Login
Pricing

Set by the market, not by us

Every card on the marketplace, with the rate it is listing at today and the range it has traded in. On-demand, interruptible or reserved — pick the one that matches how much interruption your job can take.

52GPU models listed
$0.02Lowest hourly rate
Per secondBilling granularity

Rates move with supply and demand and are shown for reference. The observed range covers listings we have seen for that model, outliers included — a single mispriced machine can stretch the top end, so read the median column first.

FAQ

Common questions

Why is the observed range so wide on some models?
Because it is a marketplace, not a price list. A provider in a cheap-power region prices differently from one in a metro data center, and occasionally someone lists a machine at a number that makes no sense. The range shows what has actually been listed, outliers included, rather than a tidied-up version.
Is the "from" price what I will pay?
It is a current reference point, not a quote. What you pay depends on which specific machine you rent — its region, reliability history, interconnect and disk. Query live offers from the CLI to see real availability at the moment you want to launch.
How much cheaper is interruptible?
It varies by model and by how contested that GPU is right now, which is why we do not publish a fixed discount. If your job checkpoints cleanly, it is usually worth checking; if a lost hour is expensive, it usually is not.
Do storage and bandwidth cost extra?
Providers bill disk and bandwidth separately from GPU time, at rates shown on each offer. For short jobs it is noise; for anything moving large datasets repeatedly it is worth reading before you launch.
Can I get a better rate at volume?
Yes — for sustained multi-node load, a reserved block is normally cheaper than paying on-demand around the clock. Tell us the shape of the job and we will come back with a number.

Start with $5

No contract, no minimum. Rent one card for an hour and see whether the numbers hold.

Get started Query from the CLI

Transparent GPU rental prices

Every GPU on Linkhome AI is priced by the market, so AI compute rental rates reflect real supply and demand. Compare hourly GPU rental prices for data-center GPUs like H100, H200 and B200 and for consumer GPUs like the RTX 4090.

Pick a card, click Rent and book it from your account — billed per second, with no long-term commitment unless you want reserved AI infrastructure.

How GPU rental worksReserved GPU clustersVolume pricing