I also share a video link with many, many, many more people than I ever could convince to buy a game. I might convince 2 or 3 friends to buy a game but I can incidently get 30,000 people to watch a video with a single tweet.

Think of it more like a loss leader. Costco sells hotdogs at a loss because it brings in more than what it costs them.

I'm sure Google/YouTube has crunched the data and came to the conclusion that it isn't worth it anymore but I also don't think data can even give a complete picture. Data can't predict how many people will leave vs subscribe when all is said and done. Results from limited rollouts are not guaranteed to follow in a full rollout.