Neocloud
RunPod
Community and secure cloud tiers, per-second billing, and serverless GPU endpoints.
Two fleets: Secure Cloud (datacenter-grade) and Community Cloud (vetted third-party hosts, cheaper). Serverless endpoints with scale-to-zero make it a favorite for indie AI products. Consumer cards (4090/5090) rent legally here at prices datacenter GPUs can't touch.
GPU pricing on RunPod
On-demand list per GPU-hour, last reviewed 2026-09. Methodology.
| GPU | $/GPU-hr | Notes |
|---|---|---|
| RTX 3090 | $0.22 | Community Cloud |
| V100 | $0.29 | |
| RTX 4090 | $0.34 | Community Cloud |
| L4 | $0.49 | |
| RTX PRO 6000 | $0.50 | Secure Cloud |
| RTX 5090 | $0.69 | Community Cloud |
| A100 40GB | $0.99 | |
| L40S | $1.09 | |
| A100 80GB | $1.39 | Secure Cloud; Community from ~$1.09 |
| MI300X | $2.39 | |
| H100 SXM | $2.69 | Secure Cloud; Community from ~$1.99 |
| H100 PCIe | $2.89 | |
| H200 | $4.59 | Secure Cloud |
| B200 | $6.79 |
Strengths
- Consumer GPUs at community prices
- Serverless with fast cold starts
- Per-second billing
Watch for
- Community tier reliability varies by host
- Multi-node training limited