Rent an RTX 5090 by the minute

The RTX 5090 is NVIDIA’s 32 GB Blackwell card. On CloudGPU you rent it by the minute: pick a template or a raw GPU, click Deploy, and in about 2 minutes you have root SSH and JupyterLab on the machine. When the job is done, destroy the instance and the unused minutes go back to your balance.

See the current RTX 5090 rate · Deploy an RTX 5090

The card

RTX 5090 32G
Architecture Blackwell
VRAM 32 GB
Memory bandwidth 1792 GB/s
FP16 209 TFLOPS
CUDA compute capability 12.0
Billing Per minute, 1-minute minimum
Rate On cloudgpu.app/pricing

See current rates on /pricing.

What people run on it

Which model fits depends on its size, quantization and context length; see how much VRAM an LLM needs.

Typical setups:

  • Image generation with ComfyUI. The ComfyUI template gives you ComfyUI and a public HTTPS endpoint; you download FLUX or other weights yourself. See Rent a GPU for ComfyUI.
  • Local LLMs with Ollama. The Ollama template comes empty or with a model preloaded and exposes an OpenAI-compatible endpoint. See choosing a GPU for Ollama.
  • Serving with vLLM. Rent a raw GPU and install vLLM over SSH. See vLLM on a rented GPU.
  • Fine-tuning. The LLaMA-Factory template, or your own stack on a raw GPU. See Rent a GPU for LoRA fine-tuning.

Not sure 32 GB is enough? How much VRAM do I need to run an LLM? and the RTX 4090 vs RTX 5090 vs A100 comparison help you choose. If your model needs more memory on one card, look at the RTX 4090 48G or the A100 80G.

How it works

  1. Sign up with an email address or a Google account.
  2. Add balance. Visa or Mastercard, Alipay (in most countries) or Google Pay / Apple Pay in one checkout, USDT on TRC-20, or bank transfer for business customers. The same balance pays for GPU minutes and API tokens. See payments.
  3. Deploy. On Deploy, choose a template (Ollama, JupyterLab + Ollama, ComfyUI, LLaMA-Factory) or Raw GPU, and the RTX 5090 32G.
  4. Work. About 2 minutes later the instance row shows root SSH and JupyterLab; API templates also get a public HTTPS URL.
  5. Destroy. Copy off what you want to keep, then destroy the instance. Billing stops and unused minutes go back to your balance.

Billing in plain words

  • Billed per minute of GPU time, with a 1-minute minimum.
  • While an instance runs, each started hour is pre-charged from your balance; unused minutes go back to your balance when you destroy the instance.
  • No subscription and no monthly fee; the balance does not expire.
  • Destroying an instance releases its GPU, and persistent storage across restarts is still on the roadmap, so copy your outputs off before you destroy.

For a longer read on per-minute billing and paying without a US card, see the blog post Rent an RTX 5090 in the cloud by the minute.

FAQ

How long does an RTX 5090 instance take to start? About 2 minutes from Deploy to a running machine.

How is the RTX 5090 billed? Per minute of GPU time, with a 1-minute minimum. Each started hour is pre-charged from your prepaid balance, and unused minutes go back to your balance when you destroy the instance. The current rate is on the pricing page.

Do I get root access? Yes. Every instance comes with root SSH and JupyterLab. Templates that serve an API, such as Ollama and ComfyUI, also get a public HTTPS URL.

Is my data kept after I destroy the instance? No. Persistent storage across restarts is on the roadmap, so copy your models and outputs off the machine before you destroy it.

How can I pay? Visa, Mastercard, Alipay, Google Pay or Apple Pay in one checkout, USDT on TRC-20, or bank transfer for business customers.

Ready to start?

Deploy an RTX 5090 · See the current rate

Sources

All read on 24 October 2026 (Beijing time, UTC+8).