Run AI workloads on any cloud, store on Hugging Face: zero-egress storage with SkyPilot

SkyPilot, an open-source tool for running AI workloads across multiple clouds, now integrates with Hugging Face to let you store results directly on Hugging Face's platform without egress fees. Egress fees are the charges cloud providers impose for moving data out of their services, which can add up quickly. This integration means you can run AI workloads on any cloud provider—like AWS, Google Cloud, or Azure—and save the outputs to Hugging Face, avoiding those costs entirely.

This matters because it makes AI development more accessible and cost-effective. Previously, you might have been locked into a single cloud provider to avoid egress fees, limiting your flexibility. Now, you can choose the best cloud for each workload—whether it's cheaper compute, better GPUs, or specific services—and still store your data where it's easiest to share and collaborate. It's like being able to shop at any store but return everything to the same warehouse for free.

To try this today, go to the SkyPilot documentation and follow the steps to set up a Hugging Face storage bucket. Once configured, you can run your AI workloads and store the outputs directly on Hugging Face without worrying about extra charges. It's a simple way to make your AI projects more flexible and cost-effective.