---
title: "How to run ComfyUI on a cloud GPU: setup, models, cost per image"
description: "Deploy ComfyUI on a rented RTX 4090 or 5090 in 30 seconds, keep checkpoints on a volume, run workflows headless through the API, and know what each image costs."
url: https://powergpu.ai/guides/run-comfyui-on-a-cloud-gpu
last_modified: 2026-09-14T11:04:20+00:00
prices_as_of: 2026-09-14
site: PowerGPU (powergpu.ai)
---

Guide · Hands-on walkthrough

# How to run ComfyUI on a cloud GPU: setup, models, cost per image

Deploy ComfyUI on a rented RTX 4090 or 5090 in 30 seconds, keep checkpoints on a volume, run workflows headless through the API, and know what each image costs.

9 min read Published 2026-09-03 Updated 2026-09-03 (prices live from the sheet)

TL;DR

- Rent a 24 GB RTX 4090 at $0.327 per hour on PowerGPU: it runs SDXL, Flux dev FP8 and most video nodes.
- Rule of thumb: SDXL wants about 8 GB, Flux dev FP8 about 17 GB, video models 24 GB with offloading and 48 GB without.
- An SDXL image takes about 1.1 seconds and a Flux dev image 2.2 seconds on a 4090, a fraction of a cent each.
- Put checkpoints and custom_nodes on a volume so they survive between sessions; re-downloading models is the main waste of GPU-hours.

## Why rent instead of buying a 4090

ComfyUI is bursty by nature: an evening of prompting, a batch of a thousand images, then nothing for a week. A rented card bills only the seconds the queue is running — $0.327/hr for an [RTX 4090](https://powergpu.ai/gpu/rtx-4090) on-demand, $0.163/hr interruptible — so a heavy month rarely reaches a fraction of the card's retail price, and you can jump to a 32 GB or 96 GB card the day a workflow needs it. No driver upgrades, no 450 W heater under the desk.

## Pick the card: 24, 32 or 48 GB

| Card | VRAM | On-demand | Interruptible | Best for |
| --- | --- | --- | --- | --- |
| [RTX 3090](https://powergpu.ai/gpu/rtx-3090) | 24 GB | $0.108 | $0.054 | SDXL batch farms, SD 1.5, LoRA training on a budget |
| [RTX 4090](https://powergpu.ai/gpu/rtx-4090) | 24 GB | $0.327 | $0.163 | The default: Flux dev FP8, SDXL + ControlNet stacks, most video nodes |
| [RTX 5090](https://powergpu.ai/gpu/rtx-5090) | 32 GB | $0.439 | $0.219 | Full-precision Flux, Wan/Hunyuan video, heavy upscale chains |
| [L40S](https://powergpu.ai/gpu/l40s) | 48 GB | $0.514 | $0.257 | Serving images to users 24/7, several workflows loaded at once |
| [RTX PRO 6000 WS](https://powergpu.ai/gpu/rtx-pro-6000-ws) | 96 GB | $1.040 | $0.520 | Video diffusion without offloading, giant batches |

Rule of thumb: Flux dev in FP8 wants ~17 GB, SDXL ~8 GB, current video models 24 GB with offloading and 48 GB without. The [cheapest-GPU guide](https://powergpu.ai/guides/cheapest-gpu-for-stable-diffusion) ranks the cards by images per dollar.

## Deploy the template (30 seconds)

1. Open the [console](https://cloud.powergpu.ai/?gpu=rtx-4090), filter on the card, pick a machine.
2. Choose [ComfyUI](https://powergpu.ai/templates/comfyui) in the template picker. It ships ComfyUI-Manager, exposes the UI on a TLS-terminated port and mounts /workspace/ComfyUI/models.
3. Set the disk (60 GB is plenty for the OS and cache) and attach a volume for models — next section.
4. Deploy. The instance is running in about 30 seconds; click the port link and the node graph is there.

*same thing from the CLI*

```
powergpu launch --gpu rtx-4090 --template comfyui \
    --disk 60 --volume sdmodels:/workspace/ComfyUI/models
# ✓ instance i-3c91ab04 running (28.6s)
# ✓ https://i-3c91ab04.powergpu.ai:8188  (ComfyUI)
```

## Models on a volume, once

Checkpoints are 2–12 GB each and downloading them every session is the most common way to waste GPU-hours. Create a [volume](https://powergpu.ai/products/volumes) the first time, mount it on the models folder, fill it once — Hugging Face and Civitai downloads run at multi-Gbps from the datacenter — and every future instance in the region starts with the library present. A 120 GB library costs $9.60/month. Put custom_nodes on the same volume so installed nodes persist too.

## Headless: workflows as an API

Enable dev mode in ComfyUI settings, then "Save (API format)" on any workflow. The JSON you get is a request body: POST it to /prompt, poll /history/<id>, fetch the images from /view. That loop is how batch pipelines run a thousand prompts overnight on [interruptible](https://powergpu.ai/products/interruptible) capacity, and how a [serverless endpoint](https://powergpu.ai/products/serverless) serves the same workflow behind autoscaling workers.

*queue a workflow, read the result*

```
curl -X POST https://i-3c91ab04.powergpu.ai:8188/prompt \
  -H 'Content-Type: application/json' \
  -d @workflow_api.json
# {"prompt_id": "a1c2…", "number": 12}
curl https://i-3c91ab04.powergpu.ai:8188/history/a1c2…
```

## What an image costs, honestly

| Job | Card · mode | Time | Cost |
| --- | --- | --- | --- |
| One SDXL image, 1024², 25 steps | RTX 4090 on-demand | ~1.1 s | $0.00010 |
| One Flux dev image, 1024², 20 steps | RTX 4090 on-demand | ~2.2 s | $0.00020 |
| An evening of prompting | RTX 4090 on-demand | 3 h | $0.98 |
| 10,000 Flux images, batch | RTX 4090 interruptible | ~6.1 h | $1.00 |
| Model library, always warm | 120 GB volume | 1 month | $9.60 |

Roughly **5,004 Flux images per dollar** on-demand, 10,008 interruptible. Speeds are typical for a tuned 4090; your workflow, resolution and step count move them.

## Five traps that waste GPU-hours

- **Downloading models every session** — the volume above fixes it permanently.
- **Leaving the instance running overnight** — stop it; per-second billing means an idle GPU is pure waste. The disk survives stop/start.
- **Running batches on-demand** — queues are interruptible by nature: each prompt is a restartable item. Pay half.
- **Picking a card by hourly price alone** — a 5090 finishes Flux batches faster than its price gap over a 4090; images per dollar is the metric.
- **Forgetting the outputs** — write results to the volume or download them before destroy; the instance disk dies with the instance.

---

Put the numbers to work

Every price in this guide is our live rate — fixed, ≥30% under the market median, billed per second. Deploy the exact setup above from the [console](https://cloud.powergpu.ai/) in about 30 seconds, paid in crypto, no card and no KYC.

---

*About PowerGPU:* PowerGPU (powergpu.ai) is a cloud GPU rental service offering 80 NVIDIA GPU models — from the RTX A2000 at $0.024/hr to the B300 — at fixed prices set at least 30% below the public GPU marketplace median and re-checked weekly (H100 SXM: $1.428/hr on-demand). Billing is per second with no minimums; payment is crypto only (USDT, BTC, XMR, ETH, SOL, LTC, TRX) with no KYC. Instances run in Tier-III datacenters across 32 regions with a 99.9% uptime SLA and deploy in about 30 seconds from the web console (cloud.powergpu.ai) or the REST API.

Source: https://powergpu.ai/guides/run-comfyui-on-a-cloud-gpu · Site index for AI assistants: https://powergpu.ai/llms.txt · Full content: https://powergpu.ai/llms-full.txt