Can I rent an RTX 5090 by the hour?

Yes. CloudGPU rents RTX 5090 GPUs with 32 GB of VRAM on demand and bills by the minute. Check the live rental page for availability.

The name on the live offer is the card you start. Deploy shows that name. RTX 5090 and RTX 5090D are different listings when both are offered. The RTX 5090 page describes the 32 GB RTX 5090 offer, and RTX 5090 vs RTX 5090D explains how those two cards differ. This page does not say which listing is available.

How an hour of work is billed

The pricing page states the billing rule in these words:

Per minute, rounded up, with a 1-minute minimum. One hour is pre-authorised while an instance runs, and unused minutes go back to your balance when you destroy the instance. There is no subscription and no minimum term.

An hour of GPU work is billed under that rule. The longer page that uses the same rule is per-minute billing. Current rates stay on the pricing page. This page does not quote them.

Top up the prepaid balance from payments before you deploy. If PyTorch reports no kernel image on this card, see how to fix “CUDA error: no kernel image is available” on an RTX 5090.

Frequently asked questions

Can I run a one-hour job?

Yes. The pricing page states: Per minute, rounded up, with a 1-minute minimum. One hour is pre-authorised while an instance runs, and unused minutes go back to your balance when you destroy the instance. There is no subscription and no minimum term.

Where do I confirm the exact GPU?

On the live offer at https://cloudgpu.app/deploy. Read the card name shown there before you start it. The RTX 5090 page at https://cloudgpu.app/rent-rtx-5090 describes the 32 GB RTX 5090 offer. RTX 5090 and RTX 5090D are different listings when both are offered. This page does not say which listing is available.

How do I stop billing?

Destroy the instance. The pricing page states: unused minutes go back to your balance when you destroy the instance.

Sources (read 2026-11-11)