Answers · data as of Oct 8 2026

How much cheaper are spot/interruptible GPUs than on-demand?

Spot/interruptible GPU capacity runs about 50% below on-demand today (median across 15 SKUs with both tiers) — in exchange, the provider can reclaim the machine mid-run.

NVIDIA H100 PCIe: 77% below on-demand.
NVIDIA A100 40GB PCIe: 65% below on-demand.
NVIDIA H200 SXM: 59% below on-demand.
NVIDIA A100 40GB SXM: 54% below on-demand.
NVIDIA A100 80GB PCIe: 52% below on-demand.
NVIDIA L40: 52% below on-demand.
NVIDIA RTX PRO 6000 Blackwell: 51% below on-demand.
NVIDIA L40S: 51% below on-demand.
NVIDIA A100 80GB SXM: 50% below on-demand.
NVIDIA H200 NVL: 45% below on-demand.
NVIDIA B300: 42% below on-demand.
NVIDIA B200: 41% below on-demand.
NVIDIA H100 SXM: 41% below on-demand.
NVIDIA H100 NVL: 41% below on-demand.
NVIDIA RTX 5090: 28% below on-demand.

Spot prices in the market table →Methodology (tiers never mixed) →

This answer is computed from live data on every load — per-provider daily medians, per single GPU per hour, pricing tiers never mixed (see methodology). Quote it with attribution: “Data: Price of Compute — priceofcompute.com” (cite).