Microsoft is reportedly testing Copilot+ AI features with discrete GPUs instead of NPUs — a feature available on Windows App SDK with a Windows Insider Experimental Channel build and Developer Mode turned on

Copilot PCs have been around for a couple of years since Microsoft launched them in 2024, and while the company tried to push NPU-equipped laptops towards users to take advantage of these new features, it seems that it’s planning to reverse course. According to Windows Latest, an experimental Windows App SDK available on GitHub now lets you run Language Model APIs on supported GPUs, starting with the Nvidia GeForce RTX 30-series cards with at least 6GB of VRAM. However, it also requires a Windows Insider Experimental Channel and Developer Mode switched on. Hence, you need to go through some hoops to turn on local AI inferencing on Windows, even if you don’t have an NPU-equipped device.

You still won’t get all the features found in a Copilot+ PC even if you went through all the hoops to activate this feature, but it’s a sign of things to come for local AI on Windows PCs, in general. It’s still not clear why Microsoft is seemingly abandoning the Copilot+ PC advantage it…