CloudGPU

✓ Verified by AITL
November 21, 2026

GPU Rental by the Minute and Model APIs on One Balance

CloudGPU rents GPUs by the minute, from an RTX 2080 to B300 cards, and resells model APIs such as Claude, DeepSeek and GLM through one OpenAI-compatible endpoint, all paid from one prepaid balance.

Category:
AI Coding
Pricing Model:
Paid
Last checked:
November 21, 2026
Reviewed by:
Marian Cojocaru
Pricing Snapshot
Billing Frequency:
Pay as you go
Paid Option From:
Prepaid top-up from $5; GPUs from $0.15/hour
Last Known Major Update:
Not publicly available

Some links on AITL are affiliate links and we may earn a commission at no extra cost to you. It never affects our reviews. How we curate

At a glance
CloudGPU
Homepage screenshot
Pros:
  • GPU rental billed by the minute, from $0.15 an hour
  • One-click templates for ComfyUI, Ollama, LLaMA-Factory and MiniMax H3
  • One OpenAI-compatible endpoint for many chat, image and speech models
  • One prepaid balance for GPUs and API calls, with no expiry
  • Top-ups from $5 by card, Alipay, Google Pay, Apple Pay or USDT
  • MCP server lets AI agents run GPU jobs within an approved budget
  • No egress fees and no subscription
Cons:
  • No free plan or free credits
  • Claude and other models are resold, not the official APIs
  • The website does not name the company behind it
  • Public beta with no uptime guarantee or service credits
  • Unused balance is not refunded in cash
  • Some GPUs run on third-party and peer-to-peer networks
  • Not available in mainland China
API Available

No public API available

About the tool

CloudGPU is a GPU cloud and model API service for developers. You can rent a raw GPU with root SSH and JupyterLab, or deploy a template such as ComfyUI, Ollama, LLaMA-Factory or MiniMax H3 and get an HTTP endpoint without setting anything up. Cards range from an RTX 2080 at $0.15 an hour to RTX 5090, A100, H100, B200 and GB300 machines, some of them hosted on third-party data centres and peer-to-peer GPU networks. Billing is per minute or per hour, and the quote shows the minimum charge before you confirm. The AI API is one OpenAI-compatible endpoint for chat, image, speech and transcription models, including Claude, DeepSeek, GLM, Kimi and MiniMax. CloudGPU lists its Claude rates at about half the official list price and states that it is not the official Claude API. A remote MCP server and a local package let Cursor, Claude Code and other AI agents check a task, show a quote and run GPU jobs within a budget the user approves. Everything is paid from one prepaid US dollar balance, topped up from $5 by card, Alipay, Google Pay, Apple Pay, USDT on TRC-20 or bank transfer. The balance does not expire, but unused credit is not refunded in cash. The service is in public beta with no uptime guarantee, is not offered in mainland China, and the website does not name the company behind it.

Key use cases

Building AI-native applications

Create your own AI products, from copilots and assistants to intelligent workflows and dynamic interfaces.

API & developer readiness

CloudGPU's AI API is one OpenAI-compatible endpoint at https://cloudgpu.app/v1 for chat, image, speech and transcription models such as Claude, DeepSeek, GLM and Kimi, billed per token from the shared balance. It works with the OpenAI SDK, LangChain, Cursor and Cline by changing the base URL and key. A REST API and a remote MCP server with OAuth also let agents rent GPUs and run jobs within an approved budget.

Pricing and plans

  • Pay as you go: top-ups from $5. One prepaid balance for GPU rental and model APIs. No subscription, no free plan.
  • GPU rental: from $0.15/hour. For example RTX 2080 $0.15, RTX 4090 $0.39, RTX 5090 $0.69, A100 80G $1.59, H100 80G $6.81, B200 192G $7.69 and GB300 288G $18.45 an hour, billed per minute or per hour.
  • Model API: per token. For example Claude Opus 5.5 at $2 input and $10 output per million tokens, Claude Sonnet 5.5 at $1 and $5, GLM 5.3 Flash at $0.135 and $0.45.

Per-minute rentals refund unused prepaid minutes when the instance is destroyed; whole-hour options charge each started hour in full. The balance does not expire and is not paid back in cash, except for faults on CloudGPU's side or where the law requires it. No egress fees.

November 21, 2026
See alternatives

Frequently asked questions

Is CloudGPU free?

No. CloudGPU has no free plan or free credits. You top up a prepaid balance from $5 by card, Alipay, Google Pay, Apple Pay or USDT, and GPU minutes and API tokens are deducted from it. The balance does not expire and there is no subscription. Checked November 21, 2026 (payments).

How much does CloudGPU cost?

GPU rental runs from $0.15 an hour for an RTX 2080 to $0.69 for an RTX 5090, $1.59 for an A100 80G and $18.45 for a GB300. Model API calls are billed per token; Claude Opus 5.5, for example, is $2 input and $10 output per million tokens. Checked November 21, 2026 (pricing).

Can I get a refund?

Unused balance is not paid back in cash, including when you close the account. Refunds apply to charges caused by a fault on CloudGPU's side or where the law requires one, and cover the amount paid, not bonus credit. Per-minute rentals return unused prepaid minutes when you destroy the instance (Refund Policy, updated October 24, 2026).

Is my data used to train AI?

No. The privacy policy says CloudGPU does not use your data to train AI models. Data inside a GPU instance is deleted by the supplier when the instance is destroyed. API prompts are forwarded to the model vendor that serves the model, and that vendor's own terms decide how long it keeps them (Privacy Policy, updated October 24, 2026).

Who owns what I create on CloudGPU?

You do. The Terms say you keep ownership of any data, models, code or output you create on or upload to a CloudGPU instance, and CloudGPU gets only the technical permissions needed to run the service (Terms of Service, updated November 19, 2026).

Does CloudGPU have an API?

Yes. Its AI API is one OpenAI-compatible endpoint at https://cloudgpu.app/v1 for chat, image, speech and transcription models, including Claude, DeepSeek, GLM and Kimi. The site says it is not the official Claude API. A remote MCP server lets AI agents rent GPUs within a budget you approve (API).

Who makes CloudGPU?

The website does not name the company behind it. Support is at support@cloudgpu.app. The service is in public beta with no uptime guarantee, and it is not offered to people in mainland China (Terms of Service, updated November 19, 2026).

Automation integrations

Native connectors with Zapier, Make and Albato, checked by AITL.

No direct integrations listed

No public direct integrations currently listed with Zapier, Make, or Albato.

Official channels

Where this tool is active and shares its updates.

Similar tools

Submit your product!

Submit your AI tool for review. The AITL team evaluates every submission by hand. Add the AITL badge to your site and send us the link to get a reply within 48 hours.

Submit now

Side by side

Facts from each tool's AITL page. AITL does not pick a winner.

Open on the Compare page

Listed on AITL? Get your badge

Show visitors that your tool is listed on AITL. Copy the code below and paste it into your website's HTML. The badge links straight to this page.

Listed on AITL badge preview
  1. 1Copy the badge code
  2. 2Paste it into your site, for example the footer
  3. 3Visitors click through to your AITL listing
Free weekly email

One AI workflow a week

Every week we combine 2–4 tools from AITL into a step-by-step workflow you can try in minutes.

  • Clear steps, from the first tool to the result
  • Only tools we reviewed and listed on AITL
  • Plus the newest tools we listed that week

No spam. Unsubscribe anytime. See our Privacy Policy

You're in. Watch your inbox for your first workflow.
Oops! Something went wrong while submitting the form.
Example issue
4 tools · about 20 min

Turn one meeting into a week of social posts

Fathom
Claude
Leonardo AI
SocialBee
  1. Fathom records the call and writes the summary.
  2. Claude turns it into 5 short posts.
  3. Leonardo AI makes one image per post.
  4. SocialBee schedules them across the week.

CloudGPU

Prepaid top-up from $5; GPUs from $0.15/hour

Visit Website