GPU rental prices: fifteen cards, three US datacenters.
GPU Rent Hub rents dedicated GPUs by the month at flat prices from $190 to $3,890 per card, or by the week from $55, paid in Bitcoin, Ethereum, USDT, USDC, Monero, Solana or Litecoin, with no identity verification. Every row below is a whole physical server: the GPU, the CPU cores, the RAM and the NVMe listed are yours alone. Prices include power, cooling, a public IPv4, unmetered inbound and 20 TB/month outbound. Stock is live across DFW-1, IAD-1 and PDX-1.
Refreshed from inventory
every 60 s · ■ live
| GPU | Architecture | VRAM | vCPU | RAM | NVMe | Semaine | Mois | DFW-1 | IAD-1 | PDX-1 | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| B200 NouveauNVIDIA B200 SXM | Blackwell Datacenter | 180 GB HBM3e | 28 | 283 GB | 3 TB | 1 090 $US/sem | 3 890 $US/mois | 6 | — | 2 | Voir |
| H200 SXMNVIDIA H200 SXM 141GB | Hopper Datacenter | 141 GB HBM3e | 24 | 251 GB | 2 TB | 640 $US/sem | 2 290 $US/mois | 11 | 8 | — | Voir |
| H100 SXMNVIDIA H100 SXM 80GB | Hopper Datacenter | 80 GB HBM3 | 20 | 251 GB | 2 TB | 475 $US/sem | 1 690 $US/mois | 32 | 19 | 12 | Voir |
| H100 PCIeNVIDIA H100 PCIe 80GB | Hopper Datacenter | 80 GB HBM2e | 16 | 188 GB | 1 TB | 420 $US/sem | 1 490 $US/mois | 14 | 9 | 4 | Voir |
| MI300XAMD Instinct MI300X 192GB | CDNA 3 Datacenter | 192 GB HBM3 | 24 | 283 GB | 2 TB | 450 $US/sem | 1 590 $US/mois | 8 | — | — | Voir |
| A100 SXMNVIDIA A100 SXM 80GB | Ampere Datacenter | 80 GB HBM2e | 16 | 125 GB | 1 TB | 280 $US/sem | 990 $US/mois | 40 | 22 | 16 | Voir |
| L40SNVIDIA L40S 48GB | Ada Lovelace Datacenter | 48 GB GDDR6 ECC | 16 | 94 GB | 500 GB | 148 $US/sem | 520 $US/mois | 26 | 18 | 30 | Voir |
| RTX 6000 AdaNVIDIA RTX 6000 Ada 48GB | Ada Lovelace Station de travail | 48 GB GDDR6 ECC | 12 | 62 GB | 500 GB | 130 $US/sem | 460 $US/mois | 9 | 12 | 5 | Voir |
| RTX 5090NVIDIA GeForce RTX 5090 32GB | Blackwell Grand public | 32 GB GDDR7 | 12 | 62 GB | 500 GB | 155 $US/sem | 540 $US/mois | 48 | 30 | 22 | Voir |
| RTX 4090NVIDIA GeForce RTX 4090 24GB | Ada Lovelace Grand public | 24 GB GDDR6X | 8 | 47 GB | 250 GB | 110 $US/sem | 390 $US/mois | 120 | 84 | 61 | Voir |
| RTX A6000NVIDIA RTX A6000 48GB | Ampere Station de travail | 48 GB GDDR6 ECC | 8 | 47 GB | 500 GB | 82 $US/sem | 290 $US/mois | 34 | 20 | 11 | Voir |
| A40NVIDIA A40 48GB | Ampere Datacenter | 48 GB GDDR6 ECC | 8 | 47 GB | 500 GB | 68 $US/sem | 240 $US/mois | 22 | 15 | — | Voir |
| L4NVIDIA L4 24GB | Ada Lovelace Datacenter | 24 GB GDDR6 | 8 | 31 GB | 250 GB | 70 $US/sem | 250 $US/mois | 40 | 28 | 36 | Voir |
| RTX 3090NVIDIA GeForce RTX 3090 24GB | Ampere Grand public | 24 GB GDDR6X | 8 | 31 GB | 250 GB | 65 $US/sem | 230 $US/mois | 96 | 52 | 38 | Voir |
| RTX A5000NVIDIA RTX A5000 24GB | Ampere Station de travail | 24 GB GDDR6 ECC | 8 | 31 GB | 250 GB | 55 $US/sem | 190 $US/mois | 28 | 14 | 9 | Voir |
Each card has its own page with the weekly price, the equivalent cost per GPU-hour, full specs and stock in DFW-1, IAD-1 and PDX-1: rent a B200 $3,890/mo · rent an H200 SXM $2,290/mo · rent an H100 SXM $1,690/mo · rent an MI300X $1,590/mo · rent an H100 PCIe $1,490/mo · rent an A100 SXM $990/mo · rent an RTX 5090 $540/mo · rent an L40S $520/mo · rent an RTX 6000 Ada $460/mo · rent an RTX 4090 $390/mo · rent an RTX A6000 $290/mo · rent an L4 $250/mo · rent an A40 $240/mo · rent an RTX 3090 $230/mo · rent an RTX A5000 $190/mo.
What does a whole 8-GPU node cost per month?
Multi-GPU configurations are always on a single machine. SXM parts get NVLink and, on 8-way nodes, NVSwitch; PCIe parts share a PCIe Gen4/Gen5 root with peer-to-peer enabled. Discounts stack with term discounts. If you own the cards already, the same HGX platforms take customer hardware under GPU colocation, from $240 per GPU per month.
| Node | Interconnect | Total VRAM | Weekly | Monthly | Per GPU-hour |
|---|---|---|---|---|---|
| 8× B200NVIDIA HGX B200 | NVSwitch · 1.8 TB/s | 1440 GB | 7 850 $US/wk | 28 010 $US/mo | 4,86 $US |
| 8× H200 SXMNVIDIA HGX H200 | NVSwitch · 900 GB/s | 1128 GB | 4 610 $US/wk | 16 490 $US/mo | 2,86 $US |
| 8× H100 SXMNVIDIA HGX H100 | NVSwitch · 900 GB/s | 640 GB | 3 420 $US/wk | 12 170 $US/mo | 2,11 $US |
| 4× H100 SXMNVIDIA HGX H100 | NVLink · 900 GB/s | 320 GB | 1 750 $US/wk | 6 220 $US/mo | 2,16 $US |
| 8× MI300XAMD Instinct platform | Infinity Fabric · 896 GB/s | 1536 GB | 3 240 $US/wk | 11 450 $US/mo | 1,99 $US |
| 8× A100 SXM 80GNVIDIA HGX A100 | NVSwitch · 600 GB/s | 640 GB | 2 020 $US/wk | 7 130 $US/mo | 1,24 $US |
| 8× L40SPCIe Gen4 | PCIe peer-to-peer | 384 GB | 1 070 $US/wk | 3 740 $US/mo | 0,65 $US |
| 8× RTX 5090PCIe Gen5 | PCIe peer-to-peer | 256 GB | 1 120 $US/wk | 3 890 $US/mo | 0,68 $US |
| 8× RTX 4090PCIe Gen4 | PCIe peer-to-peer | 192 GB | 790 $US/wk | 2 810 $US/mo | 0,49 $US |
8-way nodes include a dedicated 100G pair and a second NVMe pool. Larger clusters (16-64 GPUs, InfiniBand HDR/NDR) are quoted per case. Ask for a cluster quote
What is included in the rental price?
Every price on this page is flat and paid up front from your balance in BTC, ETH, USDT, USDC, XMR, SOL or LTC. No card on file, no wire, no name on the invoice and no identity check: see what no-KYC GPU hosting means, how crypto settlement works, or where the break-even sits against hourly GPU rental.
Compute
Dedicated GPU(s) on a single-tenant server. AMD EPYC or Intel Sapphire Rapids hosts, ECC RAM, no oversubscription of anything.
Storage
Local NVMe as listed, encrypted at rest with a per-instance key we destroy on release. Add more NVMe or a network volume from the dashboard.
Network
One public IPv4, a /64 of IPv6, 100 Gbps port (10 Gbps guaranteed per single-GPU instance), unmetered in, 20 TB out. Extra IPs and bandwidth are add-ons.
Access
Root over SSH, IPMI-style console from the dashboard, reboot and reinstall on demand, any OS image. Templates for PyTorch, vLLM, ComfyUI, Ollama, JAX and bare Ubuntu/Debian.
Can't see the card you need?
We add hardware based on the waitlist. Join it from the dashboard for any sold-out configuration and you'll be first in line. No deposit required to queue.
Which GPU should you rent?
Every pairing below is a full side-by-side: memory, bandwidth, FP8 support, interconnect, monthly price and the equivalent per GPU-hour, with a verdict derived from the numbers rather than from opinion.