The table below compares current hourly H100 pricing for a single NVIDIA H100 80GB GPU across major U.S. cloud providers.
To help you compare Nvidia H100 prices across cloud providers, this guide normalizes per-GPU rates (even when only multi-GPU instances are listed) and reflects each provider's current published U.S. on-demand rates. The information is reviewed weekly.
| Provider | GPU / Instance | On-Demand $/GPU-hr* | Notes |
|---|---|---|---|
| Vultr | H100 80GB | $2.99 | |
| Hyperbolic | H100 80GB | $3.19 | |
| Thunder Compute | H100 80GB | $3.20 | Managed availability with per-minute billing. |
| Lambda | H100 80GB | $3.29 | |
| Runpod | H100 80GB | $3.49 | |
| Nebius | H100 80GB | $3.85 | |
| Crusoe Cloud | H100 80GB | $3.90 | |
| Modal | H100 80GB | $3.95 | |
| DigitalOcean | H100 80GB | $4.41 | |
| Paperspace | H100 80GB | $5.95 | |
| Vast.ai | H100 80GB | $5.98 | Median of 3 distinct verified US/CA hosts; ranges $2.51-$6.80/GPU-hr. Priced as compute plus 100GB storage, with bandwidth billed separately. |
| CoreWeave | 8 x H100 80GB | $6.16 | Normalized from an 8-GPU node. |
| AWS | p5.48xlarge | $6.88 | Normalized from an 8-GPU node. Cheapest US region (us-east-1); ranges $6.88-$8.60/GPU-hr across 7 US options. |
| Azure | Standard_NC40ads_H100_v5 | $6.98 | Cheapest US region (eastus); ranges $6.98-$9.07/GPU-hr across 8 US options. |
| Oracle Cloud | BM.GPU.H100.8 | $10.00 | Normalized from an 8-GPU node. |
| Google Cloud | a3-highgpu-8g | $10.98 | Normalized from an 8-GPU node. Cheapest US region (us-central1); ranges $10.98-$13.20/GPU-hr across 10 US options. |
| Last reviewed on September 18, 2026. | |||
Methodology (why you can trust these numbers)
- Published list rates: Provider pages only, with normalized per-GPU pricing. Custom quote and commitment-only listings are labeled.
- Same class of silicon: All prices refer to NVIDIA H100 80GB GPUs.
- Normalized across SKUs: Where providers list only 8-GPU nodes, we divide by eight to get a single-GPU equivalent.
- USD in U.S. regions: Rates elsewhere can differ by 5-20 %.
Why this matters for developers
This table compares how much it costs to run an NVIDIA H100 for 10 hours across major cloud providers.
| Provider | On-Demand $/GPU-hr | 10-Hour Cost |
|---|---|---|
| Vultr | $2.99 | $29.90 |
| Hyperbolic | $3.19 | $31.90 |
| Thunder Compute | $3.20 | $32.00 |
| Lambda | $3.29 | $32.90 |
| Runpod | $3.49 | $34.90 |
| Nebius | $3.85 | $38.50 |
| Crusoe Cloud | $3.90 | $39.00 |
| Modal | $3.95 | $39.49 |
| DigitalOcean | $4.41 | $44.10 |
| Paperspace | $5.95 | $59.50 |
| Vast.ai | $5.98 | $59.79 |
| CoreWeave | $6.16 | $61.55 |
| AWS | $6.88 | $68.80 |
| Azure | $6.98 | $69.80 |
| Oracle Cloud | $10.00 | $100.00 |
| Google Cloud | $10.98 | $109.79 |
H100 availability is extremely low across cloud providers, so consider using an A100 instead when your workload does not strictly require H100 performance.
| Provider | GPU | Listed $/GPU-hr | 10-Hour Cost |
|---|---|---|---|
| Thunder Compute | A100 | $1.09 | $10.90 |
Price sources: provider pricing pages linked in the table above. Vast.ai values are marketplace estimates and vary by host and region.
H100 workloads are expensive by default, so availability, billing model, and operational friction matter alongside the hourly rate. Thunder Compute has managed H100 availability, per-minute billing, and fast launch workflows.
Takeaways
- Thunder Compute's H100 rate remains below Runpod, AWS, Azure, Oracle Cloud, and GCP, while Vultr and Hyperbolic list lower H100 headline rates.
- Thunder Compute's has managed availability, per-minute billing, and fewer setup steps, while marketplace H100 rates can come with more host-level variability.
- Bookmark this page, we review the numbers weekly so you do not have to.
- Building a startup? See our analysis of Startup-Friendly GPU Cloud Providers for credit offers, and spin up an H100 on Thunder Compute.
Read a full comparison of AWS P5 instances powered by NVIDIA H100 and H200 GPUs.
Compare other GPUs and maximize your computing power by reading the cheapest cloud GPU providers.