GPU.ai GPU Cloud Price Index - Live floor prices across 12+ clouds, snapshot 2026-10-04

Fetched 2026-10-04. Live floor price per GPU-hour across 12+ aggregated GPU clouds, normalized to per-GPU rates (multi-GPU node prices divided by GPU count). Zero markup - raw provider rates, billed per second. Snapshot 2026-10-04, refreshes hourly. 1,692 GPUs launchable across 27 models. SECURE (vetted datacenter) vs COMMUNITY (marketplace suppliers) pricing. Key per-GPU-hour secure/community rates: RTX A4000 $0.120/$0.070; RTX 3090 $0.500/$0.147; RTX 4090 $0.740/$0.335; RTX 5090 $0.990/$0.435; L4 $0.490/$0.330; L40 $0.780/$0.690; L40S $0.800/$0.790; A100 80GB $1.07/$0.970; RTX PRO 6000 $1.99/$1.69; H100 PCIe $1.98/—; H100 NVL $3.19/$2.59; H100 SXM $3.49/$3.36; H200 SXM $4.59/$3.59; B200 $6.79/$8.62; B300 $7.89/—. Market range (public on-demand list across major clouds): L40S $1.86-$2.50; A100 $2.70-$4.10; RTX PRO 6000 $2.50-$5.50; H100 SXM $6.16-$11.06; H200 SXM $6.30-$10.60; B200 $8.60. Data served by public no-auth API. Note: GPU.ai is itself a cloud aggregator, so these are their aggregated rates, not raw cloud list prices.