OpenAI has reportedly made a substantial purchase, acquiring tens of thousands of Mac mini and Mac Studio units to train a sophisticated Computer-use Agent adept at navigating and operating computer systems. Given that macOS virtual machines are exclusively compatible with Apple hardware and are available in limited quantities, training such an Agent necessitates a vast array of authentic Apple devices. In contrast, Anthropic has opted to lease Mac computing resources from Amazon AWS to meet its needs.

Apple's Mac revenue soared to $10.4 billion in the latest fiscal quarter, marking a significant 29% year-over-year surge. Despite this financial triumph, certain popular Mac configurations have remained out of stock for extended periods. The Mac's unified memory architecture is particularly well-suited for running large-scale models, enabling them to operate at peak performance for sustained durations. On some high-end Macs, the text generation speed of locally deployed models is already on par with that of state-of-the-art cloud-based model services, showcasing the Mac's formidable capabilities.

However, Apple has been gradually distancing itself from the enterprise market in recent years. The company lacks a dedicated enterprise customer team and has, instead, outsourced its AI computing power requirements to startups. NVIDIA's DGX Spark is strategically positioning itself to capture this market segment. In response, Apple has preemptively unveiled new Mac mini and Mac Studio models, with a keen focus on facilitating local large model deployment and multi-machine cluster applications, thereby reinforcing its commitment to innovation and versatility in the AI landscape.