Microsoft is quietly testing a significant expansion of its Copilot+ AI features, according to Tom's Hardware — one that could bring those capabilities to far more Windows users than currently have access to them.
Up to now, Copilot+ features on Windows 11 have required a device with a dedicated Neural Processing Unit, or NPU — specialized AI chips found in newer laptops from manufacturers like Qualcomm, Intel, and AMD. That requirement locked out anyone running an older machine or a desktop PC with a powerful gaming GPU but no built-in NPU.
According to Tom's Hardware, Microsoft is now experimenting with running these AI features on discrete GPUs instead. The option is available through the Windows App SDK, but with a catch: users need to be enrolled in the Windows Insider Experimental Channel and have Developer Mode enabled. That means this is firmly in early-testing territory, not something the average user can simply switch on today.
The move signals that Microsoft wants to broaden the audience for its on-device AI push beyond the narrow slice of consumers who bought a Copilot+ certified PC. Millions of Windows users already own machines with capable discrete graphics cards — particularly gamers and creative professionals — who would gain access to local AI tools without needing new hardware.
It matters because on-device AI, which processes data locally rather than sending it to the cloud, promises faster responses and greater privacy — and unlocking it for GPU-equipped PCs could accelerate mainstream adoption far more quickly than waiting for NPU-equipped devices to become the norm.