In brief
- Apple CEO Tim Cook cautioned that Mac mini and Mac Studio units could face ongoing shortages for “several months,” as AI-fueled demand dramatically outpaced the company’s projections.
- OpenClaw—the open-source AI agent platform now supported by OpenAI—made Apple’s unified memory architecture the go-to hardware choice for running large AI models locally.
- Apple’s M4 Ultra chip supports up to 192GB of unified memory, enabling developers to run models that exceed the capacity of any single consumer Nvidia GPU, which tops out at 32GB of VRAM.
The Mac mini has long been the unassuming, easy-to-overlook desktop tucked away in the corner of the Apple Store. Affordable by Apple’s standards, practical, and largely passed over by the AI community—until OpenClaw changed everything.
On Thursday, Tim Cook informed analysts that both the Mac mini and Mac Studio are completely sold out—and may remain unavailable for months. “Both of these are outstanding platforms for AI and agentic tools,” he stated during Apple’s Q2 2026 earnings call, “and customers are recognizing that far more quickly than we anticipated.”
In other words: Apple underestimated just how urgently developers would seek out these machines, particularly during a period when market-wide scarcity is already disrupting supply chains.
Mac revenue reached $8.4 billion for the quarter, reflecting a 6% increase year-over-year. Not exactly a blockbuster figure—but the bottleneck is supply, not demand. High-memory Mac mini and Mac Studio configurations aren’t merely backordered; some have been entirely removed from the Apple Store.
The entry-level $599 Mac mini is sold out across the U.S., with neither delivery nor in-store pickup options available. Upgraded models equipped with 64GB of RAM are showing estimated wait times of 16 to 18 weeks. Mac Studio units configured with 512GB of unified memory have vanished from the store entirely. eBay scalpers were quick to capitalize, listing base models at nearly twice the retail price.
So what triggered all of this? OpenClaw and the surge of memory-intensive agentic AI.
The open-source AI agent framework—created by Peter Steinberger and now backed by OpenAI following a competitive bidding war with Meta—skyrocketed to over 323,000 GitHub stars, becoming the fastest path for individuals and small teams to deploy persistent AI agents on their own hardware. And the unofficial reference platform for running it became, almost overnight, the Mac mini.
This wasn’t driven by any marketing campaign, however.
What most coverage of the Mac shortage overlooks is that Apple was sidelined from serious AI workloads for years. Before the AI agent revolution went mainstream, critics frequently pointed out that
Running large language models, Stable Diffusion, or any other kind of home AI software used to be painfully slow and nearly impossible to use. An M2 Mac performed about as well as a 2019-era GPU. Apple’s decision to avoid CUDA and Nvidia hardware, instead pushing its own MLX framework, left it just as sidelined in the AI world as it had been in gaming.
Nvidia dominated because CUDA—its proprietary GPU programming platform—became the foundation for both training and running AI models. The entire AI ecosystem was constructed on top of it. Apple had no equivalent. Nobody considered a Mac a viable option for local AI inference.
But CUDA has a well-known limitation: VRAM capacity.
Even Nvidia’s top consumer card, the RTX 5090, maxes out at 32GB of VRAM. That’s a hard wall. Any model exceeding 32GB can’t run at full speed on that GPU—it overflows into slower system RAM, gets bottlenecked by the PCIe bus, and performance collapses. Running a serious 70-billion-parameter model on Nvidia hardware means stacking multiple GPUs, setting up a server rack, drawing serious power, and spending thousands of dollars.
Apple’s Unified Memory Architecture (UMA) addresses this in a way CUDA simply can’t. On Apple Silicon, the CPU, GPU, and Neural Engine all draw from the same shared pool of RAM. There’s no dedicated VRAM. There’s no PCIe bottleneck. A Mac mini with 64GB of unified memory can load a 70-billion-parameter model that a $1,800 RTX 5090 can’t even begin to handle.
The M4 Ultra—the chip inside high-end Mac Studio models—supports up to 192GB of unified memory. That’s enough to run 100-billion-parameter models entirely on a single desktop machine. No server rack. No recurring cloud costs.
OpenClaw made this advantage crystal clear. Because it runs AI agents locally—tapping into your files, your applications, your messages—users needed hardware that could handle the computational load without relying on cloud compute. A Mac mini with 32GB of unified memory runs 30-billion-parameter models with ease. A Mac Studio with 128GB can handle models that most developers couldn’t access without an enterprise GPU cluster just a year ago.
A slower Mac that can actually run a powerful AI model is far more useful than a powerful Nvidia GPU that can’t even load the same model in the first place.
The outcome: developers began purchasing Mac minis the way they once bought Raspberry Pis—in multiples, treating them as infrastructure rather than personal machines. Apple’s supply chain was never built to handle that kind of demand.
There’s also a wider memory crunch making things worse. IDC projects global PC shipments will drop 11.3% in 2026, partly due to a memory chip shortage driven by AI server demand. Apple is now competing for the same RAM supply as hyperscale data center builders.
Cook said it could take “several months” to bring supply and demand back into alignment for the Mac mini and Studio. An M5 chip refresh is anticipated later in 2026, which might relieve some of the pressure—but for now, buyers face long wait times or inflated reseller prices.
The Mac mini generated more demand in 2026 than at any point in its two-decade history—and all it took was a boost from an open-source project that Apple had absolutely no hand in creating.
Daily Debrief Newsletter
Start every day with the top news stories right now, plus original features, a podcast, videos and more.
