Microsoft patent describes paying users to train AI on idle GPUs

A Microsoft patent application published on November 5 describes a system that would allow users to opt in to run small AI training tasks on spare GPU and CPU capacity. The filing, titled "Opportunistic Micro-training on Edge Devices Using Spare GPU Utilization," was submitted in April 2025.

Image: TechPowerUp

Microsoft explains that the system would split large AI training workloads into independent micro-tasks sized to what a device can handle. These tasks would be sent to phones, laptops, and gaming PCs during periods of low or no usage. To ensure accuracy, the system can send identical calculations to multiple devices to check for errors or tampered data.

The system would monitor GPU and CPU load, memory use, temperature, and power draw to detect spare capacity. Microsoft says tasks can be paused or stopped if a machine heats up or if the user launches demanding software like a game.

Users would need to explicitly opt in to participate. Microsoft says potential rewards for this work could include money, software, or access to trained AI models. The patent does not list specific supported hardware or detail how payout rates would be calculated. There is no guarantee Microsoft will turn this patent into a commercial product.