NVIDIA H200 GPU cloud prices
Hopper refresh with 76% more memory — for memory-bound training and inference.
Compare vs other GPUs →Not enough price history yet — the trend fills in as we log more.
Provider-balanced managed-cloud median per GPU-hour (USD) · live data
Key Specifications
ArchitectureHopper
Memory141GB HBM3e
Memory Bandwidth4.8 TB/s
ReleaseQ2 2024
Compare Cloud Provider Prices
We’re tracking 50 providers for the NVIDIA H200 — on-demand runs $0.50 to $21.20/hr per GPU, with interruptible spot from $0.50/hr.
| Provider | Instance | Location | VRAM | Billing | $/GPU/hr | Availability | |
|---|---|---|---|---|---|---|---|
 RunpodMarketplace | 1× NVIDIA H200 NVL | Location not disclosed | 143GB | STANDARD | $0.5 | Not reported | View ↗ |
| RunpodMarketplace | 1× NVIDIA H200 NVL | Location not disclosed | 143GB | BURST · SPOT | $0.5 | Not reported | View ↗ |
CCCompute CheapPublic pricing feed | 1× NVIDIA H200 | US | 141GB | BURST · SPOT | $1.39 | Not reported | View ↗ |
CCCompute CheapPublic pricing feed | 1× NVIDIA H200 | Location not disclosed | 141GB | BURST · SPOT | $1.39 | Not reported | View ↗ |
GOGPU OutletPublic pricing feed | 1× NVIDIA H200 | RU | 141GB | STANDARD | $1.4 | Not reported | View ↗ |
 Vast.aiMarketplace | 1× NVIDIA H200 | New Jersey, US | 141GB | BURST · SPOT | $1.712 | Available | View ↗ |
| Vast.aiMarketplace | 1× NVIDIA H200 | New Jersey, US | 141GB | STANDARD | $1.975 | Available | View ↗ |
 VerdaPublic pricing feed | 4× NVIDIA H200 | FIN-03 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 2× NVIDIA H200 | Location not disclosed | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 8× NVIDIA H200 | FIN-02 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaProvider API | 1× NVIDIA H200 SXM | Location not disclosed | 141GB HBM3e | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 1× NVIDIA H200 | FIN-02 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 1× NVIDIA H200 | Location not disclosed | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 4× NVIDIA H200 | FIN-01 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 2× NVIDIA H200 | FIN-03 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaProvider API | 2× NVIDIA H200 SXM | Location not disclosed | 141GB HBM3e | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 4× NVIDIA H200 | Location not disclosed | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 2× NVIDIA H200 | FIN-01 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaProvider API | 4× NVIDIA H200 SXM | Location not disclosed | 141GB HBM3e | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 2× NVIDIA H200 | FIN-02 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 8× NVIDIA H200 | Location not disclosed | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 4× NVIDIA H200 | FIN-02 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 1× NVIDIA H200 | FIN-03 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 1× NVIDIA H200 | FIN-01 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 8× NVIDIA H200 | FIN-01 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaProvider API | 8× NVIDIA H200 SXM | Location not disclosed | 141GB HBM3e | BURST · SPOT | $2 | Not reported | View ↗ |
| VerdaPublic pricing feed | 8× NVIDIA H200 | FIN-03 | 141GB | BURST · SPOT | $2 | Not reported | View ↗ |
 AWSPublic pricing feed | 8× NVIDIA H200 | ap-northeast-2 | 141GB | BURST · SPOT | $2.153Modeled | Not reported | View ↗ |
| Vast.aiMarketplace | 8× NVIDIA H200 | Japan, JP | 141GB | BURST · SPOT | $2.174 | Available | View ↗ |
EXexcesssupplyPublic pricing feed | 8× NVIDIA H200 | US, Raleigh, NC · raleigh-usa-2 | 141GB | STANDARD | $2.25 | Unavailable | View ↗ |
 amayaPublic pricing feed | 8× NVIDIA H200 | US, Dowagiac, MI · dowagiac-usa-1 | 141GB | STANDARD | $2.25 | Unavailable | View ↗ |
 LambdaPublic pricing feed | 1× NVIDIA H200 | us-west-3 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
| LambdaPublic pricing feed | 1× NVIDIA H200 | australia-east-1 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
SHShadeformPublic pricing feed | 1× NVIDIA H200 | Location not disclosed | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
| LambdaPublic pricing feed | 1× NVIDIA H200 | asia-northeast-1 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
| LambdaPublic pricing feed | 1× NVIDIA H200 | us-midwest-2 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
| LambdaPublic pricing feed | 1× NVIDIA H200 | us-east-1 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
| LambdaPublic pricing feed | 1× NVIDIA H200 | us-east-3 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
| LambdaPublic pricing feed | 1× NVIDIA H200 | asia-south-1 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
| LambdaPublic pricing feed | 1× NVIDIA H200 | us-south-2 | 96GB | STANDARD | $2.29 | Not reported | View ↗ |
Frequently Asked Questions
Why choose the NVIDIA H200?
141GB HBM3e and 4.8 TB/s bandwidth ease memory pressure for large context windows and bigger models without dropping to CPU offload.
When is the NVIDIA H200 not a good fit?
The premium over H100 rarely pays off for models that already fit comfortably in 80GB.
How much does the NVIDIA H200 cost per hour?
NVIDIA H200 on-demand pricing currently runs $0.50/hr to $21.20/hr per GPU, depending on the provider and instance type. Interruptible spot capacity starts around $0.50/hr.
How much does the NVIDIA H200 cost per month?
At 720 hours per month, one NVIDIA H200 runs about $360 to $15,264 per month on demand.
Which cloud providers offer the NVIDIA H200?
We currently track 434 NVIDIA H200 listings across 50 providers. Pricing and availability vary by region and billing model.
Alternatives to NVIDIA H200
NVIDIA A100
Previous-generation workhorse — more affordable, widely available.
From $1.29 /hr · 1 providers
NVIDIA Tesla P100
Pascal-era HBM2 card — budget training and inference.
From $1.27 /hr · 2 providers
NVIDIA H100 80GB
The standard data-center GPU for large-scale AI training and inference.
From $1.55 /hr · 55 providers