There's undoubtably more data being served but I'd be very curious how the edge caching efficiencies differ given how many people watch the same things (recommended by the algorithm) and since YouTube's content is so much shorter on average that larger catalog is somewhat balanced by any given cache node being able to hold more videos.
