# Lambda Pricing

**Source:** https://lambda.ai/pricing — **Snapshot:** 2026-08-12
**Billing:** hourly, pay-as-you-go on single-GPU instances; multi-GPU HGX clusters require a 2-week–1-year commitment. **No egress fees** (the main differentiator vs Vast/RunPod for big weight and video transfers).

On-demand single-GPU instances (self-serve):

| GPU | VRAM | $/hr |
| --- | ---: | ---: |
| Quadro RTX 6000 | 24 GB | 0.69 |
| V100 | 16 GB | 0.79 |
| A6000 | 48 GB | 1.09 |
| A10 | 24 GB | 1.29 |
| A100 PCIe 40GB | 40 GB | 1.99 |
| A100 SXM 40GB | 40 GB | 1.99 |
| A100 SXM 80GB | 80 GB | 2.79 |
| GH200 | 96 GB | 2.29 |
| H100 PCIe | 80 GB | 3.29 |
| H100 SXM | 80 GB | 3.99–4.29 |
| B200 SXM6 | 180 GB | 6.69–6.99 |

(1-, 2-, 4-, 8-GPU configs of the same card vary the rate slightly — range shown.)

Committed clusters (2 weeks – 1 year), per GPU-hr:

| Cluster | 16 GPUs | 64 GPUs | 256+ GPUs |
| --- | ---: | ---: | ---: |
| HGX H100 | 6.16 | 5.85 | 5.54 |
| HGX B200 | 9.86 | 9.36 | 8.87 |

Notes:
- H100 SXM at $3.99 is the headline value — same card as RunPod secure ($2.99) but no community-pool shenanigans and no egress fees.
- GH200 at $2.29 is interesting for video models with big context needs.
- No spot market; no idle-reclaim. Predictable, premium.

**Verify live before committing spend** — this is a snapshot.