Calculator · published prices only
What does a GPU run cost?
Cost = the provider’s own published price per GPU-hour × 8 GPUs × 730 hours. Nothing is estimated: a provider with no published price is not in the table. Prices checked 16 Sept 2026.
- GPU
- B200
- GPUs
- 8
- Hours
- 730
B200 · 8 GPUs · 730 hours (1 month)
| Provider | Plan | Per GPU-hour | This run (on-demand) | This run (spot) | Source |
|---|---|---|---|---|---|
| UpCloud | B200 x1 | $5.15/GPU-hr | $30,096 | — | upcloud.com · 16 Sept 2026 |
| Hyperstack | B200 x1 | $6.00/GPU-hr | $35,040 | — | hyperstack.cloud · 16 Sept 2026 |
| Verda (formerly DataCrunch) | B200 x1 | $6.29/GPU-hr | $36,734 | — | verda.com · 16 Sept 2026 |
| E2E Networks | B200 x1 | $6.99/GPU-hr | $40,822 | — | e2enetworks.com · 16 Sept 2026 |
| Nebius | B200 x1 | $7.15/GPU-hr | $41,756 | $23,068 | nebius.com · 16 Sept 2026 |
| Oracle Cloud Infrastructure | B200 x1 | $14.00/GPU-hr | $81,760 | — | apexapps.oracle.com · 16 Sept 2026 |
On-demand and spot prices are list prices for the region the provider publishes. They exclude storage, egress, IP addresses and taxes, which are billed separately. Spot capacity can be reclaimed by the provider, so a spot figure is what the run would cost if it ran uninterrupted at that price. Committed and cluster deals are negotiated and are not shown here.
Running this for weeks?
Committed pricing is usually negotiated below the list prices above. Tell us the GPU count, duration and region.