# infr.sh — GPU hosting

Updated: 2026-05-25

GPU rental + serverless GPU options for self-hosting models and running training jobs. Prices in USD; verify on provider pricing pages before committing.

**Cheapest H100 spot:** Vast.ai · H100 80GB · $0.34/hr
**Cheapest H100 on-demand:** Spheron / GMI · $1.03/hr

## Rentals ($/GPU-hour)

| provider | gpu | spot/hr | on-demand/hr | kind |
| --- | --- | ---: | ---: | --- |
| Vast.ai | H100 80GB | 0.34 | 1.65 | marketplace |
| Vast.ai | A100 80GB | 0.67 | 2 | marketplace |
| Vast.ai | B200 | — | 4.5 | marketplace |
| RunPod | H100 PCIe 80GB | — | 1.99 | cloud |
| RunPod | H100 SXM 80GB | 1.19 | 2.69 | cloud |
| RunPod | A100 80GB | — | 1.89 | cloud |
| RunPod | B200 | — | 4.99 | cloud |
| Lambda | H100 SXM 80GB | — | 2.99 | cloud |
| Lambda | A100 80GB | — | 1.79 | cloud |
| Lambda | B200 | — | 4.99 | cloud |
| CoreWeave | H100 80GB (8x HGX) | 2.83 | 6.16 | cloud |
| CoreWeave | B200 (8x) | 3.96 | 8.6 | cloud |
| Modal | H100 80GB | 3.95 | 3.95 | serverless |
| Modal | A100 80GB | — | 2.5 | serverless |
| Modal | A100 40GB | — | 2.1 | serverless |
| SF Compute | H100 80GB | — | 1.96 | cluster |
| SF Compute | H200 80GB | — | 2.49 | cluster |
| AWS | H100 80GB (p5) | 2.5 | 4.1 | cloud |
| AWS | A100 80GB (p4de) | 1.85 | 4.1 | cloud |
| Spheron / GMI | H100 80GB | 0.47 | 1.03 | marketplace |
| Spheron / GMI | A100 80GB | 0.6 | 1.07 | marketplace |
| Spheron / GMI | B200 | — | 2.12 | marketplace |
| Replicate (raw GPU) | H100 80GB (8x) | — | 5.49 | serverless |

## Serverless ($/second)

| provider | gpu | $/sec |
| --- | --- | ---: |
| fal | H100 80GB | 0.00079 |
| Modal Function | H100 80GB | 0.0011 |
| Modal Sandbox | H100 80GB | 0.0011 |
| Replicate | H100 80GB | 0.00153 |
| RunPod Serverless | H100 80GB | 0.00078 |

JSON: https://www.infr.sh/i/gpu.json