About GPUPerHour

Who runs GPUPerHour

GPUPerHour (https://gpuperhour.com) is operated by QuantaCloud, a GPU capacity broker. QuantaCloud brokers reserved and cluster capacity (16+ GPUs) across multiple data-center partners and returns custom quotes, typically within 24 hours; that service is offered on this site through the quote forms and the GPU cluster page. GPUPerHour itself is free to use.

The site is supported by affiliate commissions and ads. When you follow a "Deploy" link to a provider and rent a GPU, we may earn a referral fee at no extra cost to you. Affiliate relationships do not influence which providers are shown, their prices or how they are ranked; see the Affiliate Disclosure.

What the site does

GPUPerHour is a real-time price comparison for renting cloud GPUs. It tracks on-demand offers for GPUs such as the H100, H200, B200, A100, L40S and RTX 4090 across more than 25 cloud GPU providers and shows, for each GPU, the cheapest current price per GPU-hour, which provider offers it and how many providers have it in stock. Every price links to the provider so you can deploy directly.

• The homepage is the live pricing table with filters by GPU, provider, region and pricing type, plus a market snapshot of the cheapest current price for popular GPUs.
• GPU pages list every current offer for one GPU, its price range, last-seen catalog listings and a 30-day price history.
• Provider pages show one provider's full GPU catalog and availability.
• Comparison pages put two GPUs or two providers side by side: specifications, live prices and when to choose which.

How prices are collected

Our scrapers read each provider's public pricing and availability once a minute. Every observation is stored with the provider, GPU model, GPU count, region, pricing type, security tier and the time it was seen. Prices are normalised to US dollars per GPU-hour: when a provider lists a multi-GPU instance, the instance price is divided by the number of GPUs.

Headline prices (the price in a page title, the "from" price on GPU pages, the market snapshot and the structured data) are computed only from current offers. An offer is current when all of the following hold:

• the provider reports it as available (in stock) right now;
• the provider has live stock tracking, so "available" is a real-time observation rather than a catalog entry;
• it was seen within the last 15 minutes;
• it is secure on-demand pricing: on-demand billing on provider-operated (data-center) hardware;
• it has a valid per-GPU price.

Offers that fail any of these tests can still appear further down a page as last-seen catalog listings, and are always labelled as such. A GPU with no current offer never shows a "from" price in its title.

The same current-offer rule feeds the Cloud GPU Price Index, a daily snapshot of on-demand prices for a fixed GPU basket with free CSV downloads.

What is excluded from headline prices

• Peer-to-peer and marketplace hosts. Community-hosted machines offered through a marketplace are not secure on-demand capacity, so they never set a headline price. Provider pages do show them, so a provider's full catalog is visible there.
• Spot, interruptible and reserved pricing. Headline prices are on-demand only. Spot and reserved rates are shown separately when you filter for them, and a GPU that is only listed as spot or reserved is shown as a catalog listing, not as a live price.
• Stale observations. Anything not seen within 15 minutes drops out of the current set until it is observed again.
• Sub-cent artifacts. Listings that would round to $0.00/hr are skipped for page titles.

Prices are collected from public provider pages and APIs and may lag a provider's own checkout by a few minutes; confirm the final price and capacity with the provider before booking. See the Terms of Service for the full disclaimer.

How often pages update

Contact

Spotted a wrong price, want a provider added, or have a partnership question? Use the contact page or email us directly:

• General enquiries: [email protected]
• Technical support and data corrections: [email protected]
• On X: @gpuperhour

Privacy questions are covered in the Privacy Policy.