RunPod review

Facts from RunPod’s own pricing and docs pages, September 2026 — 70 read from a quoted sentence on the page we archived.

Summary

RunPod rents GPUs from $0.27 per GPU-hour on demand, bills per second, and asks no minimum commitment. Egress is free.

SOC 2 Type IIISO 27001HIPAAGDPR · DPASpend limitsVolume discountsFree egressSSOOpenAI-compatible API

From

$0.27/GPU-hr

Billing

Per second

Minimum commitment

None

Egress

Free

Uptime SLA

99.99% SLA on Enterprise agreements

Registered in

🇺🇸

Visit RunPod ↗

On-demand pricing

NVIDIA RTX PRO 6000 Blackwell 96GB

×1, on demand · RTX Pro 6000 96GB

$2.09/hr

Deploy

NVIDIA H100 SXM5 80GB

×1, on demand · H100 SXM 80GB

$3.49/hr

Deploy

NVIDIA H200 SXM5 141GB

×1, on demand · H200 141GB

$4.59/hr

Deploy

NVIDIA L40S 48GBLowest

×1, on demand · L40S 48GB

$1.09/hr

Deploy

NVIDIA B200 SXM6 180GB

×1, on demand · B200 180GB

$6.79/hr

Deploy

See all 19 RunPod offers →

What’s good

Lowest published on-demand rate is $0.27/GPU-hr (NVIDIA RTX A5000 24GB), with no minimum commitment

Egress is free — no per-GB charge on data leaving

Billing is per second

Self-serve signup — no sales call to get started

Publishes a 99.99% SLA on Enterprise agreements uptime commitment

Free credits available: $1

Plans

	Community Cloud pods	Secure Cloud pods	Serverless	Clusters
From	$0.27/hr (RTX A5000)	—	$0.58/hr (16 GB class)	$1.79/hr (A100 SXM)
Minimum term	None; 1 hour of credit to deploy	None; savings plans optional	None	None on demand; 3 months+ reserved
Billing	Per second	Per second; 3/6-month savings plans	Per second, worker start to stop	Per second on demand; reserved by contract
Uptime SLA	No uptime warranty	—	99.99% (Inference, Enterprise)	Guaranteed SLA on reserved
Tenancy	Container on a peer host's GPU	Container in partner T3/T4 datacenters	Autoscaling workers, scale to zero	Multi-node, InfiniBand/RoCE; reserved to 10,000+ GPUs
Provisioning	Under 30 seconds	Under 30 seconds	Sub-200 ms cold start (FlashBoot)	Minutes
Payment	Prepaid credits	Prepaid credits; invoicing over $5,000	Prepaid credits	Prepaid credits; Enterprise invoice

Details

Billing & access

How you pay

Prepaid credits; post-paid invoice for Enterprise

Currency

USD

Commitment discounts

3/6-month prepaid savings plans (pods); reserved clusters

Free credits

$1,000 startup credits (application); referral rewards

Approval or KYC

KYC for crypto payments; none otherwise stated

Refunds

Fees non-refundable; savings plans non-refundable

Auto-renewal

Subscriptions auto-renew; auto-pay optional

When funds run out

Pods stopped at $0; without a network volume, terminated

What you get

GPUs listed

30+ SKUs incl. B300, B200, H200, H100, A100, MI300X, L40S, RTX 5090/4090

The docs' GPU-type table lists 49 types from RTX A2000 (6 GB) to B300 (288 GB), including AMD MI300X and V100s.

Regions

31 regions across the US, Europe, Asia and Australia

Node configuration

vCPU/RAM per GPU on the rate card (e.g. H100 SXM: 20 vCPU · 125 GB)

Interconnect

1,600–3,200 Gbps InfiniBand or RoCE v2 (Clusters)

Datacenter tier

T3/T4 (Secure Cloud)

Local storage

Container disk $0.10/GB-mo (running); volume disk $0.10 running / $0.20 stopped

Persistent storage

kind: Network volume (NVMe, S3-compatible API)

price: $0.07/GB-mo under 1 TB, $0.05 above

billed while stopped: yes

Standard and high-performance tiers; network volumes are only available on Secure Cloud pods and are billed while pods are stopped.

Ingress

Free

Public endpoints

Serverless endpoints; exposed pod ports

Private networking

Tenant isolation on clusters (P_Keys / VLAN)

Images

Templates or any Docker container

Access methods

Web console, CLI (runpodctl), REST API, GraphQL API, SDKs, SSH

Console, runpodctl CLI, REST (v1/v2) and GraphQL APIs, Python/JS SDKs, and SSH into pods and serverless workers.

Model deployment

Serverless workers, vLLM worker, GitHub deploys, Hub templates

Managed inference

Serverless endpoints; Public Endpoints for hosted models

Serverless

Yes — per-second, scale to zero

FAQ

How much does RunPod cost?

As of September 2026, the lowest published on-demand rate is $0.27 per GPU-hour for the NVIDIA RTX A5000 24GB. RunPod publishes an on-demand rate for 19 GPU models. Billing is per second.

Does RunPod have an uptime SLA?

RunPod states 99.99% SLA on Enterprise agreements; none on Community Cloud.

Is there a minimum commitment on RunPod?

No. RunPod states none.

Does RunPod charge for egress?

No. RunPod states that egress is free.

Can I sign up to RunPod without approval?

RunPod states that signup is self-serve. On approval and identity checks it states: kYC for crypto payments; none otherwise stated.

Are RunPod payments refundable?

RunPod states: fees non-refundable; savings plans non-refundable.

Where are RunPod's data centres?

RunPod lists 31 regions across the US, Europe, Asia and Australia.

Who do you contract with at RunPod?

The contracting entity is Runpod, Inc..

Reviewed September 2026, from RunPod’s own site. 70 facts, each quoted from the page it was read on.