Microsoft tests Copilot+ AI on Nvidia RTX GPUs

Microsoft is experimenting with running parts of the Copilot+ PC local AI suite on Nvidia RTX graphics cards instead of relying on dedicated neural‑processing units (NPUs). Using an experimental version of the Windows App SDK 2.2, the company is testing the model on GeForce RTX 30‑series GPUs with at least 6 GB of VRAM. The trial enables features such as text summarisation, rewriting, code generation, AI‑based photo editing, super‑resolution and object removal to operate on standard Windows 11 PCs that lack an NPU.

The shift aims to lower hardware costs, broaden the user base for AI‑assisted productivity tools, and reduce dependence on cloud services, improving privacy and response times. If the GPU‑based approach proves successful, the functionality could be rolled out to Windows 11 consumers in the coming months, making advanced AI features available on a much larger range of PCs.