Here's the problem with that, however. Over a 24 hour period, it may average out to that. However, it won't be uncommon that during prime time 5 out of 10 or more of the subscribers are likely to want to game depending on whether there was a highly anticipated game release. On release day or a few days after that it might jump up to 8, 9, or 10 out of 10 people hitting the servers at prime time.
We see this happen with Steam with highly anticipated releases where the servers will struggle to service all the download request. Steam attempts to mitigate this by having pre-loads for big releases. And obviously once the game is download you are no longer subject to whether or not Steam has enough servers as long as the system hasn't crashed. You can't do that with an on demand gaming service. You either service the request to play a game or you refuse service to that customer.
So, you either risk losing gamers if you refuse service to 60-80% (using your 20% number) of the subscriber base in that situation, greatly degrade their gaming experience by having servers over 1000 miles away servicing some of the requests (again risking losing subscribers), or you have the ability to service say 80% or more of the subscriber base during prime time (which means you need to figure out what to do with that hardware on off-prime time hours).
I'm going to assume that MS aren't incompetent and that their accountants are not OK with them burning cash on server hardware going unused, so MS obviously have an idea of what they might want to do with the XBS-X blades when they aren't servicing gaming requests. The question is, what services would be conducive to potentially being evicted from those blades on a moments notice if greater than expected demand (depending on the time of day) for those blades happens?
Regards,
SB