H100-80GB vs H100-NVL-94GB
H100-80GB is the cheaper of the two: $1.7945/hr at Lium, against $3/hr for H100-NVL-94GB at Vast.ai — 1.7× the price.
Cheapest on each
Aligned facts
Same sheet, same order, both cards — including the gaps.
| Attribute | H100-80GB | H100-NVL-94GB |
|---|---|---|
| VRAM | 80 GB | 94 GB▲ |
| Bandwidth | ○ no answer | 3.9 TB/s |
| FP16 | ○ no answer | 835 TFLOPS |
| FP8 | ○ no answer | 1670 TFLOPS |
| TDP | ○ no answer | 400 W |
| NVLink | ○ no answer | NVLink 4 |
| Architecture | Hopper | Hopper |
| Released | 2022 | 2023▲ |
A blank is not a zero. Where one card states a figure and the other states none, neither wins the row — the second number is unknown, and ranking an unknown as nothing is how a comparison starts to lie.
Every seller, both cards
One row per seller, cheapest tier, on-demand list prices only. Sellers of both cards lead, cheapest combined rate first.
| Seller | H100-80GB | H100-NVL-94GB |
|---|---|---|
| Massed Compute | $2.732026-10-26 | $3.112026-10-26 |
| Lium | $1.792026-10-27JSON | ○ no answer |
| HyperAI | $1.802026-10-26 | ○ no answer |
| Sesterce | $1.832026-10-07 | ○ no answer |
| TensorDock | $2.252026-10-26 | ○ no answer |
| Hyperstack | $2.502026-10-26 | ○ no answer |
| Koyeb | $2.502026-10-26 | ○ no answer |
| Spheron | $2.642026-10-26 | ○ no answer |
| DeepInfra | $2.902026-10-26 | ○ no answer |
| Civo | $2.992026-10-27 | ○ no answer |
| Canopy Wave | $3.252026-10-27 | ○ no answer |
| Cerebrium | $3.402026-10-27 | ○ no answer |
| Beam | $3.502026-10-27 | ○ no answer |
| Oblivus | $3.642026-10-26 | ○ no answer |
| Runcrate | $3.862026-10-26 | ○ no answer |
| Enverge | $4.002026-10-26 | ○ no answer |
| DigitalOcean | $4.412026-10-11 | ○ no answer |
| Fal.ai | $4.502026-10-26 | ○ no answer |
| Lyceum | $4.692026-10-26 | ○ no answer |
| Replicate | $5.492026-10-26 | ○ no answer |
| Crusoe | $5.502026-10-27 | ○ no answer |
| Vast.ai | ○ no answer | $3.002026-10-27 |
| RunPod | ○ no answer | $3.192026-10-27 |
What this page will not tell you
On-demand list rates only · spot and committed terms priced separately and not mixed in here · no performance-per-dollar figure appears on this page, because it needs a throughput benchmark we do not hold and dividing a price by a vendor’s own TFLOPS claim produces a number that looks authoritative and means nothing