Economies of scale basically.

It's way simpler to just connect to Wikipedia.org and download the pages you want to read instead of downloading the whole Wikipedia.org database and keeping it stored and updated on your devices. Same principle.

Mostly the cost of hiring a DevOps engineer to set it up and maintain it and taking on the additional risk of having to deal with upgrading the node etc... It's just cheaper and easier to go centralized at the moment.

Though things are never cheaper to maintain than at the beginning of a tech bloom. Its only going to get more expensive to create and maintain as the node and the APIs get more complex.

Sure but these aren't crazy complicated beasts - a binary installation and some unix experience gets you 90% of the way there! I don't think they'll get far more complex in the next few years at least, and documentation/user support is really good

Its all the dependencies, security updates and keeping all of them up to date and in check. One word, Log4j.

same reason people are moving to aws, and people which used to use aws instances are moving to serverless.

Leave infrastructure work for other people.