GPU Cloud

Rent GPU compute by the hour or second for inference, training, and fine-tuning workloads.

**Provider:** [RunPod](https://www.runpod.io)

**Pricing:** Usage-based, from ~$0.25 / GPU-hour

**Best for:** Developers who need dedicated GPU machines without long-term contracts

RunPod offers serverless and pod-based GPU instances for ML workloads. Supports PyTorch, CUDA, and containerized deployments. Useful as a backend for agents running local or fine-tuned models.