Quick answer
How do I rent a GPU for AI, step by step?
To rent a GPU for AI, pick the GPU whose VRAM holds your model plus about 20%, open an account, add credit, start an instance, connect over SSH, and copy your results off before you delete it. With GPUwithCrypto the account is a number, credit is topped up in crypto from $25, and an H100 80GB costs $2.32/hour as of 29 October 2026, billed per minute.
Renting a GPU with GPUwithCrypto at a glance
| GPUs | NVIDIA H100 80GB, H200 141GB, B200, A100 80GB, L40S 48GB, GeForce RTX 4090 24GB and RTX 5090 32GB, rented by the minute |
|---|---|
| Price rule | 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 |
| Billing | 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 |
| Payment | Crypto only: BTC, USDT, USDC, ETH, SOL, LTC, minimum top-up $25 |
| KYC | None: no ID, no selfie, no card |
| Instance | Each instance gives you whole GPUs, never a time-shared slice or a MIG partition. You get root SSH access to Ubuntu with the NVIDIA driver, CUDA and PyTorch installed, or start from your own Docker image |
| How to try a GPU | There is no free tier: a 10-hour block is the smallest purchase to try a GPU, and unused hours never expire |
How do I rent a GPU for AI with GPUwithCrypto?
Choose the GPU by VRAM
Add 20% to the model's weights (parameters × bytes per parameter) and pick the smallest GPU that holds it: a 24 GB RTX 4090 for 7–8B models, a 48 GB L40S for 32B in FP8, a 141 GB H200 for 70B in FP8. The VRAM guide has the full table.
Get an account number
Sign in with an account number; an email address is optional. Your account number is how you sign in, so store it carefully. With an optional email on file you can recover it; with no email, a lost number is gone for good.
Top up in crypto
Send at least $25 in BTC, USDT, USDC, ETH, SOL, LTC, on the exact network shown on the payment page. The balance is credited after the network confirmations.
Add your SSH key
You add your SSH public key in the dashboard; we never see or store your private key.
Start an instance
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. Stock changes through the day: the dashboard shows which GPUs are free right now, and nothing is charged until an instance starts. Once your payment is confirmed, you launch an instance from the dashboard; our team prepares it at the GPU provider and its SSH details appear there as soon as it runs. Billing starts only then.
Connect over SSH and run your job
Each instance gives you whole GPUs, never a time-shared slice or a MIG partition. You get root SSH access to Ubuntu with the NVIDIA driver, CUDA and PyTorch installed, or start from your own Docker image. Run
nvidia-smifirst to check the GPU model and its memory, and start long jobs insidetmuxso a dropped connection does not stop them.Save your outputs, then stop or delete
Every instance has local NVMe disk, sized for its GPU and shown before you start. The disk is wiped when you delete the instance, so copy your checkpoints out first. 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.
# from your computer: connect (host and SSH port are shown in the dashboard)
ssh -p SSH_PORT root@HOST
# on the instance: GPU model, memory and driver version
nvidia-smi
# run long jobs inside tmux (install it with: apt-get install -y tmux)
tmux new -s job
# from your computer: copy results back before you delete the instance
rsync -avP -e "ssh -p SSH_PORT" root@HOST:/root/outputs/ ./outputs/
What should I compare when renting a GPU anywhere?
Compare like with like, then read the rules around the price. The eight points that change the real cost:
- Price type. On-demand, spot, community and marketplace prices are different products. Spot and marketplace GPUs can be taken back; compare on-demand with on-demand. See what is a spot GPU.
- Billing increment and minimum. On 29 October 2026, RunPod billed per second, Lambda and Thunder Compute per minute, DigitalOcean per second with a 5-minute minimum, and Spheron per minute after a 20-minute minimum runtime.
- Form factor. An H100 SXM moves 3,350 GB/s of memory bandwidth; an H100 PCIe 2,000 GB/s. Check which one the price is for.
- Whole GPU or slice. A MIG partition or a time-shared GPU is not the same product as a whole card.
- Storage. Disks are often billed apart: RunPod charged $0.10 per GB per month for container disk. Check whether the disk survives a stop and what happens on delete.
- Payment and ID. Whether crypto is accepted, whether paying in crypto triggers KYC, and whether a card is required.
- Refunds. Several providers never refund credit bought with crypto.
- Acceptable use. Read what the provider forbids before you build on it.
Which GPU clouds take crypto, and do they ask for ID?
From each provider's own billing docs and terms, read on 29 October 2026. "No ID step documented" means the pages we read describe none, not a promise that the provider never asks. Full list of 33 providers: GPU clouds that accept crypto.
| Provider | Accepts crypto | ID / KYC | Card needed | Credit bought with crypto |
|---|---|---|---|---|
| GPUwithCrypto | Yes: BTC, USDT, USDC, ETH, SOL, LTC | No | No | Never expires; hours of a failed instance go back to the balance |
| RunPod | Yes (Crypto.com named in its terms) | KYC before the first crypto payment | Auto-pay is card-only | Credits non-refundable |
| Vast.ai | Yes (BitPay, Crypto.com) | No ID step documented | Verified credit card per its billing FAQ | Not refunded |
| Salad | Yes (USDC, USDT, RENDER on Solana) | No ID step documented | — | Not refundable; credits expire after 12 months |
| Massed Compute | Yes (BTC) | No ID step documented | — | No auto-recharge for crypto |
| Spheron | Yes (USDC on six chains) | No ID step documented | — | No self-serve refunds; credits do not expire |
| Lambda | No: "only accepts major credit cards" | No ID step documented | Credit card (debit and prepaid declined) | — |
Prices checked on
How many GPU clouds accept crypto?
Fewer than half. According to GPUwithCrypto's check of 29 October 2026, 14 of the 33 GPU providers we checked accept crypto, and 4 of the 33 ask for ID in some cases. Only RunPod ties a KYC step to paying in crypto, in its billing docs. None documents an ID check for every customer at sign-up. More: no-KYC GPU cloud and renting a GPU without a credit card.
How do I keep the bill down?
Right-size the GPU, stop it when idle and test before a long run. A model that fits a 48 GB L40S does not need an 80 GB H100. 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, so stop the instance between sessions. 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. Before a multi-day job, run it for an hour on the same GPU and measure your steps per hour. All prices: pricing.
Questions about renting a GPU
Can I rent a GPU without a credit card?
Yes. GPUwithCrypto takes only crypto (BTC, USDT, USDC, ETH, SOL, LTC) and asks for no card and no ID. At other providers, crypto sometimes comes with conditions: RunPod asks for KYC before the first crypto payment, and Vast.ai's billing FAQ asks for a verified card.
How fast does a rented GPU start?
Once your payment is confirmed, you launch an instance from the dashboard; our team prepares it at the GPU provider and its SSH details appear there as soon as it runs. Billing starts only then. The top-up is credited once the network has confirmed the transaction; the usual time per coin is shown on the payment page.
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. Copy your outputs off before the hours run out, or buy another block.
Is my data deleted when I delete the instance?
Yes. Every instance has local NVMe disk, sized for its GPU and shown before you start. The disk is wiped when you delete the instance, so copy your checkpoints out first. For billing we record each instance’s GPU model, host, start and stop time; these lines are kept for 30 days. We do not read the files on your disk. Account number, orders and top-ups stay while the account exists, then only as long as accounting law asks. Authorities receive records only when a law compels us.
What can I not do with a rented GPU?
No cryptocurrency mining, no password or hash cracking, no attacks on other systems (DDoS, scanning, exploitation), no malware, no spam, no child sexual abuse material, no intimate or deceptive images of real people without their consent, no weapons development and no work for sanctioned people or companies. The acceptable use policy has the details.
Sources (11)
- RunPod billing information docs.runpod.io/references/billing-information — RunPod (read 2026-10-29)
- RunPod pricing runpod.io/pricing — RunPod (read 2026-10-29)
- Vast.ai billing reference docs.vast.ai/documentation/reference/billing — Vast.ai (read 2026-10-29)
- SaladCloud billing docs.salad.com/general/explanation/billing — Salad (read 2026-10-29)
- Massed Compute crypto payments vm-docs.massedcompute.com/docs/billing/crypto/automatic-payments — Massed Compute (read 2026-10-29)
- Spheron billing — Spheron (read 2026-10-29)
- Lambda billing docs.lambda.ai/public-cloud/manage-billing — Lambda (read 2026-10-29)
- Lambda pricing lambda.ai/pricing — Lambda (read 2026-10-29)
- DigitalOcean GPU Droplet pricing digitalocean.com/pricing/gpu-droplets — DigitalOcean (read 2026-10-29)
- Spheron pricing spheron.network/pricing — Spheron (read 2026-10-29)
- NVIDIA H100 specifications — NVIDIA (read 2026-10-29)