Yes. For comparison, Netflix has about $30B in revenue... paid up front for all of its content (something youtube doesnt) and accounts for a larger percentage of internet traffic (likely because of higher quality streams)... and they still made $6B net profit.
Yeah I thought that way too at first, it's an interesting take. These days maybe ML could help auto-coach promising content producers. For the other take: maybe pre-caching content that's more cost saving to cache than it is to not cache it can be aided by ML too. But replicating smaller libraries of content at all regional replica nodes is probably more straight forward at first.
