Any prediction of decentralisation of the Internet, or anything built on it, needs to explain how this will overcome the economies of scale (1 big data centre is cheaper than 1000 little ones) and the network effects (everyone buys and sells through Amazon because thats where you find the most buyers and sellers).
Network effects of centralization explain why current big companies prefer to build centralized services -- but the next big idea never comes from existing big companies, anyway.
My 30,000' view of computing history is that big companies, as a rule, don't create big ideas that survive. They pick up small ideas that work and scale them up (Gall's Law). Git and HTTP and Python didn't come from Microsoft or Google or Amazon, but those companies took them once they were already popular and made them scale -- in large part by making them work well in a centralized architecture.
The software that comes next won't win because FAANG like it better for building centralized architectures. It'll win because everybody and their dog will be using it at home. Then companies like FAANG will see that and pick it up and try to centralize it. Then the cycle will repeat.
We've already got a billion tiny data centers -- the powerful computers and smartphones in our houses and pockets. All we need is the right software, and the desire to not have every service delivered free-with-ads-over-the-web from big corporations.