RentGPU RentGPU
GPU finden →
Methodology

How we collect prices.

Every price on RentGPU is a real listing pulled from a public provider API or pricing page on a recurring schedule. We don't use estimates, third-party aggregators, or vendor-supplied figures.

Data flow.

  1. Scrape. Per-provider adapters (Vast.ai, RunPod, TensorDock, Lambda, io.net, Fluence, Clore.ai, Akash) hit each provider's listings endpoint at minimum every hour.
  2. Aggregate. For each (GPU model, provider) pair we record median, min, max hourly price plus the offer count — so a single low-ball outlier doesn't distort the headline number.
  3. Normalize. Provider-specific GPU names ("H100 SXM", "H100 PCIe", "H100 NVL") resolve through an alias table to one canonical model.
  4. Display. The hourly rate shown is the median across active offers within the last 24 hours.

Affiliate disclosure.

Some "Launch" / "Sign up" links on RentGPU are affiliate links. If you sign up through one, we may receive a referral fee. This never affects ranking — providers are sorted by lowest median price, full stop. We don't accept payment to feature, hide, or reorder providers.

Caveats.

  • Spot/bid prices on P2P markets fluctuate minute-by-minute — the median we show smooths that volatility but won't match the exact price you see on a provider's checkout.
  • Some providers (Akash, Render Network) settle in tokens. We convert to USD at the most recent token-USD rate; volatility there is on you, not us.
  • First-party clouds (Lambda, AWS, hyperscalers) publish on-demand list prices; spot or commitment discounts can be substantially cheaper.