RunPod GPU pricing

RunPod is a GPU cloud for developers running Pods, Serverless endpoints and multi-node clusters. It fits teams that want self-serve GPUs with both datacenter-backed and marketplace-style capacity. Pod billing is usage-based, with on-demand, prepaid savings plans and separate storage rules.

Live RunPod prices (2026-07-08)

Cheapest listed per-GPU-hour rates verified at the source. Before storage, egress and networking.

Pricing model

RunPod prices by workload type: Pods for dedicated GPU instances, Serverless for inference workers and Clusters for multi-node jobs. The Pods documentation says compute and storage are billed by the second, and on-demand Pods have no commitment. Savings plans are prepaid commitments for a fixed term and apply to GPU compute only, while storage continues at standard rates. RunPod also requires enough account credit for the selected configuration before an on-demand Pod can be deployed.

Capacity tiers

RunPod exposes several capacity paths. Pods are the main self-serve instance product, with Community Cloud and Secure Cloud shown as separate pricing columns on the public pricing page. The homepage describes Serverless for production inference, Pods for development and agent hosting, and Instant Clusters for training runs. Secure Cloud is described as network-isolated infrastructure for workloads with stricter compliance needs, while Community Cloud is the lower-friction marketplace-style tier surfaced in pricing.

GPU lineup

RunPod publishes a broad GPU type table in its documentation. Current listed families include NVIDIA Blackwell GPUs such as B200 and B300, Hopper GPUs such as H100 and H200, Ampere GPUs such as A100, A40 and A30, Ada and RTX workstation cards, L4 and L40 variants, older V100 cards, and AMD MI300X. The pricing page highlights high-memory datacenter GPUs alongside RTX 4090, RTX 5090 and RTX A-series options.

Storage, egress and hidden costs

RunPod says Pods have no fees for data ingress or egress. Storage is separate from GPU compute: container disk, volume disk and network volumes each have their own billing treatment, and network volumes are portable between Pods. The docs warn that storage can continue accruing on stopped Pods, savings plans do not cover storage, and RunPod is not designed for long-term cloud storage. Backups to another storage location are recommended for critical data.

Reliability and support posture

RunPod publishes several reliability and compliance statements, but buyers should separate marketing claims from the contract they sign. Its homepage states an uptime commitment and points readers to the compliance page and Trust Center for current SLA and certification details. Its documentation says Secure Cloud uses T3 and T4 data centers with enterprise-grade security, reliability, redundancy and faster response times. Compliance coverage can vary by workload, region, provider and deployment model.

Free credits and referral programs

RunPod documents referral, affiliate, creator and startup programs. The referral program pays RunPod Credits as a percentage of referred Serverless and Pod spend during the documented reward window, with bonus credits after the referred account reaches the required funding threshold. The affiliate program is for qualifying high-performing referrers and pays cash through Partnerstack. The startup page lists Starter and Growth tiers, including a credit match tier and a starter-credit tier. See RunPod referrals and RunPod Startup Program.

Who should pick RunPod

Pick RunPod if you want a practical GPU platform with both self-serve instances and serverless inference in one account. It is especially relevant when you want per-second Pod billing, no RunPod egress fee, and a choice between lower-cost Community Cloud and more controlled Secure Cloud capacity. For long-running production, check savings-plan terms, stopped-Pod storage charges and region-specific compliance before committing. For sensitive or regulated work, verify the exact Trust Center documents for your deployment model.

FAQ

Profile verified 2026-07-06. Sources: www.runpod.io · www.runpod.io · docs.runpod.io · docs.runpod.io · docs.runpod.io · www.runpod.io · docs.runpod.io · www.runpod.io