Pricing / Predictable dedicated GPU capacity
Cheapest price per gpu hour for Predictable dedicated GPU capacity
Unit: usd_per_gpu_hour only — nothing converted or blended. min sourced GPU-hour SKU among candidates; not coerced to $/1M; per-second GPU was not included; not quality-adjusted. Updated 2026-10-27 (newest asOf / re-check on this page).
Pick is not a price winner. lambda matching USD per GPU-hour is 2.79; cheapest among candidates is runpod at 0.34.
| # | Provider | SKU | USD per GPU-hour | Source | asOf / re-checked | First seen / last changed |
|---|---|---|---|---|---|---|
| 1 | RunPod | RTX 4090 Pod (Community Cloud) | $0.34 | page | 2026-10-19 / not reconfirmed | 2026-10-19 / no change recorded |
| 2 | Massed Compute | L40S (48GB) x1 | $0.97 | page | 2026-10-19 / not reconfirmed | 2026-10-19 / no change recorded |
| 3 | Crusoe | NVIDIA L40S 48GB | $1.5 | page | 2026-10-19 / 2026-10-27 | 2026-10-19 / no change recorded |
| 4 | Nebius | NVIDIA RTX PRO 6000 | $1.8 | page | 2026-10-19 / 2026-10-27 | 2026-10-19 / no change recorded |
| 5 | Verda | 1x L40S 48GB | $1.845 | page | 2026-10-27 / 2026-10-27 | 2026-10-19 / 2026-10-27 |
| 6 | LambdaDepot Pick | NVIDIA A100 SXM 80GB | $2.79 | page | 2026-10-19 / not reconfirmed | 2026-10-19 / no change recorded |
| 7 | CoreWeave | NVIDIA HGX H100 inference single GPU | $6.16 | page | 2026-10-19 / not reconfirmed | 2026-10-19 / no change recorded |
Price only: not quality-, latency- or limit-adjusted. The Depot Pick is editorial and is not chosen by price. Brief: /jtbds/predictable-gpu · Agents: GET /api/quote?jtbd=predictable-gpu