Same as for a thousand other fields essential to operating the modern world. Nobody has time to learn them all, so we specialize.
They're usually very hard to get share because machine manufacturers can smash out cheaper things via processes like castings, mouldings and stampings, then eventually lock down spares (or just don't bother).
The open source option basically only be worse (but maybe more repairable) and/or more expensive than the alternatives, except when there is no alternative in the market. And China is providing so much mid-grade affordable and fairly functional stuff there often is an alternative even in the most isolated places. In 1980, getting a decent lathe in some town in, say, Angola might have been basically impossible. Now, it's still not cheap, but it's not completely impractical. If you can get bearings and induction-hardened shafts you'd need to DIY, you can get the whole thing, and maybe even cheaper.
It's a bit depressing, because of course I want to see the world flooded with high-quality, modular, very standardised, re-usable, repairable, hackable items, but that approach has a limited market in reality.
Notably lacking from their toolkit is anything large (no refineries, no blast furnaces, no glassworks for making window panes, etc) or anything needing high precision or high purity (medicine, ball bearings, optics, high quality metals, etc). It still assumes the rest of society will be around to source those materials from.
The GVCS is like if FOSS only ever produced leftpad libraries and never a linux or a postgres.
A 500 million glassworks or foundry is just that times a million (literally). And technology has never stopped - no one would spend a hundreds of millions to build a new plant on a 20-year old design.