In the early days of the #fediverse user numbers (aka hosting costs) were low and enthusiasm was high.
Admins made their instance because they wanted to. Moderators volunteered because it felt meaningful (or whatever). We all had similar values and vibes were good. When the hat was passed around everyone happily threw a few bucks into it. Setting up a small instance can be done fairly easily by using a docker image so only basic linux skills were needed.
All that is about to change. 1/n 🧵
ActivityPub, the language used by instances to syncronise is... verbose.
Let's imagine a person who has 500k followers spread across 10k servers. Each post that such a person sends creates 10k Sidekiq jobs. Each interaction—another 10k Sidekiq jobs. If the post gets popular and gets 1k favorites (that's 1/500 of the total followership of that person) in a short period of time, we're talking about 10M Sidekiq jobs. And all of that needs to be coordinated by one server. https://softwaremill.com/the-architecture-of-mastodon/
other extrapolations from "outside" 
why are accounts with 500k followers of interest, if Twitter isn't to be recreated? RSS is a legitimate option, servers serving RSS and AP are too, clients supporting RS and AP are too.
why not ads, if that is what users are going to "pay"? would typical admins refuse to federate with any instances that, say, implicitly, mandatorily, subscribe all their accounts to an ad bot?
why aren't lighter-weight servers like Pleroma/Akkoma/whatever an option? why is Mastodon-identical behaviour assumed to be inviolable?
Vapourware, but why shouldn't storage be de-duplicated across pairs of servers? Sure, disable it if you want copies. But jointly provision IPFS pinning and exchange hashes in S2S communication.
@tetrislife I have no experience with those platforms but some of the issues arise due to the nature of the protocol, which they all share. So more efficient software lowers the bar for a tiny instance but for big complex networks the protocol design will overwhelm anything.
Yes I bet there are efficiencies that can be found by multiple instances sharing the same server/infrastructure. Mastodon does not seem to be built with that in mind but I'm sure it'll come up a some point.