Microsoft tests Copilot+ AI with discrete GPUs, bypassing NPU requirement

Microsoft is testing Copilot+ AI features with discrete GPUs as an alternative to NPU-based inference on Windows, available to developers through the Windows App SDK with Insider Experimental Channel builds and Developer Mode enabled. The move expands compatibility for enterprise deployments that rely on existing discrete GPU infrastructure.

The testing signals flexibility in Copilot+ deployment paths and acknowledges that discrete GPUs remain prevalent in enterprise data centers and workstations, particularly where specialized inference or mixed workloads demand higher parallelism.