Physix Frontier badgePhysix Frontier
PublishChinese | EN

Physix Frontier · News Briefing Card (IT Home · Oct 16, 2026)

Apple Research Pools Idle Mac GPUs for AI Inference

KEY FACTS

  • Apple's research team unveiled a system called Minions that aggregates idle GPUs across multiple Macs.
  • The system consists of four parts: a server, metadata storage, executors, and a watchdog.
  • Devices use a pull-based model to actively claim subtasks, avoiding duplicate processing of the same task.
  • In tests, inference tasks that originally queued for over 10 hours began executing within minutes under Minions.
  • The approach requires no new GPU hardware and can lower infrastructure costs.

KEY DATA

over 10 hoursOriginal queue time
minutesTime to start execution under Minions

PHYSIX OBSERVATION

Assembling idle GPUs from office Macs into a compute pool is an approach that uses existing devices to bypass GPU procurement and queueing bottlenecks, and it is quite attractive to budget-constrained enterprises. But cross-device scheduling involves data confidentiality and governance, making it hard to become a consumer-grade product in the short term; it looks more like an internal cost-reduction solution.

Source: IT Home report