Apple Ships High-End Mac Mini and Mac Studio for Enterprise AI, Touting Zero Per-Token Cost

Apple's new Mac mini and Mac Studio push the company's silicon into enterprise AI territory, with high-end configurations priced up to nearly $20,000. The pitch, delivered pointedly by hardware chief Johny Srouji, is economic: 'there's no cost per token.' In a demonstration, Apple ran a trillion-parameter model across four networked Mac Studios powered from a single wall outlet, emphasizing power efficiency and a one-time capital cost versus recurring cloud metering.

The systems run macOS 27 ('Golden Gate') and feature Siri AI — the substantially rebuilt, personal-context-aware assistant Apple detailed earlier in September, drawing on messages, emails, and photos with onscreen awareness. The strategic framing puts Apple in direct rhetorical conflict with NVIDIA (whose GPUs dominate AI compute) and Microsoft (whose Azure cloud bills per token), betting that a segment of enterprises are tired of unpredictable inference bills.

Skeptics note the roughly $20,000 entry price limits near-term adoption, and enterprise AI at scale still favors cloud elasticity for spiky workloads. On-device also caps model size and update cadence versus frontier cloud models. Separately, a settlement website went live for a class action alleging false advertising over Apple's long-delayed personalized Siri — a reminder that Apple's AI narrative has trailed rivals and that its credibility on shipping is under scrutiny. Reaction among enterprise developers was warm on the cost message but cautious on practicality; the 'no cost per token' line resonated with teams weary of cloud billing, even as many questioned whether the math holds outside steady, predictable inference loads.