Rent a GPU for AI from €0.12 an hour
For most AI work the deciding number is video memory. 12–16 GB covers SDXL, Flux in fp8, Whisper and small chat models; 24 GB adds Qwen-Image, Wan 2.2 video and 32B models in 4-bit; 48–80 GB is for 70B models, bigger video and training. VRAMfarm rents 20 NVIDIA cards from €0.12 an hour, billed by the minute.
Cards for AI work and their prices
| GPU | Type | Memory | TFLOPS FP32 | Where | Price per hour | ≈ per month | Action |
|---|---|---|---|---|---|---|---|
| RTX 3060 | consumer | 12 GB | ≈12.7 | +10 | €0.12/h | €86 per month | Rent |
| RTX 3090 | consumer | 24 GB | ≈35.6 | +15 | €0.30/h | €216 per month | Rent |
| RTX 4090 | consumer | 24 GB | ≈82.6 | +16 | €0.59/h | €425 per month | Rent |
| RTX 5090 | consumer | 32 GB | ≈104.8 | +15 | €0.76/h | €547 per month | Rent |
| RTX A6000 | workstation | 48 GB | ≈38.7 | €0.72/h | €518 per month | Rent | |
| L40S | data centre | 48 GB | ≈91.6 | +1 | €1.06/h | €763 per month | Rent |
| A100 | data centre | 80 GB | ≈19.5 | +2 | €1.46/h | €1,051 per month | Rent |
| H100 | data centre | 80 GB | ≈67 | +1 | €3.90/h | €2,808 per month | Rent |
| H200 | data centre | 141 GB | ≈67 | +2 | €5.43/h | €3,910 per month | Rent |
To launch, your balance must cover one hour of the machine. You are billed by the minute for actual use.
What you can do on a rented GPU
Images
ComfyUI with Flux, SDXL, Qwen-Image and HiDream opens in your browser, with the models already downloaded. Bring your own LoRAs and workflows.
Video
Wan 2.2 makes 5-second clips from text or a picture. The 5B model runs on 24 GB cards, the 14B model on 48 GB and up.
Chat and LLMs
A private chat on Qwen3, or your own API with vLLM. Data stays on the server.
Voice and speech
Text to speech with Qwen3-TTS, VoxCPM and Chatterbox (a voice from a few seconds of audio), transcription with Whisper.
Start from the task, then the card
An RTX 3090 and an RTX 4090 both have 24 GB: the same models fit, the 4090 is faster and costs more. Decide what has to fit first, then how fast you need it. The picker on the home page shows the cheapest card for a given model.
Three ways to work
Ready-made apps for people who want results without setup. Jupyter and SSH on a clean machine with CUDA 12 for people who write code. And an API to start and stop servers from your scripts. All three use the same cards and the same per-minute price.
Keep your files between sessions
The server disk is wiped when you stop. Save /workspace on stop, and models, LoRAs and outputs come back on the next server, even on a different card, for €0.03 per GB a month.
What a typical evening costs
Three hours of image generation on an RTX 4090 cost about €1.77. Testing a 70B model on an H100 for an hour: about €3.90. Stop the server and billing stops the same minute.
Ready-made apps for this
The app installs itself with its models and opens in your browser. You pay only for the GPU time.
Guides on this
11 November 2026 How much VRAM do I need? A table for LLMs, Flux, Wan 2.2 video and training
VRAM = weights × bytes per parameter + KV cache + overhead. A 70B model in 4-bit needs about 43 GB of weights. Tables by model, quantisation and card.
11 November 2026 Cost per image, video clip and million tokens on a rented GPU, worked out
1,000 Flux images, one 5-second Wan 2.2 clip, a million tokens from 8B to 70B, a Flux LoRA: all-in costs with live euro prices and the maths shown.
11 November 2026 Google Colab vs a rented GPU: disconnects, compute units and the real cost per hour
Colab is free or cheap until sessions drop after a few hours and units run out. When Colab is enough, and when a GPU from €0.30 an hour is better.
Frequently asked questions
Which GPU should I rent for AI as a beginner?
An RTX 4090 or RTX 3090 with 24 GB. Almost every popular open model for images, video and chat runs on 24 GB, and they are among the cheapest cards per gigabyte.
Do I need to install drivers or CUDA?
No. Every machine comes with NVIDIA drivers and CUDA 12, and the driver and CUDA version are shown before you launch. Ready-made apps install themselves.
Can I use my own Docker image?
Yes. Choose “Custom Docker image” and give the image name. A server with a custom image cannot save its folder on stop. Docker does not run inside the server, because the server is itself a container.
How do I pay?
Top up your balance from €10 by card (Apple Pay and Google Pay included) or SEPA. Billing is per minute from the balance in euros.
Pick a card and start
Sign up, top up from €10 and launch a server or a ready-made app in a few minutes.