The cheapest H200 on the table right now is $1.00 per GPU-hour at Vast.ai (spot), out of 57 offers from 27 providers. The H200 carries 141 GB of VRAM on NVIDIA's Hopper generation. That is peer/marketplace capacity — rented from other users' machines, and not directly comparable to a dedicated instance.
On-demand pricing for the same card spans $2.74 to $10.60 per GPU-hour — a 3.9× spread between the cheapest and the most expensive provider for identical silicon. That gap is the entire reason this table exists.
Cheapest by pricing model: on-demand $2.74 at Vast.ai · spot $1.00 at Vast.ai · reserved $2.10 at fal.
| Provider | Tier | $/GPU-hr | Pricing | Node | Commit | In stock | |
|---|---|---|---|---|---|---|---|
| Vast.ai | marketplace | $1.00 | spot | 1× GPU | — | yes | rent |
| Vast.ai | marketplace | $1.33 | spot | 2× GPU | — | yes | rent |
| AWS | dedicated | $1.85 | spot | 8× GPU | — | — | rent |
| Verda | dedicated | $2.00 | spot | 1–8× GPU | — | — | rent |
| fal | dedicated | $2.10 | reserved | 1× GPU | — | — | rent |
| Nebius | dedicated | $2.45 | reserved | 8× GPU | — | — | rent |
| Vast.ai | marketplace | $2.58 | spot | 4, 8× GPU | — | yes | rent |
| Vast.ai | marketplace | $2.74 | on-demand | 1× GPU | — | yes | rent |
| Seeweb | dedicated | $3.03 | on-demand | 1× GPU | — | — | rent |
| Lium.io | marketplace | $3.20 | on-demand | 8× GPU | — | yes | rent |
| DigitalOcean | dedicated | $3.40 | reserved | 8× GPU | 12 mo | — | rent |
| Civo | dedicated | $3.49 | on-demand | 1, 8× GPU | — | — | rent |
| Vast.ai | marketplace | $3.58 | on-demand | 2, 4× GPU | — | yes | rent |
| Runpod Community | marketplace | $3.59 | on-demand | 1× GPU | — | yes | rent |
| Runpod Community | marketplace | $3.59 | spot | 1× GPU | — | yes | rent |
| GPU.ai | dedicated | $3.59 | on-demand | 1× GPU | — | yes | rent |
| Massed Compute | dedicated | $3.62 | on-demand | 1–8× GPU | — | — | rent |
| Runpod | dedicated | $3.79 | on-demand | 1× GPU | — | yes | rent |
| Runpod | dedicated | $3.79 | spot | 1× GPU | — | yes | rent |
| Vast.ai | marketplace | $3.95 | on-demand | 8× GPU | — | yes | rent |
| Hyperstack | dedicated | $3.99 | on-demand | 1× GPU | — | — | rent |
| Verda | dedicated | $4.00 | on-demand | 1–8× GPU | — | — | rent |
| Beam | dedicated | $4.09 | on-demand | 1× GPU | — | yes | rent |
| Crusoe | dedicated | $4.29 | on-demand | 8× GPU | — | — | rent |
| DigitalOcean | dedicated | $4.47 | on-demand | 8× GPU | — | — | rent |
| Nebius | dedicated | $4.50 | on-demand | 8× GPU | — | — | rent |
| fal | dedicated | $4.50 | on-demand | 1× GPU | — | — | rent |
| Prime Intellect | dedicated | $4.50 | on-demand | 1× GPU | — | yes | rent |
| Modal | dedicated | $4.54 | on-demand | 1× GPU | — | yes | rent |
| LeaderGPU | dedicated | $4.76 | reserved | 1× GPU | 1 mo | — | rent |
| Spheron | dedicated | $4.79 | on-demand | 1× GPU | — | — | rent |
| Together AI | dedicated | $4.99 | reserved | 8× GPU | — | — | rent |
| Google Cloud | dedicated | $5.31 | spot | 1× GPU | — | — | rent |
| Replicate | dedicated | $5.49 | reserved | 1–8× GPU | — | yes | rent |
| Together AI | dedicated | $5.99 | on-demand | 8× GPU | — | — | rent |
| CoreWeave | dedicated | $6.30 | on-demand | 8× GPU | — | — | rent |
| OVHcloud | dedicated | $6.73 | on-demand | 4× GPU | — | — | rent |
| AWS | dedicated | $7.91 | on-demand | 8× GPU | — | — | rent |
| Google Cloud | dedicated | $9.31 | on-demand | 1× GPU | — | — | rent |
| Oracle Cloud | dedicated | $10.00 | on-demand | 8× GPU | — | — | rent |
| Azure | dedicated | $10.60 | on-demand | 8× GPU | — | — | rent |
| Azure | dedicated | $10.60 | spot | 8× GPU | — | — | rent |
How much does an H200 cost per hour?
As of the latest scrape, $1.00 per GPU-hour at Vast.ai is the cheapest H200 offer across 27 tracked cloud providers. The cheapest on-demand rate is $2.74 per GPU-hour at Vast.ai, and the most expensive on-demand rate is $10.60 at Azure.
Which cloud is cheapest for the H200?
Vast.ai at $1.00 per GPU-hour on spot capacity. Prices move constantly, so this page is regenerated from the live scrape every 15 minutes. Note that marketplace tiers are peer hardware and are not directly comparable to dedicated capacity.
Is spot H200 capacity cheaper?
Yes — the cheapest spot H200 is $1.00 per GPU-hour at Vast.ai, -63% versus the cheapest on-demand rate of $2.74. Spot instances can be reclaimed by the provider at any time, so they suit checkpointed training and batch inference rather than long-lived services.
How many GB of VRAM does the H200 have?
141 GB, on the Hopper architecture. Multi-GPU instances multiply that: an 8× H200 node exposes 1128 GB of GPU memory in total.