OpenAI has quietly assembled tens of thousands of Apple Mac mini and Mac Studio systems for reinforcement learning and computer-use agent training, a buildup also highlighted by market commentator Serenity, though neither OpenAI nor Apple has publicly confirmed the arrangement. The machines suit multi-step interface work—navigating software, drafting documents, and organizing email—because Apple’s unified memory pools RAM across CPU and GPU, a better fit for memory-bound, parallelism-light agent loops than traditional GPU clusters built for large-scale pre-training. Anthropic has taken a similar path by renting Mac mini capacity through Amazon Web Services and has begun cooperating with Raspberry Pi on new model hardware standards, with no reports yet of large-scale Raspberry Pi purchases by OpenAI or Anthropic for the same tasks. Surging demand has stretched delivery times for high-RAM configurations to weeks or months, and some customers have already shifted to Nvidia DGX Spark, raising questions about how long Apple’s edge in this niche will last. Apple refreshed both Mac lines ahead of schedule on August 25, 2026, adding M6 and M5 Pro options on the Mac mini and M5 Max and M5 Ultra on the Mac Studio, while Mac revenue reached $10.4 billion in the most recent quarter, up 29% year over year.