at $1.99 per GPU-hour; Google Cloud charges 5.5× as much.
- On-demand from$1.99per GPU-hour · Voltage Park
- One GPU, all month$1,452.70730 hours on-demand
- Spot from$1.30per GPU-hour · Google Cloud
- Price gap5.5×across 20 clouds
- Memory94 GBlargest version listed
On-demand list prices per GPU-hour in USD · checked Nov 17, 2026 · How we compare GPU prices
Ranked for your usage
Monthly H100 cost by cloud
Cheapest for your volume
H100 at Voltage Park
28 options · most expensive $7,904.95, 5.5× this price
| Cloud | Estimated / month | Per GPU-hour | Spot / GPU-hour |
|---|---|---|---|
1 Voltage ParkGPU cloud · H100Pricing pageChecked | $1,432.80 | $1.99 | not listed |
2 HyperstackGPU cloud · H100Pricing pageChecked | $1,800.00 | $2.50 | not listed |
3 GMI CloudGPU cloud · H100Pricing pageChecked | $1,872.00 | $2.60 | not listed |
4 Massed ComputeGPU cloud · H100Pricing pageChecked | $1,965.60 | $2.73 | not listed |
5 RunPod (Secure Cloud)GPU cloud · H100 PCIEPricing pageChecked | $2,080.80 | $2.89 | not listed |
6 Massed ComputeGPU cloud · H100 NVLPricing pageChecked | $2,102.40 | $2.92 | not listed |
7 OVHcloudGPU cloud · H100Pricing pageChecked | $2,152.80 | $2.99 | not listed |
8 Massed ComputeGPU cloud · H100 SXMPricing pageChecked | $2,260.80 | $3.14 | not listed |
9 ScalewayGPU cloud · H100Pricing pageChecked | $2,281.03 | $3.17 | not listed |
10 RunPod (Secure Cloud)GPU cloud · H100 NVLPricing pageChecked | $2,296.80 | $3.19 | not listed |
1 GPU × 24 h × 30 days = 720 GPU-hours at list price. Storage, network egress, taxes and reserved-capacity discounts are not included.
On-demand and spot list prices per GPU-hour in USD, checked daily (last Nov 17, 2026). Sources: GPU Rental Prices and Cloud GPU List Prices (AWS, Google Cloud and Azure price APIs), both CC BY 4.0; every row links the cloud's own pricing page. Order depends on price only. How we compare GPU prices.
Serverless H100: pay only while it runs
Billed per second of use, usually with a cold start; cheaper than renting for spiky traffic, dearer around the clock.
| Platform | Per GPU-hour of use | Source |
|---|---|---|
| Koyeb H100 | $2.50 | Pricing page |
| Cerebrium H100 | $3.40 | Pricing page |
| Modal H100 SXM | $3.95 | Pricing page |
| fal H100 | $4.50 | Pricing page |
| Replicate H100 | $5.49 | Pricing page |
| RunPod (Secure Cloud) H100 | $5.89 | Pricing page |
Before you rent
Rates are per GPU-hour; an instance with several GPUs bills all of them. Hyperscalers often sell this chip only in 8-GPU instances, so check the smallest size before comparing.
Running one H100 around the clock costs $1,452.70 a month at Voltage Park and $8,014.74 at Google Cloud. Availability, regions and network speed differ, so check capacity before you commit.
Serving an open model? Compare per-token API prices first: below a steady load, paying per token is often cheaper than a rented GPU.