In my experience, to achieve large enough scale most systems have to be used in very simple ways. There isn’t space for additional complexity and the result is usually boring systems that work.
The irony here is most modern CISC design are breaking instructions to RISC-like μOps. Moore's law also means - you have more transistors for the same area, now figure out how to use them creatively to increase performance. Workloads are constantly evolving and hardware evolves with it to make those workloads fast.
Former audio engineer here. Generally this is the result of high noise floor and poor power management on the amplifier side - not powering down when content isn't playing or incorrect power sequencing on the power down of the amp.
This is the key. Look at the roadblocks Google Fiber has hit in their own fiber deployments, pivoting to wireless to bypass the resistance they're getting from incumbents.
x86 is not easily swappable by any means. From voltage supplies/rail timing, crystal requirements, different device drivers and FW rewrites, etc. Changing CPUs on a modern design is a huge reset.
I work in the consumer device space and from what I've seen, the flexibility of in-region supply chain for components and sub-assemblies would be very difficult to match outside of China. It allows for JIT manufacturing and incredibly quick divergences from POR during development when issues occur.
However for more durable goods I could definitely see a resurgence in US manufacturing with the right investments were made in extensive semi to full automation on the design/factory side. China factories are in a tough spot in the regard since employing more people is encouraged by the government, but automation eats away at the low end positions. See Foxconn's 1 Million+ employee count [1] and recently announced slowdown of automation take over [2].