Not really a contradiction. Some things don’t scale up well for mass production. That’s why we need a 5-axis CNC with a metal powder/laser material depositing head (and plastics-depositing heads), inter alia, in every garage.
Okay, every other garage. Let’s not ask for extremes.
So if you want 10,000 pieces machined, you can either buy a factory for the CNC machines like Apple did, or do one piece in each of the 10,000 garages of people who want the thing. Both approaches work.
…then there’s the third approach, do it some other way. Nobody says it has to be milled from one piece, you can braze it from several. Or injection-mould. Or hydroform from tubes, and it will be a bit bigger but also lighter and stiffer and if you coat the inner sides with sintered metal it can also act as a heat pipe system for cooling. Apple tech is sexy but also more often than not also cumbersome to not only repair but also to make. Learn from Russian aerospace tech, instead. And if you cannot achieve the result you want, you either overspecified it or should ask if you can make the tools/jigs if you cannot make the result directly. Cannot print from metal? Print from PLA and then cast the metal.
…and the fourth one, send the file to Shapeway or their ilk and get it done there.
Take 10-20% of the IT tech development we got in last few decades and put it into the machining segment, and you’re pretty much there. It’s a rather big change of paradigm but bigger ones sneaked up to us (intercontinental air transport, multimodal global logistics, cars, computers, Internet-everywhere, laser and photo-quality inkjet printers, cellphones…) and not only we did not notice them biting our asses, we barely noticed the stink when we got crapped out.
This, just on cue. A TinyG CNC controller board. What I miss in the developments are some feedback systems for precision position sensing, preferably optical (there are some hacks using digital calipers but I consider them rather dirty). (Todo: try out with an optical mouse.)