RunPod Pricing Explained (2026), Plus the Best Alternatives
RunPod bills GPU Pods by the second with no data-transfer fees: an H100 SXM was $3.99 an hour on Secure Cloud and $2.69 on Community Cloud on 17 October 2026, and an RTX 4090 $0.89 or $0.34. The surprise costs are storage on stopped Pods and Serverless rates that run higher than Pods. Vast.ai is usually cheaper for interruptible work; Lambda and DigitalOcean for steady on-demand.
Disclosure: Some links on this page are affiliate links. If you buy through them, we may earn a commission at no extra cost to you. It never changes our verdicts or rankings. How we make money.
Pick the wrong RunPod tier and you can pay about 50% more for the same GPU hour: an H100 SXM on RunPod was $3.99 on Secure Cloud and $2.69 on Community Cloud on 17 October 2026. RunPod is where many indie builders rent GPUs (it reported $325M in annual recurring revenue in October 2026), and its pricing page is long.
Here’s what you’ll actually pay, the storage and Serverless costs that catch people out, and the cheaper alternatives.
RunPod Pod prices
Per hour, from RunPod’s pricing page on 17 October 2026:
| GPU | Secure Cloud | Community Cloud |
|---|---|---|
| H200 | $5.29 | $3.59 |
| H100 SXM | $3.99 | $2.69 |
| H100 PCIe | $2.89 | $1.99 |
| A100 SXM | $1.79 | $1.39 |
| L40S | $1.09 | $0.79 |
| RTX 5090 | $1.19 | $0.69 |
| RTX 4090 | $0.89 | $0.34 |
| RTX 3090 | $0.50 | $0.22 |
Pods are billed by the second for compute and storage, with no fees for data in or out. You need at least an hour of credit on your account to start one (RunPod docs). Live prices across providers are on our GPU price index.
Secure Cloud vs Community Cloud
- Secure Cloud: T3/T4 data centres, high redundancy, standard prices. Use it for production and sensitive data.
- Community Cloud: vetted peer-to-peer hosts, much cheaper, reliability varies. RunPod stopped accepting new Community Cloud hosts, though existing capacity stays available (RunPod docs).
For training runs you can checkpoint, Community Cloud (or spot capacity) cuts the bill sharply. For a client-facing endpoint, pay for Secure.
Storage: the cost people forget
| Storage | Price | When it’s charged |
|---|---|---|
| Container disk | $0.10/GB/month | Only while running; erased when stopped |
| Volume disk | $0.10/GB/month running, $0.20 stopped | Always, until you delete the Pod |
| Network volume | $0.07/GB/month (under 1 TB), $0.05 above | Always |
A stopped Pod with a 200 GB volume still costs $40 a month. Delete Pods you’re done with, and keep datasets on a network volume you can attach to any Pod.
Serverless
Serverless endpoints bill per second from when a worker starts until it fully stops, including cold start and idle timeout. Flex workers scale to zero; active workers run 24/7 at a discount you arrange with sales. The per-hour rates are higher than Pods: H100 $4.79, A100 $2.72, RTX 4090 $1.10 (RunPod docs). Serverless wins when traffic is spiky; Pods win for steady load.
RunPod also sells 3- and 6-month savings plans (paid upfront, GPU compute only).
RunPod alternatives
H100 on-demand prices from each provider’s own page on 17 October 2026:
| Provider | H100 price per GPU-hour | Type | Source |
|---|---|---|---|
| RunPod | $3.99 Secure / $2.69 Community (SXM) | On-demand | RunPod |
| TensorDock | From $2.25 (SXM5) | Marketplace | TensorDock |
| Lambda | $4.29 (1x) / $3.99 (8x) SXM; $3.29 PCIe | On-demand | Lambda |
| DigitalOcean | $4.41 ($3.26 with a 12-month reservation) | On-demand | DigitalOcean |
| Nebius | $4.50 | On-demand | Nebius |
| CoreWeave | About $6.16 ($49.24 per 8-GPU instance) | On-demand | CoreWeave |
| Vast.ai | Market rate (no H100 SXM offers listed when we checked) | Marketplace | Vast.ai |
- Cheapest hours: Vast.ai and other marketplaces, especially interruptible bids, which Vast advertises as 50%+ cheaper. Reliability depends on the host.
- Steady on-demand from a traditional cloud: Lambda and DigitalOcean. DigitalOcean bills GPU Droplets even when they’re powered off, with a 5-minute minimum.
- Consumer GPUs for image and video models: RunPod’s Community RTX 4090 at $0.34 is hard to beat.
Prices moved a lot in 2026: several providers raised H100 list prices between August and October. Check the GPU price index before you commit, and if you own a GPU, its calculator estimates what renting it out could earn.
Tools in this guide
RunPod
AI CodingRent GPUs by the second with Pods, Serverless endpoints and network storage.
Vast.ai
AI CodingRent GPUs from independent hosts who set their own prices: on-demand, interruptible or reserved.
DigitalOcean
AI CodingRun your own server from $4/mo, launch a hardened OpenClaw 1-Click or rent GPU Droplets.
Questions people ask
How much does RunPod cost per hour?
On 17 October 2026: H100 SXM $3.99 (Secure) or $2.69 (Community); A100 SXM $1.79 or $1.39; RTX 5090 $1.19 or $0.69; RTX 4090 $0.89 or $0.34. Billing is per second.
What's the difference between RunPod Secure Cloud and Community Cloud?
Secure Cloud runs in T3/T4 data centres with high redundancy, for production and sensitive data. Community Cloud uses vetted peer-to-peer hosts: cheaper, with variable reliability. RunPod no longer accepts new Community Cloud hosts.
Does RunPod charge for storage when a Pod is stopped?
Yes. Volume disk costs $0.10 per GB a month while running and $0.20 while stopped. Container disk isn't charged when stopped, but it's erased. Network volumes cost $0.07 per GB a month (or $0.05 above 1 TB).
What is the best RunPod alternative?
For the cheapest hours on fault-tolerant jobs, Vast.ai's marketplace. For steady on-demand H100s from a traditional cloud, Lambda ($3.99 per GPU-hour on 8x) or DigitalOcean ($4.41 on demand, $3.26 reserved). Check the live GPU price index before deciding.