Try
10 GPU-hours
$30.70
$3.07/hour
- Test your setup end to end
- Same GPU and image as a long run
- Unused hours never expire
Quick answer
How much does an H200 cost per hour?
As of 29 October 2026, an NVIDIA H200 141GB costs from $3.62 to $10.60 per GPU-hour on demand across 13 providers, with a median of $5.29. On GPUwithCrypto it costs $3.07/hour, billed per minute and paid in crypto with no KYC. Interruptible H200s start lower: Nebius lists preemptible H200s from $0.79.
How much does an H200 cost per hour at each provider?
One row per provider at its cheapest on-demand H200 configuration, sorted by price. The most expensive on-demand H200 we found (Google Cloud, $10.60 per GPU in an 8-GPU VM) costs 2.9 times the cheapest.
| Provider | Configuration | On-demand $/GPU-hour | Spot / marketplace $/GPU-hour | Billing increment | Accepts crypto | ID / KYC |
|---|---|---|---|---|---|---|
| GPUwithCrypto | H200 141GB, 1 to 8 GPUs; form factor shown per host | $3.07 | — (on-demand only) | Per minute | Yes | No |
| Massed Compute | H200 141GB NVL, 1 GPU | $3.62 | — | Hourly rate (increment not stated) | Yes (BTC) | None documented |
| Hyperbolic | H200 141GB (form factor not stated) | $3.99 | — | — | Unknown | None documented |
| Crusoe | H200 SXM (HGX) | $4.29 | Contact sales | Per hour or per second | No | None documented |
| DigitalOcean (Paperspace) | H200 SXM (HGX), 1 or 8 GPUs | $4.47 | — | Per second, 5-minute minimum | Yes (stablecoins, limited) | None documented |
| Jarvislabs | H200 SXM, 1 GPU | $4.59 | $2.29 (spot) | Per minute | No | None documented |
| Spheron | H200 141GB (form factor not stated) | $4.80 | $3.35 (spot) | Per minute, 20-minute minimum | Yes (USDC) | None documented |
| RunPod | H200 141GB (form factor not stated), 1 GPU | $5.29 | $3.59 (Community Cloud) | Per second | Yes | KYC before first crypto payment |
| Nebius | H200 SXM (HGX) | $5.40 | From $0.79 (preemptible) | — | No | None documented |
| Verda (formerly DataCrunch) | H200 SXM, 1 GPU | $6.19 | $3.09 (spot) | — | No | None documented |
| OVHcloud | H200 141GB (form factor not stated), 8-GPU instance ÷ 8 | $6.20 | — | Per second, 1-minute minimum | No | Sometimes |
| CoreWeave | H200 SXM, 8-GPU instance ÷ 8 | $6.31 | $2.62 (spot) | — | Unknown | Unknown |
| AWS | H200, 8 GPUs (p5en.48xlarge) ÷ 8 | $7.91 | — | — | No | None documented |
| Google Cloud | H200 SXM, 8-GPU VM ÷ 8 (a3-ultragpu-8g) | $10.60 | $6.36 (spot) | Per second, 1-minute minimum | No | None documented |
| Vast.ai | H200 NVL and H200, marketplace | — | $3.00 for an H200 NVL, $3.77 for an H200 (Vast.ai's public price feed, cheapest rentable offers) | Per second | Yes (BitPay, Crypto.com) | None documented |
| Shadeform | H200 SXM, 8 GPUs, resold from other clouds | — | $4.00 (cheapest available listing) | — | Yes (USDC, per Stripe) | None documented |
Prices checked on
USD per GPU-hour, taxes excluded, rounded to the cent. Hyperscaler prices cover the whole VM (AWS US East, Google Cloud Iowa). 'None documented' means no ID step in the official sign-up, billing or terms pages we read, not a promise that a provider never asks. RunPod's Community Cloud is community-hosted.
Is the H200 worth the extra cost over an H100?
Yes when your model does not fit in 80 GB. According to GPUwithCrypto's price check of 29 October 2026, a median on-demand H200 costs 27% more per hour than a median H100 ($5.29 against $4.16). It has 76% more memory (141 GB against 80 GB) and 43% more bandwidth than the H100 SXM (4,800 against 3,350 GB/s).
Compute is the same: NVIDIA gives the H200 SXM and the H100 SXM the same FP8 rating, 3,958 TFLOPS with sparsity or about 1,979 dense. So the H200 pays off for big models and long contexts, not for small ones. Llama 3.1 70B in FP8 needs about 84 GB: one H200, or two H100s. The H100 vs H200 comparison has the full verdict, and the H100 price page the H100 rows.
Why are some marketplace and spot H200s cheaper than ours?
Because the host sets the price on a marketplace, and the provider can take a spot GPU back. On 29 October 2026, the cheapest rentable H200 NVL offer in Vast.ai's public price feed was $3.00 per hour, below our $3.07. Nebius listed preemptible H200s from $0.79 and Jarvislabs an H200 spot at $2.29.
These prices suit jobs that save checkpoints and can restart. For a long run or an always-on inference server, compare on-demand prices. Instances are on-demand, not spot: nobody can take your GPU back while your balance covers it, and the instance runs until you stop it.
Which H200 sets the reference price: SXM or NVL?
The reference is Massed Compute's H200 NVL at $3.62 per GPU-hour. The NVL is the PCIe version: same 141 GB and 4,800 GB/s as the SXM, but 600 W and a lower Tensor rating. The cheapest self-serve H200 SXM we found was Crusoe's at $4.29.
GPUwithCrypto has one H200 price for every host. Instances have 1 to 8 GPUs depending on the host. Before you start, the dashboard shows each host’s country, GPU count, form factor (SXM or PCIe) and interconnect.
How much does an H200 cost per day, week or month?
Our price for one GPU next to the lowest and the median self-serve on-demand price of 29 October 2026, for the same number of hours.
| Period | GPU-hours | GPUwithCrypto | Lowest elsewhere (Massed Compute) | Median elsewhere |
|---|---|---|---|---|
| 1 hour | 1 | $3.07 | $3.62 | $5.29 |
| 1 day | 24 | $73.68 | $86.88 | $126.96 |
| 1 week | 168 | $515.76 | $608.16 | $888.72 |
| 30 days | 720 | $2,210.40 | $2,606.40 | $3,808.80 |
Prices checked on
Our day, week and month match the 24, 168 and 720 GPU-hour blocks.
What are the H200's specs?
NVIDIA's official figures for the two H200 versions. NVIDIA prints Tensor figures with sparsity only; the dense figures below are half of those, so compare dense with dense.
| Spec | H200 SXM | H200 NVL (PCIe) |
|---|---|---|
| Memory | 141 GB HBM3e | 141 GB HBM3e |
| Memory bandwidth | 4,800 GB/s | 4,800 GB/s |
| FP8 Tensor, dense | 1,979 TFLOPS (3,958 sparse) | 1,670.5 TFLOPS (3,341 sparse) |
| FP16 / BF16 Tensor, dense | 989.5 TFLOPS (1,979 sparse) | 835.5 TFLOPS (1,671 sparse) |
| GPU-to-GPU link | NVLink 900 GB/s | NVLink bridge 900 GB/s |
| Max power | 700 W | 600 W |
NVIDIA labels the H200 figures 'Preliminary specifications. May be subject to change.'
What fits on one H200?
Memory needed for common open models, using weights × 1.2 for inference (EleutherAI's rule of thumb). The VRAM guide has the formulas and more models.
| Model and job | Memory needed | One H200 141GB? |
|---|---|---|
| Llama 3.1 70B inference, FP8 | ≈ 84 GB | ✓ |
| Llama 3.1 70B, FP8 weights plus one full 128K-token context (BF16 cache) | ≈ 113 GB | ✓ |
| Qwen2.5 72B inference, FP8 | ≈ 87.2 GB | ✓ |
| Llama 3.1 70B QLoRA fine-tune | ≈ 52.5 GB | ✓ |
| Llama 3.1 70B inference, BF16 | ≈ 168 GB | No: 2× H200 |
| Llama 3.1 70B LoRA fine-tune | ≈ 140 to 168 GB | No: 2× H200 |
| Llama 3.1 405B inference, FP8 | ≈ 486 GB | No: 4× H200 |
| DeepSeek-V3 or R1 inference, FP8 | ≈ 805 GB | No: one 8× H200 node (1,128 GB) |
What does an H200 cost on GPUwithCrypto?
$3.07/hour, at least 15% under its named on-demand reference: Massed Compute at $3.62, read on 29 October 2026. We do not claim to be the cheapest for every GPU; spot and marketplace offers (Vast.ai) can be lower. Instances are billed per minute at the hourly rate while they run, from the GPU-hours you bought for that GPU. Stopping an instance stops the charge at once. GPU-hours are sold in blocks of 10, 24, 168 or 720 hours for one GPU model, at the same hourly rate; instances of that GPU draw from them per minute, and unused hours never expire.
Popular
Day
24 GPU-hours
$73.68
$3.07/hour
- One day of continuous running
- A fine-tuning run or a batch job
- Unused hours never expire
Week
168 GPU-hours
$515.76
$3.07/hour
- One week of continuous running
- Training runs and evaluations
- Unused hours never expire
Month
720 GPU-hours
$2,210.40
$3.07/hour
- 30 days of continuous running
- An inference server running all month
- Unused hours never expire
In every plan: Whole GPUs, never shared · Root SSH, CUDA and PyTorch ready · Billed per minute · Unused hours never expire
H200 price questions
What is the cheapest H200 cloud right now?
On 29 October 2026 the lowest self-serve on-demand H200 price we found was $3.62, for Massed Compute's H200 NVL, and GPUwithCrypto charges $3.07. The cheapest H200 SXM was Crusoe's at $4.29. Spot H200s go lower but can be interrupted, and marketplace hosts set their own prices.
How much does an H200 cost per month?
720 GPU-hours cost $2,210.40 on GPUwithCrypto. At the median on-demand price of 29 October 2026 the same month cost $3,808.80, and $2,606.40 at the lowest.
Can a 70B model run on one H200?
Yes in FP8 or 4-bit. Llama 3.1 70B needs about 84 GB in FP8, which leaves room for long contexts on a 141 GB H200. In BF16 it needs about 168 GB, so two H200s.
Can I rent an H200 with crypto?
Yes. GPUwithCrypto takes BTC, USDT, USDC, ETH, SOL, LTC with no ID, no selfie and no card. Of the 13 self-serve H200 providers we checked, 4 accept crypto: Massed Compute, DigitalOcean, Spheron and RunPod. RunPod asks for KYC before the first crypto payment.
Which H200 do I get on GPUwithCrypto, SXM or NVL?
It depends on the host. Instances have 1 to 8 GPUs depending on the host. Before you start, the dashboard shows each host’s country, GPU count, form factor (SXM or PCIe) and interconnect. Both versions have 141 GB of HBM3e and 4,800 GB/s.
Sources (18)
- Massed Compute — VM pricing vm.massedcompute.com/pricing — Massed Compute (read 2026-10-29)
- Hyperbolic — GPU marketplace hyperbolic.ai/marketplace — Hyperbolic (read 2026-10-29)
- Crusoe — Cloud pricing crusoe.ai/cloud/pricing — Crusoe (read 2026-10-29)
- DigitalOcean — GPU Droplets pricing digitalocean.com/pricing/gpu-droplets — DigitalOcean (read 2026-10-29)
- Jarvislabs — Pricing jarvislabs.ai/pricing — Jarvislabs (read 2026-10-29)
- Spheron — Pricing spheron.network/pricing — Spheron (read 2026-10-29)
- Runpod — Pricing runpod.io/pricing — RunPod (read 2026-10-29)
- Nebius — Prices nebius.com/prices — Nebius (read 2026-10-29)
- Verda — Pricing verda.com/pricing — Verda (read 2026-10-29)
- OVHcloud — Public Cloud prices ovhcloud.com/en/public-cloud/prices — OVHcloud (read 2026-10-29)
- CoreWeave — Pricing coreweave.com/pricing — CoreWeave (read 2026-10-29)
- AWS — EC2 on-demand price list, US East (N. Virginia) — Amazon Web Services (read 2026-10-29)
- Google Cloud — Accelerator-optimized VM pricing — Google Cloud (read 2026-10-29)
- Vast.ai public GPU pricing feed — Vast.ai (read 2026-10-29)
- Shadeform — Instance types API api.shadeform.ai/v1/instances/types — Shadeform (read 2026-10-29)
- Runpod — Billing information docs.runpod.io/references/billing-information — RunPod (read 2026-10-29)
- NVIDIA H200 Tensor Core GPU — NVIDIA (read 2026-10-29)
- Transformer Math 101 — EleutherAI (read 2026-10-29)