Rent dedicated GPUs with crypto. No KYC. Nobody looking over your shoulder.
GPU Rent Hub rents dedicated bare-metal GPU servers by the week or the month, paid only in Bitcoin, Ethereum, USDT, USDC, Monero, Solana or Litecoin, with no identity verification and no email address. H100, H200, B200, RTX 5090 and a dozen other cards in three US datacenters, from $190 a month. You get root on the whole machine. We meter power and bandwidth — never your workload.
Figures as of 2026-09-01. GPU count is the sum of DFW-1, IAD-1 and PDX-1.
DFW-1 · Cage 4 · Rack 17 — live occupancy
What you get for a flat monthly price
Hourly GPU clouds are built for bursts. We're built for the other 95% of the time: training runs that last weeks, inference that never stops, and people who would rather not upload a passport to rent a graphics card. The arithmetic is worked through on the monthly versus hourly GPU rental page.
Weekly & monthly terms
One flat price, paid up front. No per-second meter, no surprise egress bill, no spot eviction at 3 AM. A month with us costs 30–45% less than the same card on-demand elsewhere.
Crypto only
BTC, ETH, USDT, USDC, XMR, SOL and LTC. Prices are set in USD and settled at the spot rate when your invoice is generated. No cards, no wires, no chargebacks, no payment processor in the loop. See exactly how to pay for GPU rental with crypto.
No KYC
A username and a password is the whole account. Email is optional and only used for recovery if you add one. We don't ask who you are, where you live, or what you're building. That is what no-KYC GPU hosting means in practice.
No workload inspection
We don't look at what runs on your GPU. No process monitoring, no disk scanning, no traffic inspection. Our telemetry is power draw, temperature and bandwidth totals — because that is what we bill and cool.
The fleet and today's monthly prices
Every instance is bare metal: the whole card, the whole PCIe lane, local NVMe, a public IPv4 and root. Multi-GPU nodes are NVLink or NVSwitch wherever the silicon supports it. Monthly rates run from $190 for an RTX A5000 to $3,890 for a B200. To rent an H100 SXM is $1,690 a month, which is $2.35 per GPU-hour if you keep it busy for all 720 of them.
| GPU | VRAM | vCPU | RAM | NVMe | Weekly | Monthly | Availability | |
|---|---|---|---|---|---|---|---|---|
| B200 NewNVIDIA B200 SXM | 180 GB HBM3e | 28 | 283 GB | 3 TB | $1,090/wk | $3,890/mo | 8 left | View |
| H200 SXMNVIDIA H200 SXM 141GB | 141 GB HBM3e | 24 | 251 GB | 2 TB | $640/wk | $2,290/mo | 19 available | View |
| H100 SXMNVIDIA H100 SXM 80GB | 80 GB HBM3 | 20 | 251 GB | 2 TB | $475/wk | $1,690/mo | 63 available | View |
| A100 SXMNVIDIA A100 SXM 80GB | 80 GB HBM2e | 16 | 125 GB | 1 TB | $280/wk | $990/mo | 78 available | View |
| L40SNVIDIA L40S 48GB | 48 GB GDDR6 ECC | 16 | 94 GB | 500 GB | $148/wk | $520/mo | 74 available | View |
| RTX 5090NVIDIA GeForce RTX 5090 32GB | 32 GB GDDR7 | 12 | 62 GB | 500 GB | $155/wk | $540/mo | 100 available | View |
| RTX 4090NVIDIA GeForce RTX 4090 24GB | 24 GB GDDR6X | 8 | 47 GB | 250 GB | $110/wk | $390/mo | 265 available | View |
| RTX A6000NVIDIA RTX A6000 48GB | 48 GB GDDR6 ECC | 8 | 47 GB | 500 GB | $82/wk | $290/mo | 65 available | View |
How does renting a GPU with crypto work?
Four steps, and a root shell in under five minutes. The whole flow is designed so that the only thing you ever type into our site is a username, a password, and an SSH public key.
Pick a username
That's the sign-up. No email, no phone, no verification link. Add a recovery email or a PGP key later if you want one.
Fund your balance
Send any supported coin to the address we generate for you. The dollar value is credited at the spot rate once the network confirms.
Choose GPU, region, term
Weekly or monthly. Start from a template — PyTorch, vLLM, ComfyUI, Ollama, bare Ubuntu — or bring your own image.
SSH in
Provisioning usually takes under 90 seconds. You get a public IPv4, your key installed, and the entire machine to yourself.
Three US datacenters. Our racks, our hardware.
We don't resell anyone. Every GPU on this site sits in a cage we lease, on a server we own, behind switches we configure. Tier III facilities with 2N power, N+1 cooling and diverse fibre to multiple upstream carriers.
Stock numbers on the catalogue are live — they come from the same inventory system the dashboard deploys against. Cage, carrier and power detail for each site is in the facility notes.
- Opened
- March 2021
- Capacity
- 3,800 GPUs · 11 MW
- Network
- 2 × 100G · Cogent, Lumen, HE
- Power
- 2N UPS · N+1 diesel
- Opened
- November 2022
- Capacity
- 1,600 GPUs · 5 MW
- Network
- 2 × 100G · Lumen, Zayo, GTT
- Power
- 2N UPS · N+1 diesel
- Opened
- July 2024
- Capacity
- 900 GPUs · 3 MW
- Network
- 2 × 100G · Lumen, HE, Zayo
- Power
- Hydro · 2N UPS
Priced in dollars. Settled in whatever you hold.
Your balance is a USD ledger. Top it up in any coin below and we credit the dollar value at the spot rate once the network confirms. Instances renew from that balance automatically, so a single deposit can run for months. Monero deposits reach our own node and clear in ten confirmations, which is the setup described on Monero GPU hosting.
- Minimum deposit
- $20 equivalent
- Confirmations
- BTC 1 · ETH 12 · USDT/USDC 12 · XMR 10 · SOL 32 · LTC 6
- Renewal
- Drawn from balance 24 h before term end
- Grace period
- 72 h with instance stopped, data kept 7 days
- Unused balance
- Withdrawable to the coin you deposited
- Invoices
- Plain-text, downloadable, no name on them
- Chargebacks
- Not a thing. Neither are account freezes.
Own the card. We'll keep it cool.
Since 2026 we also sell GPUs outright, new and from our own retired fleet. Have them shipped, or leave them in our racks under a colocation plan and treat them like any other instance — except the monthly bill drops to power and rack space. GPU colocation starts at $95 per card per month.
Buy & ship
Freight to anywhere we can legally ship. Crypto paid up front, tracking number in your dashboard, pass-through manufacturer warranty.
Buy & colocate
From $95/GPU/month for power, cooling, 100G and remote hands. Your hardware, our cage, same dashboard, same no-questions policy.
Which GPU do you actually need?
Each guide starts from the arithmetic: how much memory the work needs, what sets the speed, and the cheapest card in the catalogue that clears the bar. No benchmarks we cannot source.
Five years of people who read the fine print.
We don't run case studies — most customers would rather not be named, and that's rather the point. These are quotes from support tickets, shared with permission.
Moved a 3-month fine-tuning job from an hourly provider. Same 8×H100 node, roughly half the bill, and nobody paused the instance for a "compliance review" in week two.
The account is a username. The invoice is a number. I've paid in XMR for three years and the only email I've ever received from them was the one I asked for as a recovery address.
Weekly 4090s for render bursts, one L40S monthly for the API. Bandwidth included, no egress math, and the status page is honest — which is rarer than it should be.
The ones we get every week.
Do I really not need an email address?
What do you actually see on my instance?
What happens when my term ends?
Can I switch from weekly to monthly?
Is the hardware really dedicated?
What if a coin's price moves while my deposit confirms?
Pick a card. Pick a month. That's the contract.
Stock shown on this site is live. If the card you want is on the waitlist, join it from the dashboard — we usually clear waitlists inside two weeks.