RunPod GPU pricing 2026
Pay-per-second GPU cloud with a Community Cloud (capacity from independent operators) and Secure Cloud (vetted data centres). Strong on inference workloads and Pod or Serverless deployment.
Published rate card
Per-GPU per-hour rates pulled from the vendor pricing page linked above, Last verified June 2026. Rates exclude storage, egress, and any managed-service uplift. Reserved-capacity contracts typically improve on these rates.
| GPU | Configuration | Per GPU per hour |
|---|---|---|
| H100 SXM | Secure Cloud, per-GPU | $2.990 |
| H100 PCIe | Community Cloud | $1.990 |
| A100 80GB SXM | Secure Cloud | $1.890 |
| A100 80GB PCIe | Community Cloud | $1.190 |
| L40S | Secure Cloud | $1.190 |
| RTX A6000 | Community Cloud | $0.490 |
Hidden costs to watch
- Community Cloud capacity is unmanaged and can be reclaimed; Secure Cloud is the SLA-backed tier at a premium.
- Network volume storage billed separately at GB-month rates.
- Serverless inference billed per-second of active GPU time and idle-state warming is metered.
What RunPod is best for
Inference, fine-tuning, and short-burst training where pay-by-the-second and the lowest-cost community cloud beats hyperscaler rates.
See the GPU cloud buying guideWorked example
Acme Vision Co. (illustrative example, not a real company) needs to train a 10 billion-parameter vision-language model on a fixed 8-GPU cluster for 30 days at 18 hours per day duty cycle.
At RunPod's cheapest published rate of $0.49 per GPU-hour, the run costs $2,117 for raw GPU compute, before storage, egress, and MLOps overhead. Add the typical 25 percent year-one uplift and the modelled spend is $2,646. Use the calculator on the homepage to model your own GPU class, cluster size, and duty cycle.
Last verified June 2026. RunPod rates change frequently. Always obtain a vendor quote before purchase.
Visit RunPod pricing page