Turning idle Macs into GPU power
Apple Machine Learning Research introduced Minions, a dynamic and distributed virtual GPU cloud. The system aggregates the GPUs of Apple Silicon–based macOS devices, which remain largely idle during normal usage.
Why it matters
According to the research, as demand for GPU-accelerated machine learning inference grows, the limits of centralized GPU infrastructure become clear: high cost, limited elasticity and long wait times under resource constraints.
What it does
Minions pools idle Apple Silicon GPUs to execute large-scale machine learning inference workloads, aiming to let teams leverage the device capacity they already own.
This is not a product launch but a research paper outlining a system. Full details and performance results are available in the paper.
