Microsoft has announced a new set of tools to make AI app development less of a headache for businesses. The centerpiece? Azure AI Foundry.

This new offering is designed to let developers switch between AI models from OpenAI, Mistral, Meta Platforms, or any other supported provider. It’s about flexibility, plain and simple—something that’s been a pain point for businesses as the pace of AI innovation outstrips their ability to adapt.

Scott Guthrie, Microsoft’s cloud computing chief, said, “Each new model—even if it’s in the same family—has benefits in terms of better answers or performance on some tasks, but you might have regressions on other things,” he said.

Fixing the broken workflow

Right now, 60,000 customers use Azure AI. That’s not a small number. And they’re leveraging the platform’s 1,700 models to power their apps. But here’s the thing: this process is clunky. Developers waste time wrestling with new models instead of innovating.

Every update or new release feels like starting from scratch, and businesses hate it. They don’t want to rip apart their workflows every time OpenAI or Meta releases something shiny.

That’s where Azure AI Foundry comes in. It’s a more streamlined system that allows companies to mix and match models without unnecessary headaches. Got an older OpenAI model that works fine? Stick with it. 

Want to try something newer from Meta? Switch it in, see if it’s better, and keep what works. It’s all about options. Parts of Foundry are an upgrade of Azure AI Studio, but new features include tools for deploying AI agents.

Despite offering more choices, Microsoft says it isn’t abandoning its tight relationship with OpenAI. Guthrie was clear: OpenAI’s models are still a big deal for Microsoft. But sometimes, businesses need alternatives, and Microsoft knows that. “Choice is going to be important,” Guthrie said.

The hardware powering it all

But of course, AI needs the muscle to run, and Microsoft knows that better than anyone. Last year, the company revealed its first homegrown AI chips, and now they’ve doubled down with two new pieces of hardware.

First up, a security microprocessor designed to protect encryption and signing keys. Starting next year, every new server in Microsoft’s data centers will include this chip.

Then there’s the data processing unit (DPU), which speeds up how data moves between networks, servers, and AI chips. It’s a direct competitor to similar hardware made by Nvidia, but Microsoft thinks its version is more efficient.

These DPUs are vital for handling the massive workloads required by today’s AI models, which, as Microsoft’s chip lead Rani Borkar put it, “are growing so big.” She said that every component in their infrastructure needs to work together seamlessly to keep things fast and efficient.

From Zero to Web3 Pro: Your 90-Day Career Launch Plan