We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
Paid plans start at $0.27/mo.
Runpod pricing starts at $0.27 per GPU hour for a Secure Cloud RTX A5000 Pod, according to the pricing page checked Nov. 6, 2026. There is no monthly or annual subscription in the listed model: Pods are usage-priced with no commitment, while Serverless is billed per worker hour. GPU choice, cloud type, runtime, and storage change the bill. Some Cluster GPUs and enterprise reservations require a sales quote.
01Every plan, line by line
| Plan | Billed annually | Billed monthly | Built for |
|---|---|---|---|
| Pods | — | Quote | Dedicated GPU instances and long-running jobs |
| Serverless | — | Quote | API inference workloads |
| Clusters | — | Quote | Multi-GPU jobs without commitments |
| Reserved Clusters | — | Quote | Enterprise deployments needing reserved capacity |
Add-ons and overages
- Storage: Container Disk $0.10/GB/mo; Volume Disk $0.10/GB/mo running or $0.20/GB/mo idle
- Network Storage (Standard): $0.07/GB/mo under 1TB; $0.05/GB/mo over 1TB
- Network Storage (High-Performance): $0.14/GB/mo
- Public Endpoints are usage-priced: examples include $0.005/request for Pruna Image T2I and $0.05/1,000 characters for Whisper V3 Large
- Enterprise and Serverless reservations: contact sales for pricing
02What the plan grid doesn't tell you
Storage is extra
Container Disk and Volume Disk list at $0.10/GB/month; Volume Disk costs $0.20/GB/month while idle.
Network storage rates vary
Standard Network Storage is $0.07/GB/month under 1TB and $0.05 over 1TB; High-Performance storage is $0.14/GB/month.
Cloud type changes rates
Community Cloud rates differ from Secure Cloud. For example, the listed H100 PCIe rates start at $1.99/hr on Community Cloud versus $2.89/hr on Secure Cloud.
Some GPU rates need sales
L40S, H100 SXM, and B200 Cluster pricing requires contacting sales. Reserved Cluster pricing is also custom.
03Which plan should you buy?
| If you are… | Get | Why |
|---|---|---|
| Dedicated GPU instances | Pods | Usage-priced per GPU hour, with no commitment; listed for dedicated instances and long-running jobs. |
| API inference workloads | Serverless | Billed per worker hour, with autoscaling and scale-to-zero when idle. |
| Multi-GPU jobs | Clusters | Listed for multi-GPU jobs without commitments, with scale up to 64 GPUs; some GPU rates require a sales quote. |
04Is Runpod worth the money?
Runpod offers a broad set of GPU options with pay-as-you-go Pods, Serverless, and Clusters, so the listed rates can be useful if you can match the GPU and billing model to your workload. But the lowest rate is not a full cost estimate: storage is billed separately, Community and Secure Cloud rates differ, and some Cluster and reserved capacity prices are custom. Reliability complaints temper the value, so weigh the operational risk alongside hourly rates.
Runpod pricing FAQ
Is Runpod on demand?
Yes. Pods are usage-priced with no commitment, and their listed rates are per GPU hour. Serverless is billed per worker hour and can autoscale to zero when idle. The pricing page lists usage-based rates rather than monthly or annual subscriptions.
How much is 1 GPU hour on Runpod?
The listed Secure Cloud Pod rates include $0.27 for an RTX A5000, $1.79 for an A100, and $2.89 for an H100 PCIe per hour. Rates vary by GPU and cloud type; Community Cloud examples include H100 PCIe from $1.99/hr.
How much does a GPU server cost on Runpod?
There is no single server price. The pricing page lists Pods per GPU hour, starting at $0.27/hr for a Secure Cloud RTX A5000. Your total depends on GPU choice and runtime, with storage billed separately; some Cluster rates require a sales quote.
How can you make money with Runpod?
Runpod provides GPU infrastructure; the listed pricing does not promise income or pay users. Any revenue would depend on what you build or run, while GPU, worker, and storage usage can add costs. Estimate those expenses against your own expected revenue before committing.