And the operating system too, so they can fully leverage all the special features on the chip to improve UX.

Once you have critical mass being able to have enough experienced engineers up and down the whole stack, and you’ve figured out an organizational structure that allows them to work together efficiently, I don’t see how you can beat the vertical integration in a world where the fabs Apple can use are actually better than Intel’s own fab.

Correct me if I’m wrong, but a decade ago Intel would probably have said their fab/process node was their moat?

I know nothing about chip design, but like most things, I feel like one exceptionally brilliant tech lead surrounded by a group of say ~100 merely very smart collaborators can make a world class ARM based processor-design. A mere $100mm annually in hiring and overhead?

The rest of Apple’s advantage comes from being able to actually hire the best, and being their own customer at scale, which means being able to buy first place in line at the fab with billions of dollars of cash.

This is perhaps less true now that the chip has so many specialized areas on the die? Like, does the neural engine get allocated a certain mm^2 and certain number of bus lanes, and then a fully separate team of 100 designs it? I suspect the neural engine part of the chip is actually super simple to design, it’s the tight coupling with the OS and getting apps to properly leverage it which is tricky.