Contrary to the prevailing belief that the future of artificial intelligence belongs exclusively to Nvidia GPU clusters, a new, unexpected infrastructure trend is emerging in the market. Over the past few months, OpenAI has acquired tens of thousands of Mac mini and Mac Studio computers. This is not a spontaneous purchase of office equipment, but a strategic move aimed at solving highly specialized tasks in the development of agentic AI.

This concerns training AI agents (computer-use agents) to independently interact with desktop operating systems. From my analytical perspective, it is obvious: for a neural network to learn how to click, type, and manage an interface, it needs not just computing power, but access to thousands of isolated, actually running systems. It is here that Apple, with its ecosystem and unified architecture, offers an elegant solution.

Notably, OpenAI's main competitor, Anthropic, is solving the same problem in a different way, renting Mac mini capacity from Amazon Web Services. This confirms that demand for such configurations is enormous, while supply from cloud providers is still limited. Incidentally, on August 25, Apple unveiled an updated lineup based on the M5 Ultra and M6 chips, which fits perfectly into this strategy.

Division of Roles: Mac mini vs. Mac Studio

An analysis of the technical specifications of the new products reveals a clear division of labor within AI laboratories:

  • Mac mini M6. The compact design allows for deploying thousands of independent working environments. This is an ideal "training ground" for agents that need a real OS to practice interface interaction.
  • Mac Studio M5 Ultra. Support for up to 512 GB of unified memory is a critical resource for locally hosting the heaviest language models. For instance, the GLM-5.3-Flash model with 320 billion parameters can take up to 300 GB of RAM depending on quantization, and the Mac Studio is capable of fitting it entirely.

For comparison: to run such a model on a traditional desktop, a setup of ten flagship RTX 5090s would be required, with enormous power consumption and a bandwidth bottleneck. Apple bypasses this limitation through unified memory with bandwidth of up to 1.2 TB/s.

Notably, as early as April 30, 2026, Tim Cook publicly acknowledged the shortage of Mac mini and Mac Studio, directly linking it to the boom in agentic AI tools. The Cupertino company is already adapting its marketing, positioning the devices as solutions for "continuously running local agentic computing."

Of course, Apple hardware will not replace Nvidia GPU clusters, on which base models are trained. However, it is creating a new virtual "office" where trained neural networks practice solving real-world tasks. This turns Apple into a full-fledged infrastructure player, rather than just a consumer electronics manufacturer.

My conclusion: the AI infrastructure market is fragmenting. While everyone watches the graphics card race, demand is quietly building for specialized platforms for the final stage of training—and here, Apple's ecosystem proves unrivaled in terms of price/performance per watt.