I think the $5/month he's envisioning is for the computing/bandwidth charges on the cloud providers he mentions. The software, container, images, etc. should all be provided for free, and the open source bug tracker should have a 1 second SLA for replying to questions on getting it set up.
But on a serious note, yes, it would be nice if there was some pre-built container that just served the OSM tiles on an HTTP port. I imagine most people integrating OSM want to use their existing static-file-serving setup, which is why it doesn't already exist. (Or maybe it does, I haven't looked.)
In the end, Google Maps is free for 25,000 map loads per day and you don't have to run any servers. I'm guessing this is why it's more popular. The $5/month the OP ballparks as a price is how much Maps would cost for 760,000 map views per month. In the end, I doubt you'd do a lot better than that unless your bandwidth is free or you're serving a lot of maps. This, I guess, is why OSM isn't much more popular. The map data ends up not being the expensive part; bandwidth/hosting is.
I think the innovation in OSM is being able to use it for whatever you want. Garmin bike GPSes don't even use Garmin's own map data anymore; it's apparently cheaper for them to just use OSM data. That's because they don't have to pay the bandwidth hosting costs, they just dump the data out of the database and give people that care an update every 6 months. Each incremental map render costs them nothing, and nobody can stop them from using the data however they want. Pretty interesting that it ended up a better deal for them than using the maps they collected themselves (and probably supplemented with other vendors' data that won't let them use it on that "download once, use forever" model.)