I think computing limitations is the big one but also software limitations in terms of how much software a company might need to develop for itself to compete with a big player. Right now there's a lot of innovation going on. Not that there won't always be innovation, but in the grand scheme of time I'd argue web-tech will one day reach a state of the art that progresses more incrementally in the way that fields like physics, biology, etc. tend to with of waves of small insights rather than the few decades of rapid evolution they underwent in their infancy.

It's at that point of incremental innovation for web-tech where I think little guys become competitive because that's where they can start smaller businesses running cheap but 'good enough' hardware and premade open-source software to provide similar services to the big players albeit at smaller margins to provide comparable but less comprehensive products for cheaper. These smaller businesses are where I imagine the decentralization manifesting. Especially if some of those businesses are things like "Join our regional hardware co-op and contribute your unused processing power to our service in exchange for a share of the profit". That would be huge on its own for de-centralization, if a profitable service could bring anybody and everybody a worthwhile amount of money for simply running their web-service software during downtime.