On-demand — guaranteed capacity, runs until you stop it.Spot — cheaper, but interruptible; can be reclaimed with little notice.
GPU
Best on-demand
Spot from
Price range /hr
Market avg
Spread
Change
Trend
All 16 live sources we track — from hyperscaler
baselines to decentralized/crypto marketplaces. Ranges are the reliability-adjusted
price across the GPUs each provider offers, pulled live from the tape.
Some “Rent” links are affiliate links: if you sign up through them we may earn a referral fee, at no cost to you. Prices and rankings on this page are computed the same way regardless — the referral never changes the numbers. Links open each platform's browse page, filtered to the GPU where supported; pick the top (cheapest) row.
Every figure is pulled live from a public API or pricing page
(Vast, RunPod, DataCrunch, Clore, Akash, Nosana, Azure, Lambda, TensorDock, CoreWeave, Nebius, Crusoe, Hyperstack, MassedCompute, Paperspace, Novita). Uptime, bandwidth, host rating and location are shown
for the sources whose APIs expose them (Vast, Clore, Akash); the rest contribute price
only. Reliability factors are explicit estimates blended with measured signals — not SLAs.
Market data, not investment advice.