RunPod

Runpod is a GPU cloud for AI development. It covers the full lifecycle on one account, from experiments and fine tuning to training and production inference, and it is aimed at developers and teams that need on demand GPU compute. There are three main products. Pods are dedicated GPU instances that start in under a minute, across more than 30 GPU types from RTX 4090 class cards to B200 and B300, and 31 regions are listed. Serverless runs inference endpoints that scale from zero to thousands of workers, with a user pushing a handler and paying only for what runs. Clusters support multi node jobs. A Global Volumes feature, in beta, lets Serverless workers start in any data center against one stored copy of model weights. Reserved capacity and contract pricing are available through sales. Pricing is per second or per hour and varies by GPU and by Community Cloud or Secure Cloud. Published Pod examples include the RTX Pro 6000 at $2.49 per hour, the H200 at $5.29 per hour, the B200 at $7.99 per hour and the B300 at $8.99 per hour. Storage starts at $0.05 per GB per month, with container disk at $0.10 per GB per month and high performance network storage at $0.14 per GB per month. Serverless rates depend on the GPU type selected.