OpenAI Buys Tens of Thousands of Apple Macs to Train Computer-Operating AI Agents
OpenAI has reportedly purchased tens of thousands of Apple Mac mini and Mac Studio computers equipped with large unified memory. These devices are being used to run reinforcement learning workloads to train AI agents capable of operating computers. This highlights a growing industry trend where AI developers leverage Apple Silicon's unified memory architecture as a cost-effective alternative for specific AI workloads. The surge in demand from AI companies has even led to supply constraints for Apple's desktop Macs. Rather than large-scale pre-training, these Macs provide a realistic desktop environment for AI agents to practice navigating software interfaces and interacting with applications. Other AI firms, such as Anthropic, are also utilizing Mac hardware by renting Mac mini compute power through AWS.
## BACKGROUND
AI agents designed for computer control are systems trained to interact directly with operating systems and applications just like a human user. Apple's unified memory architecture allows both the CPU and GPU to access a single, high-bandwidth pool of memory, making it highly efficient for running large models locally without the overhead of transferring data between system RAM and dedicated VRAM.