Azure AI Foundry Projects vs Hubs — What’s the Difference?

By Ivana Tilca · July 8, 2025 · 4 min read

Azure AI Foundry gives you two ways to organize your work — Projects and Hubs — and picking the wrong one means a painful migration later. Here's what each is for, why Microsoft is investing in Projects, and how to choose the right foundation for a new AI build.

It's been a while since Azure AI Foundry hit the scene, and Microsoft keeps evolving it. With the updates from Build this year, it's more important than ever to understand the difference between Foundry Projects and Foundry Hubs — especially if you're starting a new AI project and don't want to build on the wrong foundation. Let's break it down simply.

What Azure AI Foundry is today

Think of Azure AI Foundry as your all-in-one platform for building AI-powered applications. In the past, developers had to stitch together multiple Azure services just to get something running — one resource for models, another for storage, another for orchestration. Foundry bundles the essentials into a single resource. It's the difference between a cluttered toolbox and a streamlined power station where everything is already wired together.

But "Foundry" isn't a single shape. There are two ways to organize your work inside it — Projects and Hubs — and picking the right one early saves you a painful migration later.

Foundry Projects: safe developer playgrounds

A Foundry Project gives each developer (or each initiative) a dedicated space to build, test, and experiment. Projects inherit security settings from the parent resource, so your IT or platform team stays in control without hovering over every action. It's like giving each builder their own studio while everyone shares the same blueprints and safety rules.

This is the model Microsoft is investing in, and it comes with a set of dev-friendly features worth knowing:

A unified Foundry API. One standardized API to run agentic apps across different model providers — OpenAI, Mistral, and others. No more rewriting your integration code just to switch models.

First-class SDKs. Foundry offers SDKs in Python, C#, JavaScript/TypeScript, and Java, so you can integrate without starting from scratch in whatever language your stack already uses.

A VS Code extension. Manage agents and models right from your editor instead of browser-hopping between the portal and your code.

Granular RBAC roles. Roles like Admin, Project Manager, and Project User separate infrastructure access from data access — more secure, less mess, and easier to reason about as a team grows.

Foundry Hubs: the older, heavier model

Hubs are the earlier organizational model, built around a shared hub that multiple projects hang off of. Hubs carry more machinery and are still the right home for advanced scenarios — particularly deep model training and legacy machine-learning workflows that depend on that setup. If your work is squarely in classic ML training territory, Hubs still have a role.

The trade-off is that Hubs are more complex to set up and manage, and they aren't where the newest Foundry features are landing.

So which should you use?

Here's the practical decision:

Building AI agents or apps? Start with Foundry Projects. That's where Microsoft is investing, where the latest features appear first, and where the developer experience is smoothest.

Still doing deep model training or maintaining legacy ML? You'll want Hub-based projects, at least for now, because that's where those capabilities live.

If you're genuinely unsure, default to Projects. It's far easier to grow into more advanced needs from a modern foundation than to migrate off an older one after you've built on it.

A note on migrating and future-proofing

If you already have work in a Hub, don't panic — you don't need to rip everything out tomorrow. But when you start something new, do it in a Project so you're building on the platform's direction rather than against it. Microsoft's clear signal, reinforced at Build, is that Projects are the future of building AI apps on Azure.

Final thoughts

Foundry Projects are the future of building AI apps on Azure — they're unified, easier to manage, and packed with dev-friendly features that keep improving. Hubs still have their place, especially for advanced model training, but the momentum is unmistakably on the Projects side.

So if you're just getting started, or planning your next big build, Foundry Projects are your best bet. Choose the modern foundation now, and you'll spend your energy building instead of untangling.