What's this about?
Apple was reportedly unprepared for how much demand for the Mac Mini and Mac Studio would come from people running AI models locally. 349 comments, and most of them are about memory prices.
Why are Macs the local-AI machine?
Unified memory. In a normal PC the GPU has its own VRAM, typically 8–24GB, and a model larger than that cannot run on it. Apple Silicon shares one pool between CPU and GPU, so a machine configured with 128GB or more can hold models that would need several data-centre GPUs. It was designed for video work and turned out to be the cheapest way for an individual to run a large model at home.
What's the community saying?
HN 302 points, 349 comments
- The dominant complaint is price, and the immediate correction is that this is not Apple-specific: RAM is the new gold, prices are up across the board, and Valve raised the price of a four-year-old Steam Deck for the same reason.
- The most-challenged claim in the thread: that a non-insignificant proportion of people buying these for AI have become multi-millionaires from AI. Two separate replies ask for a single name.
- A structural suggestion that gets support: Apple designed its own silicon and its own modem after acquisitions, and could do the same for memory.
- "Time to bring back the Xserve" is the top comment, and only half a joke.