That is only the revenue, which is mostly meaningless because most of the revenue is disbursed to content creators. We also don’t know how much the whole system costs to run.
That's not the same thing. They had detailed server utilisation data when I was there years ago, and they had models that tried to translate that into some sort of dollar cost, but that didn't mean they really knew.
You probably can't know. The ground truth would be if YouTube were spun out and had to rent its infrastructure from Google directly via Google Cloud. But GCloud doesn't actually sell the infrastructure they use, as far as I know - for instance the search engine, the ads engine, the anti-abuse engines, Borg, the edge networks. AFAIK most of the stuff that YouTube relies so heavily on isn't actually available to buy at any price.
Pricing is hard work, too. What price should Google charge to license out their search engine tech for competing video sites? There's no existing market for that kind of tech that could provide an obvious price point.
For an integrated operation like Google/YouTube you can't ever truly say if it's profitable or not. The division of costs and benefits will always be rather arbitrary.
I disagree both on knowability and what needs to be known. The point of the exercise is not having a calculation down to the cent. We don’t even know if the order of magnitute of the revenue and the cost is the same. By its nature I would expect youtube to be very IO and compute heavy, which would dominate the cost function, to the point of rendering rest of your list into bells and whistles.
> Pricing is hard work, too. What price should Google charge to license out their search engine tech for competing video sites? There's no existing market
That is precisely the function of making a market, supply and demand meet and iterate over the price. Right now there is no price because there is no market.
