Cloud-hosted AI isn’t always the right answer. Privacy requirements, latency constraints, offline scenarios, and cost pressures are pushing more intelligence to the edge, directly onto user PCs and devices.

With Foundry Local, Copilot+ PCs, and NPU-accelerated models, Microsoft is enabling developers to build, test, and run AI workloads locally using the same models, prompts, and workflows they deploy to the cloud.

In this session, you’ll learn how to develop local-first AI applications that run directly on Copilot+ PCs using the on-device Neural Processing Unit (NPU). We’ll explore how Foundry Local bridges local development and cloud deployment, how NPUs change performance and power profiles, and how to design applications that seamlessly move between edge and Azure.

You’ll see practical demos running models locally, invoking them from .NET and Python, and deciding when AI should run on the device versus in the cloud, without rewriting your application.