RunPod

RunPod is a cloud platform, founded in October 2022, that gives developers on-demand access to GPU infrastructure for training, fine-tuning, and running AI models. Rather than moving between separate services for experimentation and production, users can go from a Pod running locally-like development to a scaled inference endpoint on the same platform, with the company reporting adoption from AI companies including Hugging Face, Perplexity, Replit, and Cursor. The platform offers three main compute types: Serverless endpoints that autoscale with sub-200ms cold starts and no idle cost, dedicated GPU Pods available as Reserved or Spot instances, and multi-GPU Clusters for distributed training and batch inference jobs. RunPod gives access to more than 30 GPU types, including H100, A100, L40S, B200, and RTX 4090 cards, spread across 31 regions, along with persistent network storage that does not charge for data egress. For programmatic use, RunPod exposes a REST API and a Python SDK, and it integrates with coding agents such as Claude Code and Cursor through a dedicated Skills package. Billing is usage-based across all compute types, with no contracts or minimum spending commitments, which suits developers who want to scale GPU usage up or down without long-term agreements.