GPU pricing · Hopper
NVIDIA H100 price per GPU-hour
Updated 2026-10-19
Lowest in stock: $1.77/GPU-hr (Lium), median $3.38 across 37 providers. On AWS: $6.88, and $25,000 in credits buys about 3,634 GPU-hours.
NVIDIA H100 has 80 GB of memory on the Hopper architecture.
AWS costs more per hour than the cheapest clouds, but it's the one your AWS credits pay for: the cheapest AWS option is p5.4xlarge.
- Lowest in stock
- $1.77/h
- Median
- $3.38/h · 37 providers
- On AWS
- $6.88/h
| Provider | Configuration | $/GPU-hr | Availability | Source |
|---|---|---|---|---|
| Gcore | 1x H100 SXM | $1.71 | Not reported | GetDeploying |
| Lium (marketplace) | 1x H100 SXM | $1.77 | In stock | GetDeploying |
| Hyper Ai | 1x H100 h100 | $1.80 | In stock | GetDeploying |
| Beam | 1x H100 PCIe | $1.83 | Not reported | GetDeploying |
| Gpu Ai | 1x H100 PCIe | $1.98 | In stock | GetDeploying |
| Together Ai | 1x H100 SXM | $1.99 | Not reported | GetDeploying |
| Tensordock (marketplace) | 1x H100 SXM5 | $2.36 | Not reported | GetDeploying |
| Denvr | 8x H100 SXM | $2.45 | Not reported | GetDeploying |
| Koyeb | 1x H100 gpu-nvidia-h100 | $2.50 | Not reported | GetDeploying |
| Hyperstack | 8x H100 PCIe n3-H100x8-NVLink | $2.60 | In stock | GetDeploying |
| Spheron (marketplace) | 1x H100 PCIe n3-H100x1-bigroot__CANADA-1 | $2.65 | In stock | GetDeploying |
| Massed Compute | 1x H100 gpu_1x_h100 | $2.73 | In stock | GetDeploying |
| Sesterce | 1x H100 | $2.75 | In stock | GetDeploying |
| OVHcloud | 1x H100 h100-380 | $2.99 | Not reported | GetDeploying |
| Civo | 1x H100 SXM | $2.99 | Not reported | GetDeploying |
| Canopywave | 2x H100 SXM5 | $3.25 | Not reported | GetDeploying |
| Lambda | 1x H100 PCIe gpu_1x_h100_pcie | $3.29 | In stock | GetDeploying |
| Cyfuture | 8x H100 SXM 8H100.128v.2048m SXM | $3.35 | Not reported | GetDeploying |
| Runcrate | 1x H100 SXM5 | $3.38 | In stock | GetDeploying |
| Cerebrium | 1x H100 HOPPER_H100 | $3.40 | Not reported | GetDeploying |
| Jarvislabs | 1x H100 SXM | $3.49 | In stock | GetDeploying |
| Oblivus | 8x H100 PCIe H100_NVLINK_80GB_x8 | $3.64 | In stock | GetDeploying |
| Empiriolabs | 1x H100 h100-80gb-1x | $3.80 | In stock | GetDeploying |
| Hikube | 1x H100 gpu.h100 | $3.81 | Not reported | GetDeploying |
| Daytona | 1x H100 GPU-only | $3.95 | In stock | GetDeploying |
| Enverge | 1x H100 PCIe | $4.00 | Not reported | GetDeploying |
| DigitalOcean | 1x H100 SXM gpu-h100x1-80gb | $4.41 | Not reported | GetDeploying |
| Fal Ai | 1x H100 | $4.50 | Not reported | GetDeploying |
| Nebius | 1x H100 SXM NVIDIA HGX H100 | $4.50 | Not reported | GetDeploying |
| Replicate | 1x H100 gpu-h100 | $5.49 | Not reported | GetDeploying |
| Paperspace | 1x H100 SXM H100 | $5.95 | Not reported | GetDeploying |
| CoreWeave | 8x H100 SXM | $6.16 | Not reported | GetDeploying |
| AWS | p5.4xlarge | $6.88 | In stock | AWS Price List |
| Azure | Standard_NC40ads_H100_v5 (1x GPU, eastus) | $6.98 | Not reported | Azure Retail Prices API |
| Cloudferro | 1x H100 h100.1 | $7.04 | Not reported | GetDeploying |
| Oracle Cloud | 8x H100 SXM BM.GPU.H100.8 | $10.00 | Not reported | GetDeploying |
| Google Cloud | 8x H100 SXM a3-highgpu-8g | $11.07 | Not reported | GetDeploying |
AWS instances with NVIDIA H100
| Instance | GPUs | On demand / h | Per GPU-hour | Spot avg / h |
|---|---|---|---|---|
| p5.4xlarge | 1 | $6.88 | $6.88 | $4.39 |
| p5.48xlarge | 8 | $55.04 | $6.88 | $33.16 |
What your AWS credits buy
| Credits | GPU-hours | Running 24/7 for |
|---|---|---|
| $1,000 | 145 | 6 days |
| $5,000 | 727 | 30 days |
| $25,000 | 3,634 | 5.0 months |
| $100,000 | 14,535 | 19.9 months |
Compare
Third-party prices are on-demand listings from GetDeploying (and the Azure Retail Prices API for Azure), checked 2026-10-14. We drop out-of-stock, community and spot listings, and any price under 35% of the median.
Related cost calculators
- EC2 Pricing Calculator · Compare Instance Types & Costs
- EKS Pricing Calculator · Compare Auto Mode vs EC2 Node Costs
- Amazon SageMaker AI Pricing Calculator · On-Demand Studio, Training & Inference
Questions
How much does an H100 cost per hour?
From $1.77 per GPU-hour in stock on demand, with a median of $3.38 across 37 providers we track. On AWS it's $6.88 (p5.4xlarge, us-east-1).
Can AWS credits pay for H100 instances?
Yes, Activate credits cover on-demand AWS GPU instances. Capacity for top GPUs can be limited; Capacity Blocks and Savings Plans are not payable with credits.
Prices and credit amounts are estimates and may change. Check AWS for current terms.