The once-overlooked Mac mini has emerged as a critical player in the AI space due to OpenClaw, transforming it from an under-the-radar desktop into a sought-after machine. On a recent earnings call, Tim Cook disclosed that both the Mac mini and Mac Studio are sold out, potentially remaining unavailable for months. He highlighted their popularity among developers for AI applications, surpassing initial forecasts.
Quarterly Mac revenue reached $8.4 billion, marking a 6% increase from the previous year. However, supply limitations rather than demand are constraining growth. High-RAM configurations of both models have been delayed or removed entirely due to these constraints, with some offerings completely disappearing from Apple’s inventory.
In the U.S., the base $599 Mac mini is out of stock with no immediate availability for delivery or pickup. Upgraded versions with 64GB RAM face wait times of up to 18 weeks, while certain Mac Studio models have vanished entirely. Scalpers on platforms like eBay are capitalizing on this scarcity, listing base units at nearly double their retail price.
The surge in demand is largely driven by OpenClaw—an open-source AI framework supported by Peter Steinberger and now backed by OpenAI after competing with Meta—garnering over 323,000 GitHub stars. It offers a seamless way for individuals and small teams to run persistent AI agents locally, making the Mac mini an unofficial standard for its execution.
Apple’s relevance in serious AI workloads has been minimal for years due to performance issues when running large language models or other home AI software. An M2 Mac, for instance, paralleled a 2019-era GPU in terms of speed. Apple’s reluctance to adopt CUDA or integrate Nvidia technology further marginalized its role in AI and gaming.
However, Nvidia’s proprietary CUDA framework has a significant limitation: VRAM constraints. The top consumer Nvidia GPU, the RTX 5090, maxes out at 32GB of VRAM, limiting its ability to handle larger models efficiently. In contrast, Apple’s Unified Memory Architecture (UMA) allows shared RAM among the CPU, GPU, and Neural Engine, eliminating the need for separate VRAM or PCIe bus traversal. A Mac mini with 64GB can accommodate a 70 billion-parameter model that an RTX 5090 cannot manage.
Apple Silicon chips, such as the M4 Ultra in high-end Mac Studio models, support up to 192GB of unified memory, enabling them to run extremely large models locally without additional infrastructure. OpenClaw highlighted these advantages by requiring local agent execution, which necessitated powerful, yet affordable, machines capable of handling complex computations independently.
This shift led developers to purchase Mac minis in bulk, akin to buying Raspberry Pis for infrastructural use—a pattern Apple’s supply chain was unprepared for. The global memory chip shortage, exacerbated by AI server demand, further strains availability. IDC predicts an 11.3% decline in global PC shipments in 2026 due to these shortages.
Cook anticipates a few months to balance the Mac mini and Studio’s supply and demand. Although an M5 chip refresh later in 2026 might alleviate some pressure, current buyers face long waits or inflated prices from scalpers. Remarkably, it took an open-source initiative outside of Apple’s direct influence to catapult its Mac mini to unprecedented popularity.