OpenAI, Anthropic, and other leading AI laboratories have purchased tens of thousands of Apple Mac Mini and Mac Studio desktops to support the training of autonomous computer-use agents, according to a report from The Information. High-end Mac desktop configurations are currently sold out across several retail channels due to supply constraints linked to memory chip shortages.
The labs are leveraging Apple Silicon’s unified memory architecture, high memory bandwidth, and thermal performance to execute long-running local control workloads. While Anthropic is renting Mac Minis through Amazon Web Services, other developers are utilizing open-source tools like Exo to link individual Mac units into local distributed clusters for inference and training.
The surge in enterprise demand for compact AI hardware contributed to a nearly 29 percent year-over-year increase in Apple’s Mac revenue, which reached $10.4 billion in the June quarter. Specialized infrastructure startups are also emerging to serve this market, including Mount Thor, an Apple-based cloud hosting service founded by a former OpenAI infrastructure engineer.
Why it matters
Unified memory architectures are proving highly cost-effective for agentic hardware setups, creating non-traditional server demand for consumer silicon.
Hardware shortages are expanding beyond enterprise GPUs to consumer-grade unified memory systems, bottlenecking agent research.
Infrastructure providers have an opportunity to build dedicated Apple Silicon cloud clusters to support specialized desktop automation workloads.
Source: the-decoder.com
