A quiet but landmark transformation is taking place in the world of artificial intelligence. Over the past few months, OpenAI has acquired tens of thousands of Mac mini and Mac Studio computers. This is not just a hardware purchase—it is a strategic move that is reshaping the understanding of how the most advanced AI systems are trained.

The Hidden War for Agentic AI

My analysis shows that this is not about a banal expansion of the equipment fleet. OpenAI is using these machines to train AI agents designed to independently interact with desktop operating systems. This is a fundamentally new approach: instead of abstract text tasks, neural networks are learning to click, type, and manage interfaces in a real environment.

It is telling that OpenAI's direct competitor, Anthropic, is solving a similar problem differently. The company rents Mac mini capacity from Amazon Web Services. This creates an interesting precedent: Apple, while formally not participating in the AI arms race, is becoming a key infrastructure supplier for both sides of the conflict.

Hardware Decides: M5 Ultra vs. M6

The choice of equipment is not accidental. In August, Apple introduced an updated lineup with M5 Ultra and M6 chips, and OpenAI engineers clearly saw new opportunities in this. The division of roles in AI labs is becoming obvious:

  • Mac mini M6. The compact form factor is ideal for deploying thousands of independent working environments. Agents need isolated "sandboxes" for safe training.
  • Mac Studio M5 Ultra. Support for up to 512 GB of unified memory allows running the heaviest language models locally, without relying on cloud clusters.

Take, for example, the flagship Mac Studio M5 Ultra with 512 GB of memory and 1.2 TB/s bandwidth. It can fully accommodate the open LLM GLM-5.3-Flash with 320 billion parameters, which, depending on quantization, takes up to 300 GB of RAM. On a regular desktop, such a volume would require a setup of ten flagship RTX 5090 graphics cards with monstrous power consumption and performance loss.

Apple's Strategic Turn

Notably, back in April 2026, Tim Cook, then Apple's CEO, publicly acknowledged the shortage of Mac mini and Mac Studio, directly citing agentic AI tools as the reason. The company is not just watching what is happening—it is adapting its marketing, positioning new devices as solutions for "continuously running local agentic computing."

Apple hardware is not intended to replace the Nvidia GPU clusters on which base models are trained. Instead, it forms a new infrastructure layer—a virtual "office" where trained neural networks practice performing real tasks. This is a smart move: Apple is carving out a niche that no one else can occupy.

Neither OpenAI nor Anthropic has yet disclosed the exact number of devices or purchase volumes. But the trend is obvious: Apple's ecosystem is becoming an indispensable element in the chain of creating next-generation AI agents.

My expert assessment: this step signals a fundamental shift in the industry. We are moving from the era of giant data centers to distributed systems, where the key resource is not raw computing power but the ability to effectively interact with the real digital world. Apple, without intending to, has found itself at the epicenter of this revolution.