Microsoft has filed a patent for a system that utilizes idle computing power on consumer devices to train artificial intelligence models. The application, titled Opportunistic Micro-training on Edge Devices Using Spare GPU Utilization, was submitted in April 2025 and published on November 5 under the identifier US-20260349820-A1.
The proposed technology divides complex AI training tasks into smaller micro-tasks tailored to the specific capabilities of individual devices, including gaming PCs, laptops, and smartphones. Users would need to provide explicit consent to participate in the network, which monitors hardware metrics such as CPU and GPU load, memory usage, power consumption, and thermal levels to ensure performance is not impacted.
Participants could receive compensation in the form of credits or cash for their contribution. These credits would be redeemable for software, access to trained models, or direct payments. The system includes verification mechanisms to cross-check results across multiple devices to prevent errors or tampering.
While the patent outlines the technical framework for distributed training, it does not specify payout rates or supported hardware configurations. Microsoft has not announced any official service for Windows or Xbox platforms, and the company has not confirmed if this concept will be developed into a commercial product.
