NVIDIA H100 SXM rental price
Frontier AIThe benchmark rental of the AI market. Eight of them NVLinked in a box is the unit almost every training cluster is sold in, and its hourly rate is the closest thing this industry has to a spot price.
live · updated 17d ago
VRAM
80 GB HBM3
Architecture
Hopper
Segment
Frontier AI
Providers quoting
13
best rate
$1.336/GPU-hr at Vast.ai
24h
—
7d
—
30d
—
cheapest 8× node
$25.60/hr
The cheapest NVIDIA H100 SXM you can rent right now is $1.336 per GPU-hour at Vast.ai, against $11.06 at the dearest of the 13 providers quoting it — a 8.3× spread for identical silicon. A full 8-GPU node starts at $25.60/hr at Hyperstack. Rates below refresh automatically.
H100-SXM hourly rate history
Daily low $/GPU-hour per provider, cheapest provider first · showing 9 of 13 providers — hover the chart for detail.
Where to rent the NVIDIA H100 SXM
Vast.ai
Marketplace
$1.336/GPU-hr
1× H100-SXM · $1.336/hr per node · Moldova, MD
- 1× H100 SXM x1$1.602/hr
- 2× H100 SXM x2$3.469/hr
- 1× H100 SXM x1$1.736/hr
Horizon
GPU cloud
$1.950/GPU-hr
8× H100-SXM · $15.60/hr per node · sxm5 · IS, Blonduos 2
no capacity right now
Voltage Park
GPU cloud
$1.990/GPU-hr
8× H100-SXM · $15.92/hr per node · sxm5 · US, Dallas, TX
no capacity right now
- 4× h100-sxm5-80gb · no capacity$7.960/hr
- 2× h100-sxm5-80gb · no capacity$3.980/hr
- 1× h100-sxm5-80gb · no capacity$1.990/hr
Denvr Dataworks
GPU cloud
$2.350/GPU-hr
8× H100-SXM · $18.80/hr per node · sxm5 · US, Houston, TX
no capacity right now
Massed Compute
GPU cloud
$3.140/GPU-hr
8× H100-SXM · $25.12/hr per node · sxm5 · US, Des Moines, IA
no capacity right now
Hyperstack
GPU cloud
$3.200/GPU-hr
8× H100-SXM · $25.60/hr per node · sxm5 · CA, Montreal
Verda
GPU cloud
$3.310/GPU-hr
1× H100-SXM · $3.310/hr per node · sxm5 · FI, Central
no capacity right now
Nebius
GPU cloud
$3.866/GPU-hr
8× H100-SXM · $30.93/hr per node · sxm5 · FI, Helsinki
no capacity right now
- 1× gpu-h100-sxm|1gpu-16vcpu-200gb · no capacity$3.980/hr
DigitalOcean
GPU cloud
$3.890/GPU-hr
8× H100-SXM · $31.12/hr per node · sxm5 · CA, Toronto
no capacity right now
- 1× gpu-h100x1-80gb · no capacity$4.410/hr
Lambda
GPU cloud
$3.990/GPU-hr
8× H100-SXM · $31.92/hr per node · sxm5 · US, Dallas, TX
- 4× gpu_4x_h100_sxm5$16.36/hr
- 2× gpu_2x_h100_sxm5$8.380/hr
- 1× gpu_1x_h100_sxm5$4.360/hr
AWS
Hyperscaler
$6.880/GPU-hr
8× H100-SXM · $55.04/hr per node · us-east-1
- 1× p5.4xlarge$6.880/hr
Microsoft Azure
Hyperscaler
$11.06/GPU-hr
8× H100-SXM · $88.49/hr per node · eastus
- 8× ND96is_noIB_H100_v5$88.49/hr
- 8× ND96is_flex_H100_v5$88.49/hr
- 8× ND96isr_H100_v5$98.32/hr
H100-SXM cluster nodes
Multi-GPU configurations, cheapest per GPU-hour first. For distributed training the interconnect matters as much as the rate — an NVLink or InfiniBand node and the same count of loose PCIe cards are not the same machine.
| Provider | Configuration | Interconnect | $/GPU-hr | Node/hr |
|---|---|---|---|---|
| Vast.ai | 2× H100 SXM x2 · Washington, US | — | $1.734 | $3.469/hr |
| Horizonno capacity | 8× H100-8 · IS, Blonduos 2 | sxm5 | $1.950 | $15.60/hr |
| Voltage Parkno capacity | 8× h100-sxm5-80gb · US, Dallas, TX | sxm5 | $1.990 | $15.92/hr |
| Voltage Parkno capacity | 4× h100-sxm5-80gb · US, Dallas, TX | sxm5 | $1.990 | $7.960/hr |
| Voltage Parkno capacity | 2× h100-sxm5-80gb · US, Dallas, TX | sxm5 | $1.990 | $3.980/hr |
| Denvr Dataworksno capacity | 8× H100_80GB_SXM_8x · US, Houston, TX | sxm5 | $2.350 | $18.80/hr |
| Massed Computeno capacity | 8× gpu_8x_H100_SXM5 · US, Des Moines, IA | sxm5 | $3.140 | $25.12/hr |
| Hyperstack | 8× n3-H100-SXM5x8 · CA, Montreal | sxm5 | $3.200 | $25.60/hr |
| Nebiusno capacity | 8× gpu-h100-sxm|8gpu-128vcpu-1600gb · FI, Helsinki | sxm5 | $3.866 | $30.93/hr |
| DigitalOceanno capacity | 8× gpu-h100x8-640gb · CA, Toronto | sxm5 | $3.890 | $31.12/hr |
| Lambda | 8× gpu_8x_h100_sxm5 · US, Dallas, TX | sxm5 | $3.990 | $31.92/hr |
| Lambda | 4× gpu_4x_h100_sxm5 · US, Austin, TX | sxm5 | $4.090 | $16.36/hr |