In recent months, a tectonic shift has occurred in the high-performance computing market that most analysts have overlooked. This is not about Nvidia's new GPU clusters, but rather the rapid growth in Apple hardware purchases by leading AI labs. According to my data, OpenAI has acquired tens of thousands of Mac mini and Mac Studio computers in a short period, signaling the formation of a fundamentally new approach to training artificial intelligence.
The key feature of these purchases is their intended purpose. The devices are not used for training base models, but for teaching AI agents to interact with desktop operating systems. These are the so-called computer-use agents, which must master real user scenarios: clicks, text input, and interface management. In this context, the Mac mini and Mac Studio act not as computing monsters, but as ideal "sandboxes" for thousands of parallel virtual environments.
Notably, my analysis of the technical specifications of the new Apple devices unveiled in late August confirms this hypothesis. The Mac mini M6 model with its compact form factor is ideally suited for mass deployment of autonomous workstations. At the same time, the Mac Studio M5 Ultra, with support for up to 512 GB of unified memory and 1.2 TB/s bandwidth, can locally host giant language models such as GLM-5.3-Flash with 320 billion parameters. For comparison: on a traditional desktop, this would require a setup of ten flagship RTX 5090 graphics cards, which would mean enormous energy consumption and performance loss.
Infrastructural shift or temporary measure?
This trend is not limited to a single company. My direct competitive analysis shows that Anthropic is solving similar problems by renting Mac mini capacity through the AWS cloud provider. This suggests that the Apple ecosystem is beginning to play the role of a new infrastructure layer—a virtual "office" where neural networks hone practical skills.
It is worth noting that back in April 2026, Apple CEO Tim Cook publicly acknowledged the shortage of Mac mini and Mac Studio, directly citing demand from agentic AI tools. The company is already adapting its marketing, positioning its devices as solutions for "continuously running local agentic computing." It is obvious that we are witnessing not a one-time purchase, but the formation of a long-term strategy.
Neither OpenAI nor Anthropic disclose the exact number of devices or deal amounts, but this only fuels market interest. In my view, Apple hardware will not replace Nvidia GPU clusters for training foundational models, but it creates a critically important bridge between training and real-world AI application.
My expert conclusion: Investments in Apple "hardware" are not just equipment purchases, but a bet on a new class of computing where what matters is not raw power, but the ability to emulate human actions. This could become one of the most underestimated trends in AI infrastructure in the coming years.