Consumer and A100 cards
Rent my GPU for AI: what renters pay for an RTX 4090 or an A100
Updated
Of the 8 providers on the board, RunPod, SaladCloud and Spheron publish an RTX 4090 rate per GPU-hour, and Oracle Cloud, RunPod and Spheron an on-demand A100 rate, on the pricing pages read on 2026-10-08.
The RTX 4090, provider by provider
Where a provider does price an RTX 4090, the rate is in the RTX 4090 column of the home page's price table, and the sentence it came from is beside it on the board on the dataset page, What GPU clouds charge per GPU-hour. Some of the providers pricing one are ones the board classes as marketplaces of third-party hosts, the kind of platform a card owner can list on.
The A100 on demand
Each A100 rate in the home page's price table is the provider's own figure per GPU-hour. The sentence it was read from, kept on the dataset page's board, names the machine, which can be a whole eight-card server priced per card.
What the rate means for your card
A renter's rate is what the platform charges for the card. Where a marketplace states a fee, an owner's payout comes out of that rate: TensorDock's host page states a revenue-share fee, and Vast.ai's own article says its live GPU prices are typically above what hosts earn. Clore's host pages pay rewards in CLORE on top of the rental price, so the published rate is a starting point for a listed RTX 4090 or A100, not the most it could bring in.