Published August 21, 2026 · Updated August 21, 2026
GPU Rental Prices
GPU rental prices are what cloud providers charge to rent NVIDIA H100, B200, H200, A100, AMD MI300X, and other AI accelerators on an hourly basis. Prices vary by provider type, GPU model, region, and commitment structure. There is no central exchange. Each provider sets its own rates.
ComputeWatcher tracks GPU rental prices across hyperscalers and neoclouds. The live price tracker is under development. This page provides market context and provider coverage. Subscribe to the weekly digest for price movement summaries.
2026 GPU Rental Price Ranges (On-Demand)
| GPU Model | Generation | On-Demand Range | Price Trend |
|---|---|---|---|
| NVIDIA H100 SXM | Hopper | $2.00 - $3.50 / GPU / hr | Declining as B200 expands |
| NVIDIA B200 | Blackwell | $4.00 - $6.50 / GPU / hr | Settling as supply ramps |
| NVIDIA H200 | Hopper+ | $3.00 - $4.50 / GPU / hr | Premium for higher memory |
| NVIDIA A100 SXM | Ampere | $1.50 - $2.50 / GPU / hr | Declining, displaced by H100 |
| AMD MI300X | CDNA3 | $2.00 - $3.50 / GPU / hr | Competitive with H100 |
Price ranges based on publicly available on-demand pricing. Actual rates depend on provider, region, and availability. Committed and reserved rates are typically 20 to 40 percent lower. Unknown or confidential negotiated rates are excluded.
Provider Coverage
ComputeWatcher tracks the following providers. The live pricing table will show current on-demand rates, last-checked timestamps, and 7-day/30-day price movement.
| Provider | Type | GPU Models |
|---|---|---|
| CoreWeave | Neocloud | H100, B200, H200, A100 |
| Lambda | Neocloud | H100, A100, H200 |
| Crusoe | Neocloud | H100, B200 |
| Nebius | Neocloud | H100, H200 |
| RunPod | GPU Marketplace | H100, A100, RTX 4090 |
| Vast.ai | GPU Marketplace | H100, A100, RTX series |
| AWS | Hyperscaler | H100, A100, Trainium |
| Azure | Hyperscaler | H100, A100, NDv2 |
| Google Cloud | Hyperscaler | H100, A100, TPU v5 |
| Oracle OCI | Hyperscaler | H100, A10, MI300X |
| TensorDock | GPU Marketplace | H100, A100, RTX series |
| Latitude.sh | Bare Metal | H100, A100 |
| FluidStack | GPU Marketplace | H100, A100, RTX series |
| Voltage Park | Neocloud | H100 |
| Together AI | Inference Platform | H100 (inference) |
Why Prices Differ Across Providers
The same GPU model can differ significantly in price across providers for several structural reasons:
- Hyperscaler premium. AWS, Azure, and Google Cloud bundle compute with networking, storage, compliance tools, and enterprise support. This bundle commands a premium. AI companies that need tight integration with hyperscaler data services often accept this premium.
- Neocloud focus. CoreWeave, Lambda, and Crusoe specialize in GPU compute with less bundled overhead. Lower cost structure translates to lower per-GPU prices. The tradeoff is fewer integrated services.
- Marketplace models. RunPod and Vast.ai aggregate supply from independent GPU owners, creating a spot-market dynamic with the lowest prices and variable reliability.
- Geography. Compute in Texas, Virginia, and Oregon is priced differently than Europe or Singapore, reflecting local power costs and demand patterns.
- Hardware vintage. Providers who purchased H100s at 2024 prices can sustain lower margins than those who bought at peak 2023 prices. Cost basis differences drive pricing strategy.
Frequently Asked Questions
How much does it cost to rent an H100 GPU?
On-demand H100 SXM GPU rental costs range from approximately $2.00 to $3.50 per GPU per hour in 2026. Neoclouds like CoreWeave and Lambda price toward the lower end; hyperscalers like AWS and Azure price toward the upper end. Committed capacity is typically 20 to 40 percent lower.
How much does it cost to rent a B200 GPU?
B200 Blackwell GPU rental prices range from approximately $4.00 to $6.50 per GPU per hour on-demand. Supply is still ramping and prices are expected to decline as Blackwell capacity expands through 2026 and 2027.
Which GPU cloud provider is cheapest?
Marketplace providers including RunPod, Vast.ai, and TensorDock typically offer the lowest on-demand prices. CoreWeave and Lambda offer competitive rates with stronger reliability and enterprise contracts. Hyperscalers charge a premium for integrated services.
Get the weekly compute market digest
GPU prices, futures updates, ETF filings, financing deals - every Monday.
No sponsors. Unsubscribe anytime.