If the clock speed improvements had happened over a much longer stretch of time then we probably would have seen much earlier multi-core capable tooling. We are still mostly optimized for single threaded applications, extracting the maximum from a CPU is really hard work. So also consider the backlog in tooling, there is so much work that still needs to be done there.

And so much of our tech stack runs on old abstractions.

Do you even need an MMU, if you have memory safe languages?

Honestly, maybe semi scaling is a problem for AI, but for most other software today, the problem is bugs, bloat, latency.

MMUs allow for overcommit and make it possible to dynamically allocate memory to applications that need it.