Rent NVIDIA H100
per GPU-hour
📊 Pricing at a Glance
NVIDIA H100 on-demand pricing, in stock right now, ranges from $2.59/GPU/hr to $5.95/GPU/hr across 18 offers from 8 providers with live stock tracking (updated October 2026). A further 252 tracked listings are shown below as listed / last-seen pricing.
Looking for a specific provider? See QuantaCloud NVIDIA H100, Massed Compute NVIDIA H100, or Ori NVIDIA H100.
Available Offers
The 5 cheapest on-demand offers reported in stock right now, from 8 providers with live stock tracking.
| Provider | GPU Model | VRAM | Host Specs | Region | Price | Status | Action | |
|---|---|---|---|---|---|---|---|---|
 QuantaCloud | NVIDIA H100 PCIe 80GB VRAM | 80GB | 20 vCPU 128GB RAM 1250GB Storage | Midwest | $2.59/GPU/hr | Available | ||
QuantaCloud | 4×NVIDIA H100 PCIe 80GB VRAM | 80GB | 64 vCPU 512GB RAM 5000GB Storage | Midwest | $2.59/GPU/hr $10.36/hr total (4×) | Available | ||
QuantaCloud | 2×NVIDIA H100 PCIe 80GB VRAM | 80GB | 40 vCPU 256GB RAM 2500GB Storage | Midwest | $2.59/GPU/hr $5.18/hr total (2×) | Available | ||
 Massed Compute | 4×NVIDIA H100 PCIe 80GB VRAM | 80GB | 64 vCPU 512GB RAM 5000GB Storage | Iowa | $2.73/GPU/hr $10.92/hr total (4×) | Available | ||
Massed Compute | NVIDIA H100 PCIe 80GB VRAM | 80GB | 20 vCPU 128GB RAM 1250GB Storage | Iowa | $2.73/GPU/hr | Available |
Other Tracked Listings (Last-Seen Pricing)
252 further listings GPUPerHour tracks for this GPU. These catalog rows do not qualify as current secure on-demand offers; stock observations may be missing or old. They are not confirmed in stock; check availability with the provider.
| Provider | Listed per GPU / hr | Pricing type | Availability |
|---|---|---|---|
| ThunderCompute | $1.38 | on-demand | Check provider |
| Vast.ai | $1.47 | on-demand | Check provider |
| Vast.ai | $1.47 | on-demand | Check provider |
| Vast.ai | $1.47 | on-demand | Check provider |
| Vast.ai | $1.47 | on-demand | Check provider |
| Vast.ai | $1.47 | on-demand | Check provider |
| Vast.ai | $1.60 | on-demand | Check provider |
| Vast.ai | $1.60 | on-demand | Check provider |
| Vast.ai | $1.60 | on-demand | Check provider |
| Crusoe | $1.60 | on-demand | Check provider |
H100 Availability Now
8 providers report H100 in stock on-demand right now (18 offers across 10 regions). Last stock observation: .
| Region | Cheapest $/GPU-hr | Provider | Offers in stock |
|---|---|---|---|
| us-midwest-2 | $2.59 | QuantaCloud | 3 |
| us-central-3 | $2.73 | Massed Compute | 3 |
| dallas-2 | $2.90 | Ori | 4 |
| Czechia, CZ | $3.42 | Vast.ai | 1 |
| global | $3.99 | RunPod | 1 |
The 5 cheapest of 10 regions with stock. Regions are as the providers name them. Prices on this page were observed on (UTC) and are quoted for that day; they are rechecked every minute and can change before booking.
H100 Price History (daily)
Over the last 14 days the cheapest on-demand H100 was $2.50 per GPU-hour across up to 8 providers with stock.
| Date (UTC) | Cheapest $/GPU-hr | Median $/GPU-hr | Offers in stock | Providers with stock |
|---|---|---|---|---|
| $2.50 | $2.90 | 22 | at least 5 | |
| $2.50 | $3.56 | 24 | at least 7 | |
| $2.50 | $3.73 | 27 | at least 7 | |
| $2.50 | $4.10 | 19 | at least 6 | |
| $2.50 | $2.90 | 26 | at least 7 | |
| $2.50 | $2.90 | 25 | at least 7 | |
| $2.50 | $3.13 | 25 | at least 6 | |
| $2.50 | $3.26 | 25 | at least 7 | |
| $2.50 | $3.11 | 27 | at least 8 | |
| $2.50 | $3.19 | 20 | at least 5 | |
| $2.50 | $3.10 | 24 | at least 6 | |
| $2.50 | $3.11 | 26 | at least 6 | |
| $2.50 | $3.02 | 20 | at least 7 | |
| $2.50 | $2.90 | 25 | at least 7 |
One row per captured day (00:05 UTC). Cheapest is the lowest secure on-demand price per GPU-hour reported in stock by a provider with live stock tracking; median is the median of those offers. Rows combine 3 H100 models (h100-nvl, h100-pcie-80gb, h100-sxm-80gb): cheapest is the lowest across them, median is the median of the per-model medians, and the provider count is the largest single-model count, so it is a floor. Days without a capture are not shown. Method and full series: Cloud GPU Price Index.
H100 Price by Provider
Cheapest on-demand offer per provider that is reported in stock right now. Prices are per GPU per hour; multi-GPU instances show the whole-instance price alongside. Deploy opens the DeployGPU console for providers on the platform, otherwise the provider's own site.
| Provider | Region | GPUs | Per GPU / hr | Instance / hr | Deploy |
|---|---|---|---|---|---|
| QuantaCloud | us-midwest-2 | 1 | $2.59 | $2.59 | Deploy this GPU |
| Massed Compute | us-central-3 | 4 | $2.73 | $10.92 | Deploy this GPU |
| Ori | dallas-2 | 2 | $2.90 | $5.80 | Deploy this GPU |
| Vast.ai | Czechia, CZ | 1 | $3.42 | $3.42 | Deploy this GPU |
| RunPod | global | 1 | $3.99 | $3.99 | Deploy this GPU |
| Lambda Labs | us-southeast-1 | 1 | $4.29 | $4.29 | Deploy this GPU |
| DigitalOcean | Amsterdam 3 | 1 | $4.41 | $4.41 | Deploy this GPU |
| Paperspace | ny2 | 1 | $5.95 | $5.95 | Deploy this GPU |
Notify me when H100 drops below a price
One email when the cheapest in-stock on-demand price per GPU-hour falls below your threshold. Cheapest right now: $2.59/GPU-hr.
H100 Catalog Listing History (30 days)
Catalog listings, not live prices.
These averages are of every tracked catalog listing, including spot, reserved, out-of-stock and older rows. They do not measure live on-demand prices or availability (the daily price history table above does that) and cannot establish savings against the current offers. Each day averages its recorded catalog snapshots; the median gives each day equal weight. Missing days are excluded. Figures combine 3 variants: NVIDIA H100 SXM5, NVIDIA H100 PCIe, NVIDIA H100 NVL.
| Lowest daily catalog average | $2.24/GPU/hr |
|---|---|
| Median of daily catalog averages | $2.28/GPU/hr |
| Days with recorded catalog prices | 30 of 30 |
| Latest catalog snapshot | October 30, 2026 |
QuantaCloud
H-series supply is constrained.
On-demand H-series stock comes and goes by the hour. If you need 16+ GPUs reserved or cluster capacity in the next 90 days, we can quote Hopper or B300 inventory across our partners within 24 hours.
Technical Specifications
Strengths & Limitations
- Exceptional performance for AI training and inference tasks.
- High memory bandwidth enables faster data processing.
- Optimized for large-scale deployments in data centers.
- Supports the latest NVIDIA software stack and libraries.
- PCIe form factor allows for easy integration into existing server infrastructure.
- High cost compared to consumer-grade GPUs.
- Requires significant power and cooling infrastructure.
- Not suitable for gaming or desktop applications.
- Performance benefits are most pronounced in highly parallel workloads.
Top Use Cases
Accelerate the training of large and complex deep learning models for natural language processing, computer vision, and other AI applications. The H100's Tensor Cores and high memory bandwidth significantly reduce training times.
Perform complex simulations and scientific computations in fields such as weather forecasting, drug discovery, and materials science. The H100's parallel processing capabilities enable faster and more accurate results.
Process and analyze large datasets for business intelligence, fraud detection, and other data-driven applications. The H100's high memory bandwidth and processing power enable faster data exploration and insights.
Real-World Benchmark
Market Analysis
Generated market commentary, not current pricing. Live prices and stock are in the sections above.
The NVIDIA H100 PCIe is positioned as a premium GPU for data centers and enterprises requiring the highest levels of performance for AI, HPC, and data analytics.
Frequently Asked Questions
What is the difference between the H100 PCIe and the H100 SXM5?â–¾
The H100 PCIe uses a PCIe interface for connectivity, while the H100 SXM5 uses a higher-bandwidth SXM5 interface. The SXM5 version typically offers higher performance due to its faster interconnect.
What software is compatible with the H100 PCIe?â–¾
The H100 PCIe is compatible with NVIDIA's CUDA toolkit, TensorRT, and other software libraries for AI and HPC. It also supports popular deep learning frameworks such as TensorFlow, PyTorch, and MXNet.
What are the power requirements for the H100 PCIe?â–¾
The H100 PCIe has a typical power consumption of around 350W. Ensure your server has adequate power supply and cooling to support the GPU.
Alternative GPUs
Journalists, bloggers, and researchers: You're welcome to cite our data in your articles with attribution. Our pricing database is checked periodically from 27 cloud providers tracking this GPU.