GPU compute markets
GPU rental
H100 NVL
94 GB of memory per GPU. Rental prices from 3 providers, compared.
About this GPU
Its specification, and where each provider publishes its own prices.
| Maker | NVIDIA |
|---|---|
| Architecture | Hopper |
| Introduced | 2022 |
| Memory per GPU | 94 GB HBM3, or HBM2e on the PCIe card |
| Power draw | Up to 700 W (SXM) or 350 W (PCIe) |
Providers listing it
Our own calculation, not a published figure. At each point in time it is the middle price among the cloud listings that have no line of their own (2 for this GPU), named in the table below the chart. Each listing counts once, whatever its capacity, so the line can move when a provider starts or stops listing this GPU. Spot prices are left out.
Clouds in the median
Best deal right now
$2.59/h
RunPod community · updated 3 min ago
$0.58/h (18%) less than the median, $0.73/h (22%) less than the dearest
Listings shown
4 of 4
provider and tier pairs
Per hour
US dollars per GPU over the last 7 days. Monthly prices assume 730 hours. Times are local.
| Listing | Latest | Lowest | Highest | Data points |
|---|---|---|---|---|
| Market median | $3.17 | $2.77 | $3.17 | 43 |
| RunPod secure | $3.19 | $3.19 | $3.19 | 43 |
| RunPod community | $2.59 | $2.59 | $2.59 | 43 |
| Vast.ai marketplace | $3.33 | $2.59 | $3.38 | 43 |
| Cloud median | $3.16 | $2.95 | $3.16 | 43 |
| Massed Compute on-demand | $3.16 | $3.16 | $3.16 | 43 |
| Spheron on-demand | $2.75 | $2.75 | $2.75 | 3 |
Listings right now
The latest price from every provider shown above, cheapest first.
| Provider | Tier | Per hour | Per month | Listed range | Supply | Updated | Via |
|---|---|---|---|---|---|---|---|
| RunPod | community | $2.59 | $1,891 | 3 min ago | RunPod | ||
| Massed Computesold out | on-demand | $3.16 | $2,307 | Out of stock | 3 min ago | Shadeform | |
| RunPod | secure | $3.19 | $2,329 | 3 min ago | RunPod | ||
| Vast.ai | marketplace | $3.31 | $2,420 | 6 of 42 free | 3 min ago | Vast.ai |
Machines
Whole machines from sources that describe them: Shadeform's clouds, Spheron, AWS and Azure. A machine sold in several regions of one area shows once, at its cheapest.
3 of 3 machines, cheapest per GPU first
| Provider | Tier | GPUs | vCPUs | RAM | Disk | Region | Per GPU | Machine |
|---|---|---|---|---|---|---|---|---|
| Massed Compute | on-demand | 2 | 40 | 256 GB | 2500 GB | US, Des Moines, IA +1 more · none free | $3.16/h | $6.32/h |
| Massed Compute | on-demand | 4 | 80 | 512 GB | 5000 GB | US, Des Moines, IA +1 more · none free | $3.16/h | $12.64/h |
| Massed Compute | on-demand | 8 | 140 | 1500 GB | 10000 GB | US, Des Moines, IA · none free | $3.16/h | $25.28/h |