You're talking about Wikipedia: note, Wikipedia isn't ad-laden! Despite serving up a huge volume of material, Wikipedia manages to do so on a community-supported model without advertising and tracking scumbaggery embedded in every page. You want a large-scale functioning example of alternate models, and you've just provided one yourself!
Wikipedia is an example of a still centralized, but advertiser independent (donation supported) distribution model. If you wanted a more decentralized Wikipedia-like system, you could adopt a bittorrent-like model: lightweight centralized indexes of content, but generally downloading the bulk of content from peers. For the thousands of people downloading "Algebra" between edits, the central servers would only need to distribute a handful of updated copies, then direct future requests to grab the page from distributed shared sources. Various distributed servers could take "responsibility" for hosting ranges of (alphabetically organized) words.
And what about search engines, do we just have thousands of search engines who each only catalogue a tiny portion of the web?
If a serious need arose, it's possible to devise distributed indexing models. Consider: what Google does doesn't require one massive supercomputer with a globally shared memory space to process every request; their algorithms already work with more loosely coupled distributed computing systems. Many people banding together could generate distributed indices. Furthermore, a peer-to-peer reputation based ranking system could help fight back against SEO douchebaggery screwing with search results --- the distributed cataloging system could include much more "real human" evaluation of "this is a good and relevant site for this search term, not just a keyword list on a domain squatter's site."