Hoarding lots of data is the kind of thing that becomes really, really cheap to do (per GB) once you do it at scale, and is not-so-cheap otherwise. I think we need both, the big centralized hosters like the Archive, and the distribution that IPFS provides. But the latter mostly as a last resort, to keep lots of copies around of the most socially valuable stuff, that we wouldn't want to lose even in a disaster scenario.