About RunPod
Our mission is to deliver the best user experience in core GPU computing. Whether you're an experienced ML developer training a large language model, or an enthusiast tinkering with stable diffusion, we strive to make GPU resources as seamless and affordable as possible.
Big Cloud (AWS, GCP, Azure) has made it incredibly costly for developers, startups, and enthusiasts to access GPU resources. When we first started RunPod, we knew we wouldn't get far as a fancy UI wrapper on top of Big Cloud - we had to build our own GPU infrastructure from the ground up. One that was cheaper and more efficient, that would allow anyone to access dozens of compute hours on the most state-of-the-art machines for the price of a cup of coffee.
We turned to the GPU cloud community for help, and received an overwhelming amount of support. Hundreds of GPU owners across the world deeply resonated with our mission and listed their GPUs on Community Cloud - one of the first decentralized networks of GPU hosts of its time.
As we grew and built services on top of Community Cloud, we saw more and more users reach out about scaling up their GPU requirements. Developers and startups needed larger clusters, higher reliability, and extremely fast networking speeds to train AI models and bring them to production.
So we introduced Secure Cloud to the platform - GPUs we source and manage in some of the most reliable data centers across the world. With Secure Cloud, developers can access clusters of up to 1000x GPUs with incredibly high data transfer speeds, RAID 2 redundancy, localized network volumes, and best-in-class security, all at a 50%+ lower rate than Big Cloud.
Since then, we've built Serverless - robust architecture that abstracts away all of the devops complexity required to scale GPU usage up and down for AI inference/training. Our AI Endpoints are built on top of Serverless, giving users the ability to run robust generative AI models in a few lines of code.
There is still a lot of work left to be done to create the best user experience in core GPU computing. We will continue to work relentlessly to make the platform as seamless and accessible as possible.
Big Cloud (AWS, GCP, Azure) has made it incredibly costly for developers, startups, and enthusiasts to access GPU resources. When we first started RunPod, we knew we wouldn't get far as a fancy UI wrapper on top of Big Cloud - we had to build our own GPU infrastructure from the ground up. One that was cheaper and more efficient, that would allow anyone to access dozens of compute hours on the most state-of-the-art machines for the price of a cup of coffee.
We turned to the GPU cloud community for help, and received an overwhelming amount of support. Hundreds of GPU owners across the world deeply resonated with our mission and listed their GPUs on Community Cloud - one of the first decentralized networks of GPU hosts of its time.
As we grew and built services on top of Community Cloud, we saw more and more users reach out about scaling up their GPU requirements. Developers and startups needed larger clusters, higher reliability, and extremely fast networking speeds to train AI models and bring them to production.
So we introduced Secure Cloud to the platform - GPUs we source and manage in some of the most reliable data centers across the world. With Secure Cloud, developers can access clusters of up to 1000x GPUs with incredibly high data transfer speeds, RAID 2 redundancy, localized network volumes, and best-in-class security, all at a 50%+ lower rate than Big Cloud.
Since then, we've built Serverless - robust architecture that abstracts away all of the devops complexity required to scale GPU usage up and down for AI inference/training. Our AI Endpoints are built on top of Serverless, giving users the ability to run robust generative AI models in a few lines of code.
There is still a lot of work left to be done to create the best user experience in core GPU computing. We will continue to work relentlessly to make the platform as seamless and accessible as possible.
The Team
RunPod GPU Cloud Ecosystem
Secure Cloud
Tier 4 Datacenters
Our secure cloud runs in tiered datacenters located in Kansas, USA and Oslo, NOR. RunPod HQ is located in the Philadelphia area.
Enterprise Grade hardware
Our Secure Cloud Machines are equipped with fast SSD storage, enterprise interconnects between nodes, and fast redundant business internet.
Trust and Security
Our data centers are currently using UV&S for co-location services. Our racks share physical space with servers that are used by the government and for other sensitive workloads. Our personnel are required to adhere to strict privacy and security measures.
Community Cloud
The Best Pricing
Our community cloud prices are unbeatable in the industry. If you have a workload that does not require the additional guarantees that Secure Cloud offers, then this is the place to get unbeatable savings.
Highly Reliable
We require our hosts to keep at least 98% uptime to stay listed. This means that you are guaranteed a good user experience, regardless of whether you decide to go with Secure Cloud or Community Cloud.
Decentralized
Our Community Cloud hosts are vetted by us and are located all over the world. Join us in democratizing GPU computing.
Serverless
Autoscaling
Workers scale from 0 to 100 on our Secure Cloud platform, highly available and distributed globally.
AI Inference
We handle millions of inference requests a day and can scale to handle billions. Scale your machine learning inference while keeping costs low.
AI Training
Run machine learning training tasks that can take up to 24 hours. Spin up GPUs per request and scale down once done.
AI Endpoints
Built for Production
AI Endpoints are fully managed and scaled to handle any workload. They are designed for a variety of applications including Dreambooth, Stable Diffusion, Whisper, and more.
Get Started in Minutes
Start running inference on our AI Endpoints with a few lines of code.
Execution Based Pricing
Pay only for request execution time.