The NVIDIA H100 (Hopper, 80 GB HBM3) rents from $2.00/hr, averaged across 58 cloud providers, updated 2026-09-28. On-demand pricing has risen 18% over the past 90 days.
The workhorse of the current AI build-out — strong FP16/FP8 training and inference, and the most widely available data-center GPU. Hopper · 80 GB HBM3 · 3.35 TB/s memory bandwidth · launched 2022.
▮ on-demand /hr · ▮ spot /hr
| Date | On-demand | Spot |
|---|---|---|
| 2026-09-28 | $3.39 | $2.00 |
| 2026-09-27 | $3.40 | $2.00 |
| 2026-09-25 | $3.42 | $2.00 |
| 2026-09-24 | $3.40 | $2.40 |
| 2026-09-23 | $3.39 | $2.21 |
| 2026-09-22 | $3.39 | $2.30 |
| 2026-09-21 | $3.35 | $2.15 |
| 2026-09-20 | $3.38 | $2.15 |
| 2026-09-19 | $3.40 | $2.15 |
| 2026-09-18 | $3.36 | $2.08 |
| 2026-09-17 | $3.38 | $2.00 |
| 2026-09-16 | $3.25 | $2.08 |
| 2026-09-15 | $3.27 | $2.00 |
| 2026-09-14 | $3.25 | $2.00 |
| GPU | On-demand /hr | Spot /hr | Memory | |
|---|---|---|---|---|
| H100 | $3.39 | $2.00 | 80 GB HBM3 | |
| B200 | $6.49 | $4.78 | 192 GB HBM3e | H100 vs B200 → |
| H200 | $4.48 | $2.62 | 141 GB HBM3e | H100 vs H200 → |
| MI355X | $5.99 | $2.97 | 288 GB HBM3e | H100 vs MI355X → |
| MI325X | $3.06 | $— | 256 GB HBM3e | H100 vs MI325X → |
| MI300X | $4.41 | $1.60 | 192 GB HBM3 | H100 vs MI300X → |
The H100 averages $3.39/hr on-demand, $2.00/hr spot, and $3.09/hr reserved across 58 cloud providers (updated 2026-09-28).
The lowest tier is spot at about $2.00/hr on average; reserved (1–3 year commitment) runs ~$3.09/hr. Compare live per-provider rates on the Bargo GPU price tracker.
At on-demand $3.39/hr, one H100 runs about $2,475/month (730 hours); reserved pricing lowers that to roughly $2,256/month.
On-demand, the H100 is $3.39/hr vs the B200 at $6.49/hr — cheaper per hour. See the full H100 vs B200 comparison for specs and cost-per-job context.
Spot ($2.00/hr) is interruptible spare capacity at a discount to on-demand ($3.39/hr); use spot for checkpointed or fault-tolerant jobs and on-demand or reserved for long, uninterrupted runs.
Cross-provider averages from public cloud GPU listings. Informational only.