In short: The AI boom needs every free gigabyte of RAM and GPU cycle it can get. Just days after OpenAI CEO Sam Altman controversially complained about memory going unused in consumer devices, Microsoft filed a patent describing how it could allow users to train AI on their idle devices.
The patent, filed earlier this month and quickly noted by Patentlyze, would monitor consenting users’ devices and assign AI training workloads to them during periods of low or no use. As compensation, users may receive credit or expanded access to AI models.
As the patent explains, training AI models requires massive computing power, and GPUs are uniquely suited to these tasks. In addition, the computational effort for training models increases with their complexity.
These costs also strain GPU production capacity. Memory shortages due to AI data center expansion aren’t the only reason consumer GPUs are becoming unaffordable. The situation has also made server GPUs a more lucrative wafer use for GPU makers than RTX 5090 manufacturing. AMD CEO Lisa Su recently described the GPU maker as a data center-focused company.
The RAM market is experiencing an even more extreme version of the problem, leading to rising prices for various consumer devices. Production capacity is becoming so scarce that Sam Altman recently claimed that RAM is more useful in AI companies’ servers than in home devices, where it often goes unused. The comments were met with widespread outrage. Microsoft’s patent would reclaim some of that unused GPU capacity.
The first diagram shows a training network made up of servers, desktop PCs and smartphones, suggesting that Microsoft’s system could also access idle mobile devices.
Microsoft would monitor their CPU and GPU utilization, memory usage, temperature, power consumption, memory bandwidth, and compute queue depth to determine how much they can save and when.
Training workloads would be divided into microtasks and assigned to subscribed devices based on their availability and reputation for accurate results. Users who agree to loan their spare bikes for AI training may receive financial compensation, software credits, or access to the models they train.
It remains unclear whether this patent will see the light of day, but AI-related patents have become a trend in the years since the boom began. A recent release from Microsoft suggests editing live streams in real time to rewrite or translate words while adjusting speakers’ lip movements.
