Methodology
How GPU Cloud Prices gathers and normalizes the prices you see.
Where the data comes from
Prices are collected automatically from providers' public pricing APIs — for example RunPod's GraphQL endpoint and Vast.ai's marketplace bundles endpoint. We do not scrape behind logins or paywalls. Each price links back to the provider so you can verify it yourself.
How prices are normalized
- Per GPU, per hour, in USD. Where a provider quotes a multi-GPU machine, we divide by the GPU count so every number is a single-GPU hourly rate and directly comparable.
- GPU models are canonicalized (e.g. "NVIDIA H100 80GB SXM5" → model H100, variant SXM 80GB) so the same chip from different vendors lines up.
- Price types are labeled — secure/dedicated capacity vs community/marketplace capacity, which is usually cheaper but less reliable.
How often it updates
The dataset is refreshed on a schedule, multiple times per day. The "last updated" timestamp in the footer reflects the most recent successful collection. GPU prices change constantly, especially on marketplaces — always confirm the live price and availability on the provider's own site before renting.
How we make money
Some outbound provider links are referral links, which may earn us a commission at no extra cost to you. This never affects how prices are ranked — listings are always ordered by price, lowest first.