Try
10 GPU-hours
$23.20
$2.32/hour
- Test your setup end to end
- Same GPU and image as a long run
- Unused hours never expire
Quick answer
How much does a GPU cost per hour at GPUwithCrypto?
GPUwithCrypto rents cloud GPUs by the minute, paid in crypto with no KYC. As of 29 October 2026, one GPU-hour costs $2.32 for an H100 80GB, $3.07 for an H200 and $4.61 for a B200. An A100 80GB costs $0.92, an L40S $0.82, an RTX 4090 $0.28 and an RTX 5090 $0.42. Each price is at least 15% under a named on-demand reference read on 29 October 2026.
How does each price compare with other GPU clouds?
Our hourly price next to the lowest and the median self-serve on-demand price of the providers we checked on 29 October 2026. We do not claim to be the cheapest for every GPU; spot and marketplace offers (Vast.ai) can be lower. Each GPU has its own page with every provider row: H100, H200, A100, L40S, RTX 4090, RTX 5090.
| GPU | GPUwithCrypto | Lowest on-demand elsewhere | Median elsewhere | Providers checked | Below the median |
|---|---|---|---|---|---|
| NVIDIA H100 80GB | $2.32/hour | $2.73 (Massed Compute, H100 80GB (form factor not stated)) | $4.16 | 16 | 44% |
| NVIDIA H200 141GB | $3.07/hour | $3.62 (Massed Compute, H200 141GB NVL) | $5.29 | 13 | 42% |
| NVIDIA B200 | $4.61/hour | $5.43 (Massed Compute, 8-GPU node ÷ 8) | $8.50 | 9 | 46% |
| NVIDIA A100 80GB | $0.92/hour | $1.09 (Thunder Compute, A100 80GB (form factor not stated)) | $2.23 | 13 | 59% |
| NVIDIA L40S 48GB | $0.82/hour | $0.97 (Massed Compute) | $1.67 | 11 | 51% |
| RTX 4090 24GB | $0.28/hour | $0.33 (Salad, High Priority) | $0.61 | 2 | 54% |
| RTX 5090 32GB | $0.42/hour | $0.50 (Salad, High Priority) | $0.84 | 3 | 50% |
Prices checked on
Per GPU-hour, USD, taxes excluded. Each provider counts once, with its lowest on-demand price for that GPU in any form factor. Prices listed only for multi-GPU instances are divided by the GPU count. Salad's RTX prices rise on 1 November 2026; the table shows the price valid on 29 October.
What does an H100 80GB cost?
$2.32/hour, sold in blocks of 10 to 720 GPU-hours and billed per minute. Training and fine-tuning up to 70B parameters, fast LLM inference. All H100 prices across 16 providers.
Popular
Day
24 GPU-hours
$55.68
$2.32/hour
- One day of continuous running
- A fine-tuning run or a batch job
- Unused hours never expire
Week
168 GPU-hours
$389.76
$2.32/hour
- One week of continuous running
- Training runs and evaluations
- Unused hours never expire
Month
720 GPU-hours
$1,670.40
$2.32/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
What does an H200 141GB cost?
$3.07/hour, sold in blocks of GPU-hours and billed per minute. Large-model inference and long contexts: a 70B model in FP16 fits on two GPUs. All H200 prices across 13 providers.
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
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
What does a B200 cost?
$4.61/hour with 180 GB HBM3e per GPU. Frontier-scale training and FP4/FP8 inference. Compared with 9 providers on the GPU price overview.
Try
10 GPU-hours
$46.10
$4.61/hour
- Test your setup end to end
- Same GPU and image as a long run
- Unused hours never expire
Popular
Day
24 GPU-hours
$110.64
$4.61/hour
- One day of continuous running
- A fine-tuning run or a batch job
- Unused hours never expire
Week
168 GPU-hours
$774.48
$4.61/hour
- One week of continuous running
- Training runs and evaluations
- Unused hours never expire
Month
720 GPU-hours
$3,319.20
$4.61/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
What does an A100 80GB cost?
$0.92/hour, billed per minute. Fine-tuning and inference on a budget, the most rented data-center GPU. All A100 prices, 40 GB and 80 GB.
Try
10 GPU-hours
$9.20
$0.92/hour
- Test your setup end to end
- Same GPU and image as a long run
- Unused hours never expire
Popular
Day
24 GPU-hours
$22.08
$0.92/hour
- One day of continuous running
- A fine-tuning run or a batch job
- Unused hours never expire
Week
168 GPU-hours
$154.56
$0.92/hour
- One week of continuous running
- Training runs and evaluations
- Unused hours never expire
Month
720 GPU-hours
$662.40
$0.92/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
What does an L40S 48GB cost?
$0.82/hour, billed per minute. Image and video generation, inference of models up to about 30B parameters. All L40S prices.
Try
10 GPU-hours
$8.20
$0.82/hour
- Test your setup end to end
- Same GPU and image as a long run
- Unused hours never expire
Popular
Day
24 GPU-hours
$19.68
$0.82/hour
- One day of continuous running
- A fine-tuning run or a batch job
- Unused hours never expire
Week
168 GPU-hours
$137.76
$0.82/hour
- One week of continuous running
- Training runs and evaluations
- Unused hours never expire
Month
720 GPU-hours
$590.40
$0.82/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
What does an RTX 4090 cost?
$0.28/hour, billed per minute. ComfyUI, Stable Diffusion, Flux and LoRA training at the lowest hourly price. All RTX 4090 prices.
Try
10 GPU-hours
$2.80
$0.28/hour
- Test your setup end to end
- Same GPU and image as a long run
- Unused hours never expire
Popular
Day
24 GPU-hours
$6.72
$0.28/hour
- One day of continuous running
- A fine-tuning run or a batch job
- Unused hours never expire
Week
168 GPU-hours
$47.04
$0.28/hour
- One week of continuous running
- Training runs and evaluations
- Unused hours never expire
Month
720 GPU-hours
$201.60
$0.28/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
What does an RTX 5090 cost?
$0.42/hour, billed per minute. Image and video generation and small-model inference with 32 GB of VRAM. All RTX 5090 prices.
Try
10 GPU-hours
$4.20
$0.42/hour
- Test your setup end to end
- Same GPU and image as a long run
- Unused hours never expire
Popular
Day
24 GPU-hours
$10.08
$0.42/hour
- One day of continuous running
- A fine-tuning run or a batch job
- Unused hours never expire
Week
168 GPU-hours
$70.56
$0.42/hour
- One week of continuous running
- Training runs and evaluations
- Unused hours never expire
Month
720 GPU-hours
$302.40
$0.42/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
How does per-minute billing work?
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. A 90-minute job is charged 90 minutes, not two full hours. When the GPU-hours of a GPU run out, its instances stop and their disks are kept; a new block of hours lets you start them again.
Billing steps differ from one cloud to the next. On 29 October 2026:
- Per second: RunPod, Salad and Vast.ai.
- Per second with a minimum: DigitalOcean (5 minutes), Google Cloud and OVHcloud (1 minute).
- Per minute: Lambda, Jarvislabs and Thunder Compute.
- Per minute after a 20-minute minimum runtime: Spheron.
- Hourly rate, increment not stated: Massed Compute.
For short test runs the increment matters as much as the hourly price. Stock changes through the day: the dashboard shows which GPUs are free right now, and nothing is charged until an instance starts.
How do blocks of GPU-hours work?
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. A block costs its hours at the hourly rate, with no volume discount and no premium. It fixes a budget for one job, such as a fine-tuning run or a month of inference.
Blocks are bought from your crypto balance. An instance of the block's GPU model draws from it per minute. Unused GPU-hours and balance never expire.
What does each block of GPU-hours cost?
| GPU | 10 GPU-hours | 24 GPU-hours (a day) | 168 GPU-hours (a week) | 720 GPU-hours (30 days) |
|---|---|---|---|---|
| H100 80GB | $23.20 | $55.68 | $389.76 | $1,670.40 |
| H200 141GB | $30.70 | $73.68 | $515.76 | $2,210.40 |
| B200 | $46.10 | $110.64 | $774.48 | $3,319.20 |
| A100 80GB | $9.20 | $22.08 | $154.56 | $662.40 |
| L40S 48GB | $8.20 | $19.68 | $137.76 | $590.40 |
| RTX 4090 | $2.80 | $6.72 | $47.04 | $201.60 |
| RTX 5090 | $4.20 | $10.08 | $70.56 | $302.40 |
Per GPU. Each block is for one GPU model, at that GPU's hourly rate.
How is the 15% rule computed?
Each GPU is priced at least 15% under a named reference: the on-demand price of the same GPU at a self-serve provider, read on 29 October 2026 (H100: Massed Compute at $2.73 per GPU-hour; each GPU's reference is named on its price page). Spot, interruptible, marketplace and contact-sales prices are not used as references, and some of them are lower. We do not claim to be the cheapest for every GPU.
The method has four steps:
- We read each provider's own pricing page, or the public price feed that page loads, on 29 October 2026.
- For each provider we keep its lowest self-serve on-demand price for the GPU, in any form factor. A price listed only for an 8-GPU instance is divided by 8.
- The lowest of those prices is the reference. Our price is the reference × 0.85, rounded down to the cent.
- Stale rows are dropped. TensorDock's public page says 'Last Updated: July 24, 2024', and Spheron marks its RTX 4090 price 'last known Sep 17'.
Example: the H100 reference is $2.73 at Massed Compute (H100 80GB (form factor not stated)), so our H100 costs $2.32.
According to GPUwithCrypto's price check of 29 October 2026, our seven prices rest on 67 self-serve on-demand prices from 18 providers. The gap to the median is widest on the A100 80GB (59% below) and narrowest on the H200 (42% below). Hyperscaler prices (AWS US East, Azure East US, Google Cloud Iowa) cover the whole virtual machine. All prices exclude taxes.
Which prices are left out, and can they be lower than ours?
Spot, preemptible, community, marketplace, reserved and contact-sales prices are left out of the reference, and some of them are lower than ours. What we found on 29 October 2026:
- Spot and preemptible: Nebius lists preemptible H100s from $0.79 per GPU-hour and Jarvislabs an H100 spot at $1.54. The provider can stop these instances.
- Marketplace: Vast.ai's public price feed showed a rentable H100 SXM offer at $1.73 and an RTX 4090 offer at $0.13. Independent hosts set these prices and run the machines.
- Community-hosted: RunPod's Community Cloud listed an H100 PCIe at $1.99, against $2.89 on its Secure Cloud.
- 'Starting at' and contact sales: Voltage Park's FAQ quotes an H100 'starting at' $1.99, but its on-demand pricing cards say 'Contact for pricing'. Crusoe's B200 is also contact sales.
- Commitments: DigitalOcean's 12-month reserved H100 costs $3.26 per GPU-hour, and CoreWeave advertises up to 60% off for committed usage.
We compare like with like. 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. The spot GPU guide explains when an interruptible GPU is the better deal.
How do I pay for a GPU in crypto?
Top up $25 or more
Choose the amount and the coin (BTC, USDT, USDC, ETH, SOL, LTC). The crypto amount and its rate are locked for 30 minutes.
Send it to the address shown, on the exact network
The payment page shows our deposit address for the coin and network you picked: BTC on Bitcoin, USDT on Tron (TRC-20) or Ethereum (ERC-20), USDC on Ethereum (ERC-20), ETH on Ethereum, SOL on Solana, LTC on Litecoin. Send only that coin on that network: funds sent on another network or in another coin can’t be recovered. The network fee comes on top of the amount.
Send the exact amount shown, then click “I’ve sent the payment”
Each top-up has its own exact amount (down to the last decimals): that is how we find your payment and link it to your account.
Credited once your payment is checked on the blockchain
Your balance is credited once we have checked the transaction on the blockchain, after the network has confirmed it. Typical network confirmation times:
Crypto Network Confirmations Usual time BTC Bitcoin 2 ≈ 20 min USDT Tron (TRC-20) 20 ≈ 1 min USDT Ethereum (ERC-20) 12 ≈ 5 min USDC Ethereum (ERC-20) 12 ≈ 5 min ETH Ethereum 12 ≈ 5 min SOL Solana 1 ≈ 1 min LTC Litecoin 6 ≈ 15 min Start your GPU
Pick a block of GPU-hours first and the top-up is set to its price, or to the minimum for a cheaper block. Once the payment is confirmed, launch an instance from the dashboard: its SSH details appear as soon as it runs. Credit on your balance does not expire and buys blocks of GPU-hours for any GPU. When a top-up is only made to reach the minimum, the unspent part comes back on request within 30 days of payment, in USDT or USDC. The rest of a balance is not paid out unless the law requires it.
Closed the tab? The pending top-up stays under Billing in your account: open it to see the address and the exact amount again.
Questions about GPU pricing
Is there a free trial or a minimum spend?
There is no free tier: a 10-hour block is the smallest purchase to try a GPU, and unused hours never expire. The minimum top-up is $25, and the plain balance pays for any GPU.
Which cryptocurrencies can I pay with?
BTC, USDT, USDC, ETH, SOL, LTC, with no ID, no selfie and no card. Of the 33 GPU clouds we checked on 29 October 2026, 14 accept crypto, and RunPod asks for KYC before the first crypto payment. The crypto comparison lists them all.
Are prices per GPU or per instance?
Per GPU. An instance with four H100s costs four times $2.32/hour, billed per minute. 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.
What happens when my GPU-hours run out?
When the GPU-hours of a GPU run out, its instances stop and their disks are kept; a new block of hours lets you start them again. Unused GPU-hours and balance never expire, so a paused project keeps its credit.
Why is a spot or marketplace GPU sometimes cheaper than yours?
Because the provider or the host can take it back, or sets its own price for its own machine. On 29 October 2026, Nebius listed preemptible H100s from $0.79 per GPU-hour. Our reference is the on-demand price, where the GPU stays yours while you pay.
Can I get my money back?
If an instance fails to start, or its GPU does not work and support cannot fix it, the hours it used go back to your balance. Unused GPU-hours and balance never expire. Top-ups are not paid out unless the law requires it; the part added only to reach the minimum top-up comes back on request within 30 days of payment, in USDT or USDC, as long as it is unspent. The refund policy has the details.
Sources (10)
- Massed Compute — VM pricing (H100, H200, B200, L40S references) vm.massedcompute.com/pricing — Massed Compute (read 2026-10-29)
- Thunder Compute — Pricing (A100 80GB reference) thundercompute.com/pricing — Thunder Compute (read 2026-10-29)
- SaladCloud price changes, November 2026 (RTX 4090 and RTX 5090 references) blog.salad.com/saladcloud-price-changes-november-2026 — Salad (read 2026-10-29)
- Runpod — Pricing (Secure Cloud and Community Cloud) runpod.io/pricing — RunPod (read 2026-10-29)
- Nebius — Prices (on-demand and preemptible) nebius.com/prices — Nebius (read 2026-10-29)
- Vast.ai public GPU pricing feed — Vast.ai (read 2026-10-29)
- Voltage Park — Pricing voltagepark.com/pricing — Voltage Park (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)
- Runpod — Billing information (crypto and KYC) docs.runpod.io/references/billing-information — RunPod (read 2026-10-29)