To get network effects, it's all about low friction and low barriers to entry.
There is still a long way to go to make it work for your grandparents, however I think we already made quite some progress. The act of getting a domain and a server itself is still something most people might not be comfortable with though.
In the end I think there are only technical issues to be solved to get to a state where having a server and installing apps, just like most people now do with mobile apps on their phone, is entirely viable.
The internet was meant to be peer-to-peer not server-to-consumer.
Adding more servers, even at $5 won't change that.
From the network standpoint it is peer-to-peer as the internet is a network of dumb pipes with intelligence put in the periphery of the network. This is different from say the european minitel network[1] where dumb terminals were connected to a central intelligence.
This in of itself changes the costs and expectations of the parties involved which means things will likely slump back to a permutation of this status quo even with a different standard.
The current last-mile problem is that there is not enough demand for high-bandwidth applications to justify rapid improvement (we'd all have 10 gig connections by now if the 2000's trend had continued).
The server farm is part of the periphery of the network and has a last mile connectivity same a residential home connection. It is really expensive and only makes senses if you stockpile machines to have economies of scale.