Runpod

Runpod is a GPU cloud for AI developers and ML teams that rents on-demand GPU pods, autoscaling serverless endpoints and multi-node instant clusters with per-second billing and no commitment minimums. It suits teams that need training or inference capacity this week and do not want a hyperscaler contract or their own cluster.

Runpod is a developer-focused GPU cloud founded in 2022 and led by CEO Zhen Lu. The platform has four main compute products. Pods are dedicated GPU or CPU containers billed by the hour or second. Serverless endpoints autoscale a containerized worker and scale to zero when idle. Instant Clusters provide multi-node GPU capacity with high-speed networking and Slurm or PyTorch setups, up to 64 GPUs without a commitment. Public Endpoints expose pre-deployed image, video, audio and text models through an API. A newer Flash workflow, shipped in the MIT-licensed runpod-python SDK, lets a developer decorate a Python function and run it on a remote GPU from a local terminal. Capacity comes in two tiers: Community Cloud, which is cheaper capacity from vetted third-party hosts, and Secure Cloud in partner data centers with compliance certifications. The GPU list on the pricing page runs from an RTX A5000 at $0.16 an hour on Community Cloud to a B300 at $8.99 an hour on Secure Cloud. In June 2026 Runpod raised $100 million led by Summit Partners at a $1 billion valuation, bringing total funding to $122 million after a May 2024 seed co-led by Intel Capital and Dell Technologies Capital. The company reports more than one million developers and over 20 billion serverless inference requests to date. Named users include Deep Cogito, which trained its Cogito v1 open models on Runpod. It competes with neoclouds such as CoreWeave and Lambda, marketplaces such as Vast.ai, and serverless platforms such as Modal and Replicate.

Category

Infrastructure & Cloud

Pricing

Usage-based, Contact for pricing

Target Market

ML Engineers, Data Scientists, Enterprise Developers, AI Startups, CTOs

Deployment

Cloud-only, API-based

Founded

2022

Headquarters

Mount Laurel, United States

Customers

More than 1 million developers (company-reported, June 2026)

• ✓
  GPU Pods

  Dedicated GPU or CPU containers from RTX A5000 to B300, billed per hour or per second, for interactive development, training and fine-tuning.

• ✓
  Serverless endpoints

  Autoscaling job-queue endpoints that run a containerized Python handler and scale to zero when idle, so inference costs track actual traffic.

• ✓
  Instant Clusters

  Managed multi-node GPU clusters with high-speed networking and Slurm or PyTorch setups, up to 64 GPUs with no commitment, for distributed training.

• ✓
  Community and Secure Cloud tiers

  Lower-priced Community Cloud capacity from vetted hosts alongside Secure Cloud data centers covered by SOC 2 Type II and ISO 27001, so teams pick price or compliance per workload.

• ✓
  Flash and the runpod-python SDK

  An MIT-licensed SDK and rp CLI that turn a decorated Python function into a remote GPU job or a deployed endpoint, cutting container boilerplate.

• ✓
  Public Endpoints

  API access to pre-deployed image, video, audio and text generation models for teams that want to call a model without hosting it.

• •
  Open-model training

  A model lab trains and iterates on an open LLM family on rented GPUs, as Deep Cogito did for Cogito v1, without building its own cluster.

• •
  Bursty inference APIs

  A product team deploys a custom model as a serverless endpoint that scales up under load and to zero overnight, paying only for active seconds.

• •
  Fine-tuning sprints

  An enterprise data science team rents H100 or H200 pods for a few days to fine-tune a domain model, then shuts them down and stops billing.

• •
  Bulk embedding and batch jobs

  Engineers rent cheap GPUs in short time slices to generate embeddings or process media in bulk, a pattern Hacker News users describe as cheaper than owning hardware.

List prices are public and metered per GPU by the hour or second, with no commitment minimums. Community Cloud is cheaper than Secure Cloud for the same GPU (H100 SXM: $2.69 vs $3.99 an hour). Storage is billed separately: $0.10/GB/month for container and running volume disks, $0.20/GB/month for idle volumes and $0.05-0.14/GB/month for network storage, a line some Trustpilot reviewers say surprised them. Reserved clusters, serverless reservations and contract discounts go through sales. No free tier is listed.

This page was written from 9 sources, 7 on domains other than runpod.io.

1. 1.runpod.io — pricingvendor
2. 2.runpod.io — compliancevendor
3. 3.docs.runpod.io — overview
4. 4.github.com — runpod python
5. 5.siliconangle.com — runpod raises 100m build leading cloud platform ai developer
6. 6.summitpartners.com — runpod raises 100m led by summit partners to accelerate the
7. 7.sacra.com — runpod
8. 8.trustpilot.com — runpod.io
9. 9.hn.algolia.com — search