---
title: "RunPod review: pricing, pros and cons"
description: "RunPod rents single NVIDIA GPUs by the second in two tiers, vetted partner data centres and a marketplace of third-party hosts, to teams that sign up with a card."
canonical: "https://saaskly.com/category/gpu-compute/runpod"
last-updated: "2026-09-22"
---

# RunPod review: pricing, pros and cons

> Saaskly verdict: **4.0/5**. RunPod rents single NVIDIA GPUs by the second in two tiers, vetted partner data centres and a marketplace of third-party hosts, to teams that sign up with a card.

- Category: [GPU & High Performance Compute](https://saaskly.com/category/gpu-compute.md)
- Vendor website: https://www.runpod.io
- Last updated: 2026-09-22

## Pros and cons

**Pros**

- One H100 SXM costs $3.49 an hour in Secure Cloud or $2.69 in Community Cloud, billed per second, and RunPod charges nothing for data transfer in or out
- Self-serve with a card through Stripe: add credit and deploy, with no sales call and no quota form, only an $80 an hour spending limit that rises automatically
- Holds ISO/IEC 27001:2022 under certificate RUN-ISMS-20260827 to 26 August 2029, plus SOC 2 Type II and SOC 3, audited by Sensiba LLP

**Cons**

- Storage on a stopped pod costs $0.20 per GB a month, double the $0.10 running rate, and a pod without a network volume is destroyed rather than paused when credit runs out
- No UK region: the nearest European sites are the Netherlands, France, Sweden, Czechia, Romania and Iceland
- Community Cloud hardware belongs to third-party hosts and the terms say RunPod "is only the venue to connect Hosts with consumers" with no availability warranty, and RunPod has stopped accepting new Community hosts
- Trustpilot sits at 3.8 from 315 invited reviews with 19% at one star, and the dated complaints are GPU supply and support replies of 36 to 72 hours

## Pricing

Month-to-month price first (what you pay without committing), then the annual-commit rate.

- **H100 SXM 80 GB, one GPU, Secure Cloud**: $2,548/mo, same price on any term
  - Includes: One H100 SXM 80 GB at $3.49 an hour on demand in Secure Cloud, $2,548 a month at 730 hours. The same GPU in Community Cloud, where the hardware belongs to third-party hosts, is $2.69 an hour or $1,964 a month, a 30% saving. Savings Plans take a 3 or 6 month upfront commitment but the discount is not published. USD, VAT not stated, read 22 Sep 2026.
- **RTX A5000 24 GB, one GPU, Community Cloud**: $117/mo, same price on any term
  - Includes: One RTX A5000 24 GB at $0.16 an hour in Community Cloud, $117 a month at 730 hours; $0.27 an hour or $197 a month in Secure Cloud. Volume storage $0.10 per GB a month while running and $0.20 while stopped; network volumes from $0.05 per GB a month; data transfer free both ways.

Minimum commitment: **$2,548** ($2,548/mo × 1 seat × 1 month (no commitment)).

## Specification

| Attribute | Value |
| --- | --- |
| H100 per GPU-hour* | $3.49/h (NVIDIA H100 SXM 80 GB (Secure Cloud)) |
| Committed H100 | Sales only |
| Cheapest GPU | $0.16/h (NVIDIA RTX A5000 24 GB (Community Cloud)) |
| 8xH100 node | Not on demand |
| Spot / interruptible | No |
| Cheapest dedicated server | None |
| Billing | Per second |
| Egress | Included |
| Storage (per GB/mo) | $0.100 |
| Self-serve signup | Yes |
| GPU quota | Automatic limit raise |
| Hardware** | Own hardware, colocated |
| UK location | No |
| H100 in the UK | No |
| EU locations | 5 |
| Locations | 18 |
| Owned in | US-owned |
| Owner | Independent |
| Cluster networking | Yes |
| Kubernetes | No |
| Slurm | Yes |
| Bare metal | No |
| API | Yes |
| Terraform | No |
| MCP server | Yes |
| Trustpilot (reviews) | 3.8 (315) |
| Committed pricing | Sales only |
| Committed term (months) | 3 |
| Stopped instance billed | Yes |
| Free credit | None found. A referral gives both sides a random $5 to $500 credit once the referred user spends $10. |
| Sales-only above | Reserved capacity, contract pricing and invoiced deposits above $5,000 |
| HQ | US |
| Publicly listed | No |
| Sovereign offering | None |
| ISO 27001 | Yes |
| SOC 2 | Yes |
| NVIDIA GPUs | Yes |
| AMD GPUs | No |
| Multi-GPU nodes | Yes |
| Dedicated servers | No |
| Object storage | No |
| Block storage | Yes |
| Private networking | Yes |
| CLI | Yes |
| 2FA | No |
| Phone support | No |
| Support channels | Ticket, Live chat, Discord community |

* One NVIDIA H100 80 GB per GPU-hour, on demand, in the currency the vendor shows a UK visitor; where the vendor sells no H100 the nearest GPU is named. Committed prices are the cheapest published rate and its minimum term.

** Who owns and runs the GPUs. "Third-party hosts" means the vendor is a marketplace for other people's machines and the SLA is theirs, not the vendor's.

## Our review

## Who owns it

RunPod is run by Runpod, Inc., a private US company founded in 2022 by Zhen Lu, its chief executive, and Pardeep Singh, its chief technology officer. Its terms of service give an address at 329 Bryant St, San Francisco, apply Delaware law and send disputes to the courts of New Jersey, so a UK or EU customer contracts with a US company under US law. There is no UK or EU contracting entity. The company raised a $20m seed round announced on 8 May 2024, co-led by Intel Capital and Dell Technologies Capital, with Intel Capital's Mark Rostick taking a board seat. Where the hardware sits depends on which tier you buy. Secure Cloud runs in what RunPod's documentation calls "T3/T4 data centers" operated by "vetted infrastructure partners who meet enterprise security standards including SOC 2, ISO 27001, and PCI DSS certifications", so the facilities are colocation partners rather than halls RunPod built. Community Cloud is a marketplace of what the documentation calls "peer-to-peer providers", and the terms of service are blunt about it: "Runpod is only the venue to connect Hosts with consumers, and Runpod is not a party to any interaction between compute consumer and any Host", with no warranty that Community capacity will be available or error-free. RunPod has also closed that tier to new supply, stating that it "is no longer accepting new hosts for Community Cloud".

## What it costs

Prices are in US dollars for a UK visitor, with no VAT position stated, read on 22 September 2026 from a page RunPod marks as updated on 13 September 2026. One H100 SXM 80 GB costs $3.49 an hour in Secure Cloud, which is $2,548 a month if you leave it running for 730 hours, and $2.69 an hour in Community Cloud, or $1,964 a month. The Secure tier therefore carries a 30% premium on the same GPU model, and the gap is wider on consumer cards: an RTX 4090 is $0.74 an hour in Secure against $0.34 in Community, more than double. The cheapest GPU of any kind is an RTX A5000 24 GB at $0.16 an hour in Community, $117 a month, or $0.27 in Secure. In between sit the H200 at $4.59 or $3.59, the A100 80 GB SXM at $1.59 or $1.39, and the L40S at $1.09 or $0.79. Savings Plans exist and take a 3 or 6 month commitment paid upfront, but RunPod does not publish the discount anywhere public, only inside the console, and the plan covers compute alone: storage stays at the standard rate and the plan is non-refundable. Reserved capacity and contract pricing need a sales conversation. No spot or interruptible rate appears on the pricing page. There is no published hourly price for an eight-GPU H100 node either: multi-node work goes through Instant Clusters, which price per GPU according to type and node count.

Storage is $0.10 per GB a month for container and volume disk while a pod runs, and $0.20 per GB a month once it is stopped, so idle data is billed at double. Network volumes are $0.07 per GB a month under 1 TB, $0.05 above it, and $0.14 for the high-performance tier, and they keep billing whatever the pod is doing. Data transfer costs nothing in either direction. Billing is per second, with the ledger settling every five minutes, against prepaid credit that RunPod states is non-refundable and cannot be withdrawn. You need at least an hour's worth of credit for the configuration before a pod will start, and new accounts carry a default spending limit of $80 an hour that rises automatically as the account builds history. Cards go through Stripe, cryptocurrency needs identity checks, and invoicing only opens up for deposits above $5,000 through sales. For the kind of buyer this site is written for, the arithmetic is small: 40 hours on a Secure Cloud H100 SXM in a month is $139.60, a 200 GB volume adds $20, and moving the results out adds nothing, so about $160 for a fine-tuning run.

## What you get and what to watch

A new account can launch the reference GPU the same day. You add credit, pick a region and deploy, with no sales contact, no business verification for card payments and no quota form. The only ceiling is that $80 an hour spending limit, which lifts itself rather than needing a ticket. The real constraint is supply rather than paperwork: one-star reviews dated 21 August and 15 September 2026 describe "near ZERO GPU supply" at busy times.

There is no UK region. RunPod's own networking announcement names six European sites, in Romania, Czechia (Prague), France (Paris), the Netherlands, Sweden and Iceland (Keflavik), alongside United States, Canadian and Australian regions, 18 in total. RunPod does not publish which GPU models sit in which region, so availability has to be checked live in the console. On the platform, you get single-GPU pods, Instant Clusters spanning 2 to 8 nodes and 16 to 64 GPUs with 1,600 to 3,200 Gbps between nodes, managed Slurm, serverless GPU endpoints, a REST API, the runpodctl command line tool and a first-party hosted MCP server at mcp.getrunpod.io. There is no managed Kubernetes service and no RunPod-published Terraform provider, only community ones on the registry. On compliance, RunPod holds ISO/IEC 27001:2022 under certificate RUN-ISMS-20260827, valid from 27 August 2026 to 26 August 2029, plus SOC 2 Type II, SOC 3, and HIPAA and GDPR programmes, all audited by Sensiba LLP. Its documentation says it "is fully compliant with the General Data Protection Regulation (GDPR) for data processed in European data center regions". The CLOUD Act is not addressed anywhere on the site, and the contracting party is American.

Support runs through a ticket portal and in-console chat, with a busy Discord that RunPod itself describes as a community space rather than an official support channel. There is a public status page at uptime.runpod.io. RunPod does not publish a standalone service level agreement, and Community Cloud carries no availability warranty at all under the terms of service. On Trustpilot's UK site the score is 3.8 out of 5 from 315 reviews, on a claimed profile where the company invites customers to review, with 64% at five stars and 19% at one. The dated one-star themes are GPU supply in August and September 2026, support replies taking 36 to 72 hours while pods keep billing, hardware not matching what was advertised ("broken nvswitch, locked gpus, less ram", 15 September 2026), and pods being deleted rather than paused when credit runs out (24 August 2026), which matches RunPod's documented behaviour. G2 blocked automated access, so there is no score from there. No regulatory action, lawsuit or reported data-loss incident was found.

*Not yet tested by Saaskly. Assessed on published pricing, vendor documentation, company filings, the vendor's own status and trust pages, and third-party reviews.*

---

Researched and drafted with AI assistance; reviewed and approved by Tim Meredith. See https://saaskly.com/ai-transparency.md and https://saaskly.com/editorial-policy.md.

Canonical HTML: https://saaskly.com/category/gpu-compute/runpod · Markdown index: https://saaskly.com/llms.txt