While all attention is focused on the race of Nvidia GPU clusters, a new, less noticeable trend is emerging in the AI infrastructure market. 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 could redefine the approach to training AI agents.

Why Mac specifically?

The key task OpenAI is solving is training so-called computer-use agents capable of independently interacting with desktop operating systems. This requires not supercomputers, but thousands of isolated environments where the neural network can learn to click, type, and control the interface through trial and error. It is here that compact Mac minis become the ideal tool for deploying such "virtual offices."

Notably, OpenAI's main competitor, Anthropic, is solving a similar problem differently by renting Mac mini capacity from Amazon Web Services. This confirms that the demand for such configurations is enormous, and Apple has unexpectedly found itself at the center of this process.

Technical foundation

The announcement of the updated lineup based on M5 Ultra and M6 chips (presented on August 25) has only strengthened Apple's position in this segment. An analysis of the specifications reveals a clear division of roles:

  • Mac mini M6. Thanks to its compactness and energy efficiency, these devices are ideal for mass deployment of thousands of independent working environments for AI agents.
  • Mac Studio M5 Ultra. Support for up to 512 GB of unified memory with a bandwidth of 1.2 TB/s allows the heaviest LLMs to be run and hosted locally. For example, the flagship model can entirely fit the open-source GLM-5.3-Flash model with 320 billion parameters (up to 300 GB of RAM depending on quantization). On a traditional desktop, this would require a setup of ten RTX 5090s—with colossal power consumption and performance loss.

A new infrastructure layer

It is noteworthy that back on April 30, 2026, when Apple was led by Tim Cook, the company publicly acknowledged a shortage of Mac mini and Mac Studio, directly linking it to the growing demand for agentic AI tools. The marketing of the new devices has already been adapted to this scenario—they are positioned as solutions for "continuously running local agentic computing."

Of course, Apple hardware will not replace Nvidia GPU clusters for training base models. But it forms a fundamentally new infrastructure layer—a virtual environment where already trained neural networks practice solving real-world tasks. Neither OpenAI nor Anthropic discloses the exact number of purchased devices or deal amounts, which only fuels interest in this story.

My view: this is a classic example of how the market underestimates non-obvious assets. While everyone is looking at data centers, Apple is quietly capturing the niche of "training gyms" for AI. Investors should take a closer look at the Apple ecosystem not as a consumer brand, but as a serious infrastructure player of a new kind.