OP's post is an observation of the real world: the inter-related evolution of technology and humanity's use of it. That use isn't primarily driven by system architecture. It's sociological, commercial, political, emotional.
The conception of the fediverse was itself a sociological and commercial response to social media power being vested in too few commercial entities. The distributed nature of it isn't cost free: there is no single "Mastodon" that everyone signs up to. It remains to be seen whether federation is a net winner, or whether the incremental usability friction prevents large scale adoption.
None of that is first-order technological. The underlying internet may well be federated, and it's eminently possible to create federated services on top of it. Email is perhaps the signiature example. But _just being_ federated is neither necessary nor sufficient for successful adoption. Witness the very centralisation that the fediverse is a response to.
One difference we see already is that Mastodon has instances focused on specific topics. There are instances focusing on art, news, tech, functional programming, and so on. Federated approach is showing itself to be much more conducive towards smaller communities focused on niche interests.
Meanwhile, I contest the claim that Fediverse is being centralized. In fact, server admins actively ensure that this doesn't happen by closing registration when too many users end up flooding to a particular instance.
Furthermore, the fact that servers are maintained by volunteers via donations means that there is a limit on how many users each instance can support in terms of cost. Once again, this creates incentives towards further federation to amortize costs.
Finally, the terms of success need to be defined to talk about success in meaningful terms. I would argue that Fediverse is already successful because it has millions of users producing content, it has lots of developers working on the tech powering it, and a sustainable model for running servers.
At this point the Fediverse has shown itself to be sustainable and will be around indefinitely. It will likely outlive every commercial social media platform that's around today.
"What are you doing here, believing in centralization, on a centralized news site?"
- It's hosted on servers presumably operated by a centralized entity (YCombinator/AWS/etc)
- It has moderation centralized to a specific set of people (which is weird sentence construction but I had to fit centralized in there for the point)
- Post content (even links), comments, accounts, etc, are stored in a centralized database (might be sharded or whatever, but you get the point, the control is centralized)
- It has a domain name and DNS records that are centralized to ICANN/whoever-hosts-the-dns
Centralization, to me at least, is about how content is delivered, not about how likely the product is to bounce you to an external source.