Comment by ellis0n

Exactly. Chips are incredibly expensive and extremely complex—containing billions of transistors and you can't realistically build them at home. Memory is a good example: it keeps growing in both capacity and cost.

The same thing will happen to software. Eventually, the sheer volume and diversity of code will exceed human comprehension and only massive data centers will be able to process it. We'll reach the point where software complexity and programmer capabilities hit an economically practical limit. The next-gen Moore's Law for AI?

Today, chips are designed largely through automated processes and software is increasingly being created the same way. The main difference is the timescale: software evolves billions of times faster than hardware.