Welcome to Foundry

Foundry is a self-serve cloud platform for machine learning research and infrastructure teams

Foundry gives ML practitioners the flexibility to spin up short-term GPU instances for model experimentation, large-scale training, batch jobs, live inference, and more without long-term contracts.

Foundry Cloud Platform is currently invite-only as we scale up capacity. To request access, join the waitlist.

Getting started

Quickstart guide

Set up your account and start running your workloads in minutes.

Reserving compute

Reserve interconnected nodes for as little as 3 hours for preplanned workloads.

Spot compute

Burst capacity by placing limit-price bids for spot instances.

API overview & quickstart

Create and access compute programmatically

Instance types & storage options

Instance types & specs

See what compute options you have on Foundry.

Persistent storage

Learn about persistent storage options for your workloads.