Comparison · GPU vs GPU · read 00:32 UTCH100 NVL vs GH200
Specifications, today's prices from 20 providers, price per unit of memory and compute, and who rents each cheapest. Better values are highlighted; lower is better for prices.
Side by sideSpecs, prices, value
Highlighted cells are the better of the two: more memory and compute, lower prices, lower cost per unit.
| H100 NVL | GH200 | Difference | |
|---|---|---|---|
| Memory & compute | |||
| VRAM | 94 GB | 96 GB | -2% |
| Memory bandwidth | 3.94 TB/s | 4.00 TB/s | -2% |
| FP16 tensor | — | 989 TFLOPS | — |
| FP32 | — | 67 TFLOPS | — |
| Power | — | 700 W | — |
| Prices today | |||
| Cheapest on-demand | $1.88 | $2.29 | -18% |
| Median on-demand | $3.94 | $2.75 | +43% |
| Highest on-demand | $8.98 | $6.50 | +38% |
| Cheapest spot | $0.63 | — | — |
| Cheapest reserved | $1.99 | — | — |
| Current offers (n) | 36 | 5 | +620% |
| Providers with a fresh offer | 15 | 5 | +200% |
| Value | |||
| Per GB of VRAM · hour (cheapest) | $0.020 | $0.024 | -16% |
| Per GB of VRAM · hour (median) | $0.042 | $0.029 | +46% |
| Per PFLOP FP16 · hour (median) | — | $2.779 | — |
| Per TB/s · hour (median) | $1.001 | $0.687 | +46% |
| Trend | |||
| 30-day median change | +18.2% | — | — |
| 90-day median change | +16.9% | +0.2% | — |
Cheapest providers today · on-demand · current offersWho rents each one cheapest
- 1
Airon AI1× $1.88 - 2
GPU.ai8× $2.06 - 3
Vast.aiTrust 781× $2.34
- 1
LambdaTrust 801× $2.29 - 2
Sesterce CloudTrust 651× $2.52
- 3
SpheronTrust 531× $2.75
Verdict · computed from the numbers aboveChoose one, choose the other
Choose the H100 NVL if…
- you want choice: 15 providers have a fresh offer today against 5.
Choose the GH200 if…
- your model needs more than 94 GB per GPU: 96 GB against 94 GB.
- the hourly price matters most: median $2.75 against $3.94, 30% less.
- you serve LLMs: memory bandwidth per dollar is better ($0.687 per TB/s-hour against $1.001).
Questions about H100 NVL vs GH200
Which is cheaper to rent, the H100 NVL or the GH200?
Today the cheapest current on-demand H100 NVL is $1.88 per GPU-hour and the cheapest GH200 is $2.29; medians are $3.94 and $2.75. Cheapest excludes offers marked unavailable and counts one floor per marketplace.
How much memory do the H100 NVL and GH200 have?
94 GB of HBM3 on the H100 NVL, 96 GB of HBM3 on the GH200, at 3.94 and 4.00 TB/s.
Where do these numbers come from?
From 41 current offers read on the providers' own pages and APIs over the last 45 days (reference 00:32 UTC), or in our October 2026 market study for the few providers our collector does not read. Specifications come from manufacturer datasheets. Every price on the GPU pages links to its receipt and shows its read date.