A new Microsoft patent suggests the company could soon let you earn money by using your idle GPU for AI training.
Microsoft recently published a patent called “Opportunistic Micro-training on Edge Devices Using Spare GPU Utilization” on November 5, 2026, describing a new system that breaks down major intensive tasks into smaller ones and sends them to different devices during their idle time. In short, your idle GPU could soon be used for AI training, and Microsoft could pay you to do it.
Microsoft Wants Your GPU
“A system and method for leveraging spare GPU capacity on edge devices through opportunistic micro-training of machine learning models. The system monitors edge devices to detect available compute capacity and decomposes training tasks into independently trainable micro-tasks by analyzing model dependencies and identifying optimal split points based on communication cost, memory overhead, and data locality.”
The system essentially shrinks large tasks designed for big data centers and splits them into micro-tasks that are “dynamically adjusted based on device capabilities” and distributed to devices for “execution during periods of spare capacity.” The patent says this approach is effective and has zero impact on the device’s functionality. However, such tasks are bound to have long-term effects on the GPU. Thankfully, this may not be forced, as users must authorize it.
Microsoft wants to compensate users who lend their devices, so if you opt in, “the system may track the amount of computations” on your device and compensate each “result in a credit being applied” to your account. These credits can be exchanged for free software, model access, money, etc.
One thing to clarify: this isn’t just for Microsoft to advance its own AI ventures. Instead, Microsoft is offering a solution to third-party entities that want to train their models but lack the computational resources to do so. The patent describes how these resources are expensive and consume too many resources, which has led to GPU shortages.
As such, when a person lends their GPU for AI training, Microsoft acts only as a middleman between the third-party entity and the GPU owner. This raises questions about privacy and how much of the GPU must be used. Does the individual give up some degree of permissions when lending their device?
The patent doesn’t answer those questions directly, but it does go in great detail about what constitutes an idle GPU and how the system determines that a device is in this state. It uses several metrics, such as GPU and/or CPU utilization, memory usage, temperature, power consumption, etc.
“The system monitors devices to identify when each of the devices has computational resources available. The system avoids disrupting the user experience such as by slowing down applications, draining batteries, or making devices overheat. Its goal is to utilize spare capacity that would otherwise be wasted… the sample rate is adjusted so that the system checks less frequently when the GPU is busy or hot and more frequently when the GPU is idle and/or the system is cool.”
Source: Patent number US-20260349820-A1 on USPTO
What do you think of Microsoft potentially soon offering money to use your GPU for AI tasks? Let us know your thoughts on the Insider Gaming Discord server.
For more Insider Gaming coverage, check out whether it’s better to buy the Ultimate Edition of GTA 6. And for even more Insider Gaming delivered directly to your inbox, sign up for our newsletter
Mudassir carries years of experience as a journalist and embodies that in his articles. His motto is "All great ideas come while writing." When he's not interviewing someone, he's on…
More from Mudassir Hussain
Comments