FrançaisFR
enEnglishfrFrançaisdeDeutschesEspañolptPortuguêsitItalianonlNederlandsplPolskitrTürkçeruРусскийukУкраїнськаja日本語ko한국어zh中文arالعربية
Connexion Déployer un GPU
GPU Rent HubGPUsRTX 5090

Rent a dedicated RTX 5090 by the month, paid in crypto

GPU Rent Hub rents a bare-metal server with one NVIDIA GeForce RTX 5090 32GB for 540 $US a month or 155 $US a week, in three United States datacenters, paid in Bitcoin, Ethereum, USDT, USDC, Monero, Solana or Litecoin, with no identity verification and no inspection of what you run on it.

Updated 2026-09-01
100 cards in stock now
540 $USPer month, flat
155 $USPer week
0,75 $USEquivalent per GPU-hour
32 GBGDDR7 at 1.79 TB/s
01 Specification

What you get with an RTX 5090 server

Every instance is a whole physical machine. The card is not shared, sliced or virtualised, and you get root on the host. The CPU, memory and NVMe listed here are yours alone, not a quota against a larger pool.

Best suited to:

  • Fast single-card inference on 7B–32B models
  • Diffusion and video generation
  • The highest memory bandwidth per dollar we rent

Compare every GPU we rent

RTX 5090 Blackwell / 2025
VRAM
32 GB GDDR7
Bandwidth
1.79 TB/s
Cores
21,760 CUDA · 680 Tensor · 170 RT
FP8
838 TFLOPS (sparse)
FP16
104.8 TFLOPS (FP32 shader)
Interconnect
PCIe Gen5 peer-to-peer
TDP
575 W
Host vCPU
12 dedicated
Host RAM
62 GB ECC
Local NVMe
500 GB encrypted
Network
100 Gbps port, 10 Gbps guaranteed
Traffic
Unmetered in, 20 TB out
02 Availability

Where the RTX 5090 is in stock

Stock comes from the same inventory the dashboard deploys against. A dash means the card is on the waitlist in that region, which we usually clear inside two weeks.

RegionIn stockLatency
DFW-1Dallas, TX48Chicago 22 ms · Los Angeles 34 ms · New York 38 ms
IAD-1Ashburn, VA30New York 7 ms · London 74 ms · Frankfurt 88 ms
PDX-1Hillsboro, OR22Seattle 5 ms · San Francisco 14 ms · Tokyo 92 ms

Read the datacenter specifications

03 Multi-GPU

How much is a multi-GPU RTX 5090 node?

Multi-GPU configurations sit on a single machine with PCIe Gen5 peer-to-peer between the cards. Two, four or eight cards on one node also take 5%, 8% or 10% off the per-card price.

NodeInterconnectWeeklyMonthly
8× RTX 5090PCIe Gen5PCIe peer-to-peer1 120 $US/wk3 890 $US/mo
04 Alternatives

Cards at a similar price

Every price below is a flat monthly rate for a dedicated card in the same datacenters, on the same terms.

GPUVRAMBandwidth MonthlyPer GPU-hour
L40SAda Lovelace48 GB864 GB/s520 $US/mo0,72 $US
RTX 6000 AdaAda Lovelace48 GB960 GB/s460 $US/mo0,64 $US
RTX 4090Ada Lovelace24 GB1.01 TB/s390 $US/mo0,54 $US
RTX A6000Ampere48 GB768 GB/s290 $US/mo0,40 $US
L4Ada Lovelace24 GB300 GB/s250 $US/mo0,35 $US

When monthly rental beats hourly billing

05 Questions

Renting an RTX 5090

How much does it cost to rent an RTX 5090 per month?
A dedicated NVIDIA GeForce RTX 5090 32GB server is $540 per month at GPU Rent Hub, or $155 for a one-week term. The price is flat and includes power, cooling, 12 vCPU, 62 GB of RAM, 500 GB of local NVMe, a public IPv4 and 20 TB of outbound traffic. Across a 720-hour month that works out to $0.75 per GPU-hour.
Can I rent an RTX 5090 with Bitcoin or Monero?
Yes. GPU Rent Hub accepts Bitcoin, Ethereum, USDT, USDC, Monero, Solana and Litecoin, and accepts nothing else. You fund a USD balance, the balance pays for the term, and the instance renews from it automatically. There is no deposit fee and no card, bank or payment processor anywhere in the flow.
Do I need to verify my identity to rent an RTX 5090?
No. A GPU Rent Hub account is a username and a password. No email address, phone number or identity document is requested at sign-up or afterwards, and none is stored. A recovery email, PGP key or TOTP secret can be added later if you want one.
Is the RTX 5090 dedicated or shared?
Dedicated. One tenant per physical server, with the whole card, the whole PCIe lane and root access. There is no vGPU layer, no time-slicing and no other customer on the machine. The interconnect is PCIe Gen5 peer-to-peer.
What can an RTX 5090 run?
32 GB of GDDR7 at 1.79 TB/s, faster memory than an A100 for a fifth of the price. GPU Rent Hub ships templates for PyTorch, vLLM, ComfyUI, Ollama, JAX and bare Ubuntu, and you can boot any image you like instead.
Where are the RTX 5090 servers located?
In three United States datacenters: DFW-1 in Dallas, Texas, IAD-1 in Ashburn, Virginia, and PDX-1 in Hillsboro, Oregon. The price is identical in all three, so pick the one closest to your users. Current stock per region is listed above.

Deploy an RTX 5090 in about ninety seconds

Pick a username, fund the balance in any of seven coins, choose the region and the term. No email address, no identity check, no card.

Create an account →