OpenAI and other AI labs have purchased tens of thousands of Mac minis and Mac Studios to train computer-use agents that handle multi-step tasks on their own. The buying spree is fueling a sharp lift for Apple’s Mac business, which jumped nearly 29% to $10.4 billion in the June quarter, driven largely by demand from AI developers and researchers.<\/p>\n
OpenAI is using the machines to train agents designed to navigate software autonomously and complete chained actions without step-by-step human input. Other labs are following the same playbook, drawn to the combination of strong Apple silicon performance, shared unified memory, and the thermal headroom needed for long-running AI workloads.<\/p>\n
The appetite outstrips supply. The most powerful Mac configurations have been sold out for months, with the bottleneck traced to a memory chip shortage rather than to Apple’s assembly lines. OpenAI wants additional units, but waits for inventory.<\/p>\n
Anthropic is also in the mix, renting Mac minis through Amazon Web Services rather than buying the hardware outright. The approach gives the lab elastic capacity without the capital expense of a fleet purchase, and it sidesteps the long lead times on premium configurations.<\/p>\n
Outside the major labs, the Mac mini is gaining traction as a local AI computer. A wave of open-source tooling has made it easier to run models at home or in small clusters. The open-source software Exo, for example, links multiple Macs into a cluster so users can run large models locally. Combined with Apple’s strong chips, unified memory, and solid cooling, the platform has become a popular choice for AI hobbyists and independent researchers.<\/p>\n
Nvidia offers a competing compact workstation, the DGX Spark, aimed at the same audience. The two platforms take different routes: Apple’s hardware leans on unified memory that the CPU and GPU share, while the DGX Spark depends on dedicated GPU power through CUDA and Tensor cores. Early reviews of the Spark suggest Nvidia has found another path to move its chips into smaller form factors, though the unified-memory design remains the main reason developers reach for a Mac when they need a single machine to handle both training and inference on large models.<\/p>\n
The shift signals a quiet split in how labs think about small-form-factor AI hardware. Large-scale training still runs on GPU clusters in data centers, but agent training and local experimentation increasingly favor machines with generous unified memory and the ability to run for hours without thermal throttling. Apple’s quarterly revenue jump suggests the pattern is now visible on the company’s books, and not just in the purchasing behavior of a few labs.<\/p>\n
For developers, the practical takeaway is that the Mac mini and Mac Studio have moved from general-purpose workstations to recognized infrastructure for AI work. Memory supply, not chip design, is the rate-limiting factor, and that constraint is shaping how quickly labs can expand their agent-training pipelines.<\/p>\n
OpenAI and rival labs are using Mac minis and Mac Studios to train computer-use agents that autonomously handle multi-step tasks. The hardware offers strong chips, unified memory, and solid cooling suited for long-running AI workloads.<\/p>\n
Apple’s Mac revenue jumped nearly 29% to $10.4 billion in the June quarter, with the increase driven in part by demand from AI labs and developers.<\/p>\n
The DGX Spark relies on dedicated GPU power through CUDA and Tensor cores, while the Mac mini uses Apple’s unified memory architecture shared across the CPU and GPU. Both target compact AI workloads but take different design approaches.<\/p>\n