Interestingly to your example, there seems to be a rough consensus that the bronze age collapse (~12th century BC) was related to the fact that iron production was less centralized than bronze, which required multiple long-distance supply chains that could be monopolized by regional powers. Different folks point the causality in different directions.

Some technologies are inherently more centralizing, some are less so.

I agree that most technologies can be applied for imperialistic interests, and the tech space is full of things that seemed decentralizing and have since been captured, but it's not quite law.

IC manufacturing is basically a modern global palace economy full of shipping and single-player choke points and large capital investments, so it leans hard to the centralizing side. It's less clear that AI will work that way, if it turns out that it is relatively practical to do useful work on locally hosted models it might end up being somewhat decentralizing.

Or, we'll get technological and/or regulatory capture, end up paying rents to technofeudalist douchebags with data centers and proprietary models who bribe governments and peek at user behavior and try to use their already vast resources to preemptively capture any value before anyone else can, until the next major perturbation.