Apple has developed a system called Minions that pools unused Mac GPU performance for AI tasks, Techzine reported on Oct. 14 local time.

According to the report, six Apple researchers described in a paper how they combined the computing power of multiple Macs to create a virtual cloud.

Minions redirects idle GPU capacity on Macs to AI workloads. It helps companies use their hardware more efficiently and reduce reliance on dedicated GPU servers.

Minions searches for available computing resources and automatically distributes tasks across multiple devices. It also takes into account situations in which some Macs are temporarily offline or already running at maximum performance, Techzine said.

According to Apple, Minions is used mainly for AI inference. Unlike large language model training, inference workloads can largely be handled independently, making them well suited to distributed environments, Techzine said.

The researchers said multiple teams inside Apple already use Minions.

Minions does not turn multiple Macs into a single powerful GPU that shares memory. It distributes independent tasks across devices.

Apple is also studying whether the technology can be used for AI model training.

The researchers said data security is a challenge to be addressed because running workloads on Macs used by other teams could expose sensitive information, Techzine said.