I don't think they were caught off guard. It's clear that Apple is the cheapest option, or even the only option in many cases, for running larger local models, especially with the RAM crunch. With the standard "PC design model" you have to buy RAM twice - once for the system to load the GPU and again in the GPU itself. For Apple, and a few other options like DGX Spark and some laptops, you only need to buy it once. I think the Apple/DGX Spark "unified RAM design model" is maybe the only way forward because of RAM pricing/availability over the next few years.