Your dichotomy ignores other possibilities. To what extent are we stuck on costly distribution models specifically because those are the ones you can tap for a revenue stream, to what extent are the costs driven by the monetization itself?
I used to help out with the technical infrastructure for wikipedia and even as the 8th most trafficked site on the internet the cost was radically lower than that of some of my dayjob customers with comparable levels of traffic. A non-trivial part of the reason for that was that it didn't need all the surveillance and realtime per-request bespoke generation that advertisement encumbered services require. Part of it was just that there is an awful lot of fat in big commercial operations, since they have enough revenue that it's not a concern. Some is probably because when you're chasing ad dollars the extra spend required to get a 30 ms response instead of a 60ms response might be justified on the basis of the marginal clicks it generates, but those costs shouldn't properly be considered part of the underlying distribution costs.
I think "yes, I'd be happy to pay for all of it" would be significantly more attractive if paying meant just the component of the cost that actually went to the authors of the material -- on the order of a hundredth of a cent per view. Here is $1, that'll cover the authors share of all my household's youtube viewing for the next several years.