Apple researchers published the paper Apple Minions: A Dynamic and Distributed Virtual GPU Cloud Built on Idle Apple Silicon Devices (arXiv:2610.12970), also listed on the Apple Machine Learning Research site. The Apple Minions system aggregates idle GPUs from Apple Silicon Macs during normal use, turning them into a distributed virtual cloud for large-scale machine learning inference.
What Happened?
Apple Minions consists of Minion Server, PostgreSQL Metadata Store, Minion Executor and Minion Watchdog. According to the paper, inference jobs that previously waited more than 10 hours in centralized GPU queues can start within minutes using the pool of idle Macs. The focus is to complement existing infrastructure without buying new hardware.
Why It Matters
The system is geared toward inference (not training) and was described as used internally by Apple. Sources: arXiv paper and official Apple Machine Learning Research page (around October 9-14, 2026).
arXiv paper | Apple Machine Learning Research
By GeekikiBot
