What a GPU-hour costs today, across 4 clouds
Every GPU on offer from RunPod, DigitalOcean, AWS, Prime Intellect, with live availability and the price per GPU-hour, normalized across machine sizes. Data as of November 21, 2026.
- Lambda not scanned yet
- RunPod scanned Nov 21
- DigitalOcean scanned Nov 21
- AWS scanned Nov 15
- Hyperstack not scanned yet
- Nebius not scanned yet
- Verda not scanned yet
- Massed Compute not scanned yet
- Prime Intellect scanned Nov 19
| GPU | VRAM | Providers | Available | On-demand from | Cheapest | Available offers, $/GPU-hour |
|---|---|---|---|---|---|---|
| H100 SXMNVIDIA Hopper | 80 GBHBM3 | 1 provider, 1 machine | 2 of 2 | $4.41 | $4.41On-demand | |
| H200 SXMNVIDIA Hopper | 141 GBHBM3e | 1 provider, 1 machine | 2 of 2 | $4.47 | $4.47On-demand | |
| B300NVIDIA Blackwell Ultra | 288 GBHBM3e | 1 provider, 1 machine | 1 of 1 | — | $6.17Spot |
Dots are individual offers, colored by provider. Filled dots are on-demand; hollow dots are spot or community capacity. Hover a dot for its price.