The cheapest H100 PCIe 80GB cloud rental right now is $1.66/GPU/hr from Shadeform. Full ranked list below — published on-demand list prices, updated hourly.
| Provider | Region | Billing | Total $/GPU/hr | Quality | Updated | |
|---|---|---|---|---|---|---|
S Shadeform marketplace | US, Dallas, TX | on_demand | $1.66/hr | Community | ● … | ▶ |
S Shadeform marketplace | US, Dallas, TX | on_demand | $1.79/hr | Community | ● … | ▶ |
S Shadeform marketplace | CA, Montreal | on_demand | $1.90/hr · 4 listings | Community | ● … | ▶ |
H Hyperstack neocloud | — | on_demand | $1.95/hr | Reliable | ● … | ▶ |
S Shadeform marketplace | CA, Montreal | on_demand | $1.95/hr | Community | ● … | ▶ |
R RunPod marketplace | — | community | $1.99/hr | Community | ● … | ▶ |
S Shadeform marketplace | US, Dallas, TX | on_demand | $1.99/hr | Community | ● … | ▶ |
V Vast.ai marketplace | Florida, US | on_demand | $2.00/hr | Community | ● … | ▶ |
S Shadeform marketplace | US, Dallas, TX | on_demand | $2.24/hr | Community | ● … | ▶ |
V Vast.ai marketplace | , US | on_demand | $2.40/hr | Community | ● … | ▶ |
S Shadeform marketplace | US, Des Moines, IA | on_demand | $2.48/hr · 8 listings | Community | ● … | ▶ |
R RunPod marketplace | — | community | $2.59/hr | Community | ● … | ▶ |
R RunPod marketplace | — | on_demand | $2.89/hr | Community | ● … | ▶ |
S Shadeform marketplace | US, Des Moines, IA | on_demand | $3.16/hr · 3 listings | Community | ● … | ▶ |
R RunPod marketplace | — | on_demand | $3.19/hr | Community | ● … | ▶ |
L Lambda neocloud | US | on_demand | $3.29/hr | Reliable | ● … | ▶ |
S Shadeform marketplace | FR, Paris | on_demand | $3.30/hr · 2 listings | Community | ● … | ▶ |
S Shadeform marketplace | IN, Mumbai | on_demand | $4.29/hr | Community | ● … | ▶ |
S Shadeform marketplace | US, New York, NY | on_demand | $5.99/hr | Community | ● … | ▶ |
The lowest price is often a spot/community-tier listing with interruption risk. The index median uses reliable datacenter providers. Verify availability and terms with the provider before provisioning. Methodology →