GPU servers, billed by the minute
GPU rental on VRAMfarm starts at €0.12 an hour. Pick a card, from an RTX 3060 for experiments to an H100 or B200 for training large models, top up your balance in euros, and you can usually connect over SSH or Jupyter within 2–5 minutes.
Consumer GPUs
The best price-to-speed ratio for image generation and models up to 32B.
| GPU | Type | Memory | TFLOPS FP32 | Where | Price per hour | ≈ per month | Action |
|---|---|---|---|---|---|---|---|
| RTX 5090 | consumer | 32 GB | ≈104.8 | +15 | €0.76/h | €547 per month | Rent |
| RTX 5080 | consumer | 16 GB | ≈56.3 | +7 | €0.33/h | €238 per month | Rent |
| RTX 5070 Ti | consumer | 16 GB | ≈43.9 | +3 | €0.24/h | €173 per month | Rent |
| RTX 4090 | consumer | 24 GB | ≈82.6 | +16 | €0.59/h | €425 per month | Rent |
| RTX 4080 | consumer | 16 GB | ≈48.7 | +7 | €0.31/h | €223 per month | Rent |
| RTX 4060 Ti | consumer | 16 GB | ≈22.1 | +5 | €0.16/h | €115 per month | Rent |
| RTX 3090 | consumer | 24 GB | ≈35.6 | +15 | €0.30/h | €216 per month | Rent |
| RTX 3080 | consumer | 10 GB | ≈29.8 | +6 | €0.17/h | €122 per month | Rent |
| RTX 3060 | consumer | 12 GB | ≈12.7 | +10 | €0.12/h | €86 per month | Rent |
Professional GPUs
More memory per card: 24–96 GB for serving large models and rendering.
| GPU | Type | Memory | TFLOPS FP32 | Where | Price per hour | ≈ per month | Action |
|---|---|---|---|---|---|---|---|
| RTX PRO 6000 | workstation | 96 GB | ≈125 | +8 | €2.12/h | €1,526 per month | Rent |
| RTX A6000 | workstation | 48 GB | ≈38.7 | €0.72/h | €518 per month | Rent | |
| RTX A5000 | workstation | 24 GB | ≈27.8 | €0.33/h | €238 per month | Rent | |
| RTX A4000 | workstation | 16 GB | ≈19.2 | +1 | €0.18/h | €130 per month | Rent |
Data-centre GPUs for training
A100, H100, H200 and B200 with HBM memory for training and fast LLM inference.
| GPU | Type | Memory | TFLOPS FP32 | Where | Price per hour | ≈ per month | Action |
|---|---|---|---|---|---|---|---|
| B200 | data centre | 192 GB | ≈80 | €16.12/h | €11,606 per month | Rent | |
| H200 | data centre | 141 GB | ≈67 | +2 | €5.43/h | €3,910 per month | Rent |
| H100 | data centre | 80 GB | ≈67 | +1 | €3.90/h | €2,808 per month | Rent |
| A100 | data centre | 80 GB | ≈19.5 | +2 | €1.46/h | €1,051 per month | Rent |
| L40S | data centre | 48 GB | ≈91.6 | +1 | €1.06/h | €763 per month | Rent |
| L40 | data centre | 48 GB | ≈90.5 | €0.46/h | €331 per month | Rent | |
| L4 | data centre | 24 GB | ≈30.3 | €1.37/h | €986 per month | Rent |
To launch, your balance must cover one hour of the machine. You are billed by the minute for actual use.
Why renting a GPU server beats buying one
Buy or rent
An RTX 4090 costs about €1,800, and a server with H100s costs hundreds of thousands. If you need a GPU for tens or hundreds of hours a month, renting is far cheaper: you don’t pay for idle time, power or maintenance, and when new models come out you simply switch cards.
Where the cards come from
VRAMfarm brings available machines from several data-centre providers into one catalogue. That gives you a wider choice than a cloud with a single data centre, at lower prices: we show the best offers.
Renting a GPU server
What you get
A machine with one or more graphics cards, a CPU, RAM and an NVMe disk, where you work as root inside an isolated container. Open a ready-made app for images, video or chat, take a developer environment (PyTorch with Jupyter, vLLM, an empty ComfyUI) or run your own Docker image. A server is usually ready in 2–5 minutes, or 8–15 minutes with your own image.
Configuration
Choose by video memory, because it decides which model fits. One 12–16 GB card is enough for Stable Diffusion and LLMs up to 8B. 32B models in 4-bit fit in 24 GB, and 70B needs an 80 GB card or several cards in one machine. Large models train and fine-tune fastest on A100 and H100 cards with HBM memory.
Billing
You pay by the minute: if a server ran for 40 minutes, you are charged 40/60 of the hourly price. The minimum charge per launch is 15 minutes. To launch, your balance must cover one hour of the machine; the rest stays on your balance. Billing stops when you stop the server. The disk is wiped when you stop; to keep your models and results, click Save and stop and we keep /workspace for €0.03 per GB per month.
- Top up By card or SEPA in euros. Your balance must cover one hour of the machine. from €10
- Choose what to run A ready-made app, PyTorch with Jupyter, vLLM, ComfyUI or your own Docker image.
- Pick a card by memory From 12 GB for tests to 80 GB and more for 70B models. from €0.12/h
- Connect SSH, Jupyter or the app button in the dashboard. Billed by the minute. 2–5 minutes
| Configuration | Total VRAM | Good for | Price from, per hour |
|---|---|---|---|
| 1× RTX 3060Tests, study projects, SD 1.5 | 12 GB | Tests, study projects, SD 1.5 | €0.12 |
| 1× RTX 4090Stable Diffusion, Flux, LLMs up to 32B in 4-bit | 24 GB | Stable Diffusion, Flux, LLMs up to 32B in 4-bit | €0.59 |
| 2× RTX 4090LLM 32B with long context, parallel generation | 48 GB | LLM 32B with long context, parallel generation | €1.18 |
| 1× RTX A6000LLMs up to 72B in 4-bit, Wan 2.2 14B video, 13B fine-tuning | 48 GB | LLMs up to 72B in 4-bit, Wan 2.2 14B video, 13B fine-tuning | €0.72 |
| 1× A100LLM 70B in 4-bit, fine-tuning and fast inference | 80 GB | LLM 70B in 4-bit, fine-tuning and fast inference | €1.46 |
| 4× A100Training 30–70B models | 320 GB | Training 30–70B models | €5.84 |
| 1× H100Training and inference at full speed | 80 GB | Training and inference at full speed | €3.90 |
| 8× H100LLM training, full fine-tune of 70B | 640 GB | LLM training, full fine-tune of 70B | €31.20 |
The price of a multi-GPU server is a guide: number of cards × the price of one card. The exact cost of a specific machine is shown in the list of available servers before you launch.
Renting a server with a graphics card
When you need one
When your own computer runs out of memory or time: the model doesn’t fit in the 8–12 GB of a home card, training on a laptop takes a day, or rendering an animation ties up your workstation. A rented server takes that load off.
How it differs from a VPS
Unlike a virtual server from a regular host, the graphics card is yours alone and not shared, and the NVIDIA drivers and CUDA are already installed. Connect over SSH or through Jupyter Lab in your browser, and move files with scp, rsync or Git.
Your balance
Your balance is in euros and can be topped up by card or SEPA from €10. Machines you can launch right now are on the available servers page.
- SSH
- Jupyter Lab
- scp and rsync
- Git
- The graphics card is yours alone, not shared
- NVIDIA drivers and CUDA already installed
- Root access in an isolated container
- CPU, RAM and an NVMe disk
GPU rental questions
How much does a GPU server cost?
From €0.12 an hour for entry-level cards, from €0.59 for an RTX 4090 and from €3.90 for an H100. You only pay for the time it runs, by the minute.
Why is hourly GPU rental better than buying a server?
You don’t lock money into hardware that will be outdated in two or three years, and you only pay for the hours you actually use. For up to a few hundred hours a month, renting is almost always cheaper.
Which GPU should I choose for my model?
Video memory matters most. 12–16 GB is enough for Stable Diffusion and models up to 8B, 32B in 4-bit needs 24 GB (RTX 3090, 4090), and 70B in 4-bit needs 80 GB (A100, H100) or two 48 GB cards.
Can I rent several GPUs at once?
Yes, you can run several machines in parallel. Machines with several GPUs are listed separately in the catalogue.
How do I pay for a GPU server?
Your balance is in euros. Top it up by card or SEPA transfer from €10 through a payment provider we will announce before launch. VAT, where it applies, is shown at checkout.
Need an AI app with no setup?
Ready-made apps for images, video and chat install together with their models.