Apple began shipping its new Mac mini and Mac Studio desktop computers on Tuesday and is marketing the machines to corporate customers as a lower-cost alternative to renting data center capacity. The pitch centers on local AI: companies can run models on their own hardware rather than paying per-token fees to cloud providers such as OpenAI and Anthropic. Where cloud billing meters each request a model processes, owning the hardware replaces recurring usage fees with an upfront purchase.

A four-Mac cluster runs a trillion-parameter model off one outlet

Apple is aiming the desktops at large AI workloads, such as writing code and handling intricate business tasks. The top configurations cost about $20,000.

“There’s no cost per token. You’re just using the machine again and again,” said Johny Srouji, Apple’s chief hardware officer.

Apple demonstrated a setup of four Mac Studios connected together to run a trillion-parameter AI model that found and fixed a graphics coding bug — a job that would typically require a data center. The cluster runs off a single wall outlet.

The machines link over Thunderbolt 5 with remote direct memory access, which Apple claims enables distributed AI inference up to three times faster than a single machine. RDMA lets linked machines read each other’s memory directly, trimming the overhead of moving data between computers in a cluster.

The M5 Ultra Mac Studio can be configured with up to 512GB of unified memory, though that configuration will not ship until late October. Unified memory pools the RAM that a Mac’s processors share, a key factor in how large a model a single machine can hold. Apple says the Mac mini with M6 starts at $899, and the Mac Studio with M5 Ultra at $5,499. In August, Cryptopolitan reported that the entry-level Mac mini’s price rose by $100, an increase Apple attributed to memory costs.

Srouji added that the chips share common designs, so AI models built on Apple’s machines scale from the most expensive Mac Studios down to iPhones and iPads.

Windows holds 91.3% of the enterprise market

Apple faces a steep climb. It holds roughly a 4.6% share of the corporate desktop and laptop market, compared with 91.3% for Windows.

Microsoft is courting the same buyers with on-device AI. CEO Satya Nadella has said the company plans to merge many of its AI features into a single super app for Windows. Nvidia and PC makers are rolling out new AI desktops, expected to be the headline announcement at a Microsoft Windows event in San Francisco next month. With Apple’s machines now shipping and the Windows event a month away, the two companies’ local-AI pitches are set to reach enterprise buyers in quick succession.

Microsoft said it has worked with chip partners to simplify AI work through its Windows ML tools and that RDMA and other features are an area of active investment. Nvidia CEO Jensen Huang downplayed any suggestion the company would go head-to-head with Apple at the launch of its new PC chip this summer.

Apple’s opening traces back to 2020, when the company moved to Apple Silicon, combining computing and memory in a single design to save battery power. That architecture made Macs well suited to AI, and Mac minis sold out as the open-source agent tool OpenClaw spread in markets such as China.

Sources: Reuters, Apple Newsroom, Cryptopolitan