In recent months, a tectonic shift has occurred in the AI infrastructure market that most analysts have overlooked. OpenAI, one of the leaders in the artificial intelligence race, has made a massive purchase of tens of thousands of Mac mini and Mac Studio computers. This is not just an upgrade of its hardware fleet—it is a strategic maneuver that fundamentally changes the landscape of training agentic models.

The new role of Apple hardware in AI labs

Instead of relying solely on giant Nvidia GPU clusters, OpenAI is using Mac computers to train AI agents designed to interact with desktop operating systems. These are the so-called computer-use agents—systems that mimic human actions: clicking, typing, and managing interfaces. This is a fundamentally different task from training base models, and it requires not raw computing power but a huge number of isolated environments.

Notably, OpenAI's main competitor, Anthropic, has chosen a different path by renting Mac mini capacity from Amazon Web Services. This confirms that Apple hardware is becoming a critically important resource for developing autonomous AI, one that is more advantageous to rent than to buy.

Why Mac: a technical analysis

The recently unveiled Apple lineup based on the M5 Ultra and M6 chips clearly demonstrates the division of roles in AI infrastructure:

  • Mac mini M6. The compactness of these devices allows for deploying thousands of independent virtual environments. For training agents, a realistic environment is critical, not just mathematical computations.
  • Mac Studio M5 Ultra. Support for up to 512 GB of unified memory is a key advantage. For example, the flagship model can fit the open-source LLM GLM-5.3-Flash with 320 billion parameters, which takes up to 300 GB, entirely in RAM. On a regular PC, this would require a setup of ten RTX 5090 graphics cards with monstrous power consumption and performance loss.

Indirect confirmations and strategic significance

Back in April 2026, Tim Cook, then CEO of Apple, publicly acknowledged the shortage of Mac mini and Mac Studio, directly citing agentic AI tools as the reason. The company's marketing now positions these devices as solutions for "always-on local agentic computing." This is not just words—it is the formation of a new infrastructure layer.

Apple hardware will not replace GPU clusters for training base models, but it creates a virtual "office" where trained neural networks practice on real-world tasks. This is a smart move: instead of investing in energy-intensive data centers, labs gain an efficient environment for the final refinement of agents. Neither OpenAI nor Anthropic discloses the details of their purchases, but the very fact of such interest suggests that we are on the threshold of a new era where AI infrastructure is not only GPUs but also intelligent, specialized computers.

My view: this trend is undervalued by the market. If agentic AI becomes a mass-market product, Apple could transform into a key infrastructure provider, creating a new revenue stream independent of smartphone sales. Investors should closely watch this dynamic.