🖥️ GPU Cloud Pricing
Where to rent AI compute cheapest — $/GPU/hour across clouds, as of June 2026. On-demand rates are stable; spot and marketplace rates move with supply.
NVIDIA H100
| Provider | Type | $ / GPU / hr | $ / GPU / day |
|---|---|---|---|
| Vast.aicheapest | marketplace | $0.67 | $16 |
| Spheron | spot | $1.03 | $25 |
| RunPod | spot | $1.19 | $29 |
| RunPod (PCIe) | on-demand | $1.99 | $48 |
| Lambda | on-demand | $2.49 | $60 |
| RunPod (SXM) | on-demand | $2.69 | $65 |
| Google Cloud (A3) | on-demand | $3.00 | $72 |
| AWS (p5) | on-demand | $6.88 | $165 |
| Azure | on-demand | $12.29 | $295 |
NVIDIA A100 80GB
| Provider | Type | $ / GPU / hr | $ / GPU / day |
|---|---|---|---|
| Spheroncheapest | spot | $0.60 | $14 |
| TensorDock | marketplace | $0.75 | $18 |
| Spheron | on-demand | $1.07 | $26 |
| RunPod | on-demand | $1.19 | $29 |
NVIDIA B200
Curated from provider pricing pages and cross-provider roundups; reviewed quarterly. Marketplace rates (Vast.ai, TensorDock) vary by host reliability — treat as a floor, not a quote. For token prices instead of compute, see the API Pricing page.