GPU rental prices · Hopper
NVIDIA H200 rental starts at $4.47 per GPU-hour (DigitalOcean, HGX H200) across the 2 offers we track from 2 providers. The typical price is $4.88, and one GPU running all month costs about $3,263 at the lowest rate. It is an H100 with 76% more memory and 43% more memory bandwidth, so it pays off when a model doesn't fit in 80 GB or when long contexts fill the cache.
On-demand prices
Every H200 offer in our catalog, cheapest first. Multi-GPU machines are divided by their GPU count, so you can compare one GPU with one GPU. Monthly cost assumes the GPU runs all 730 hours.
| Provider and offer | Version | GPU memory | GPUs | Per GPU-hour | Per month, 1 GPU | Checked |
|---|---|---|---|---|---|---|
| DigitalOceanNVIDIA HGX H200 GPU Droplet · 1 GPU | HGX H200 | 141 GB | 1 | $4.47 | $3,263 | 3 Nov 2026 |
| RunpodRunpod Pod · NVIDIA H200 | H200 | 141 GB | 1 | $5.29 | $3,862 | 3 Nov 2026 |
On-demand list prices before tax, from each provider's pricing page. Community and reserved prices, where a provider offers them, are lower; see the notes below. Full offer details, storage and billing terms.
Before you rent
The H200 SXM has the same 3,958 TFLOPS of sparse FP8 as the H100 SXM. You pay for memory: 141 GB instead of 80 GB, at 4.8 TB/s instead of 3.35 TB/s. Generating text reads the whole model from memory for every token, so the faster memory usually means quicker replies on the same model.
DigitalOcean charges $4.47 per GPU-hour for the H200 and $4.41 for the H100, so the extra memory costs almost nothing there. Runpod charges $5.29, against $3.99 for its H100 SXM.
Runpod's Community Cloud lists the H200 at $3.59 an hour. DigitalOcean offers it at $3.40 an hour on a 12-month commitment.
AI models
Memory to load each model size on 141 GB GPUs, with a 20% allowance for runtime buffers, and what that many GPUs cost per hour at the lowest rate above. Long contexts and many parallel users need more memory. More than one GPU only works if your inference server splits the model across them.
| Model size | 4-bit | 8-bit | 16-bit |
|---|---|---|---|
| 8BLlama 3.1 8B, Qwen3 8B | 4.5 GB1 GPU · $4.47/hr | 9.0 GB1 GPU · $4.47/hr | 17.9 GB1 GPU · $4.47/hr |
| 32BQwen3 32B, Gemma 4 31B | 17.9 GB1 GPU · $4.47/hr | 35.8 GB1 GPU · $4.47/hr | 71.6 GB1 GPU · $4.47/hr |
| 70BLlama 3.3 70B, Qwen 72B | 39.2 GB1 GPU · $4.47/hr | 78.3 GB1 GPU · $4.47/hr | 156.5 GB2 GPUs · $8.94/hr |
| 120Bgpt-oss-120b class | 67.1 GB1 GPU · $4.47/hr | 134.2 GB1 GPU · $4.47/hr | 268.3 GB2 GPUs · $8.94/hr |
| 685BDeepSeek-V3.2 | 382.8 GB3 GPUs · $13.41/hr | 765.6 GB6 GPUs · $26.82/hr | 1,531.1 GB11 GPUs · more than one 8-GPU server |
Weights only, from the parameter count. Check a specific model, context length and user count in the model hardware calculator.
Hardware
| Specification | H200 SXM | H200 NVL |
|---|---|---|
| GPU memory | 141 GB | 141 GB |
| Memory bandwidth | 4.8 TB/s | 4.8 TB/s |
| FP8 Tensor (sparse) | 3,958 TFLOPS | 3,341 TFLOPS |
| Max power | Up to 700 W | Up to 600 W |
| NVLink | 900 GB/s | 2- or 4-way bridge, 900 GB/s per GPU |
From NVIDIA's product pages. Tensor figures marked “sparse” assume structured sparsity; dense throughput is half.
Questions
$4.47 per GPU-hour is the lowest on-demand price we track, from DigitalOcean. The typical price is $4.88, and one GPU running all month costs about $3,263 at the lowest rate. Community and reserved prices go lower.
Yes. The weights take about 78.3 GB with a 20% allowance, which leaves roughly 60 GB of the 141 GB for the conversation cache. At 16-bit a 70B model needs 156.5 GB, or two H200s.
Compare
Prices and availability change, and we don't check stock. Confirm the current price and region with the provider before you rent. Sources: www.nvidia.com, www.runpod.io, www.digitalocean.com . See our affiliate disclosure.
