So lets imagine this. COD , Battlefield , halo and destiny sequels all launch between the end of Sept and the end of Nov. That's just those games , there will be a few dozen other games released then also. So how much over build do they need to do to handle all this ?
Like I said at big MMO or multiplayer games services still crash and that is with the end user having their own hardware to run the game.
Significant. If your goal is to appeal to 2B gamers, that's a lot more hardware than all the consoles and PC hardware combined today.
But at the same time, you don't expect concurrent players to be 2B, so you can save there.
Where if you have 2B hardware owners, you must produce 2B devices.
Large cloud providers should be providing solutions to spool up servers and take down servers as required.
Players playing Halo can't simultaneously be playing COD and Battlefield. So there's going to be upper limits to the situation you describe.
Let's do some napkin ballparking:
On Steam, at the peaks Halo is ~150K, Battlefield is 50K, COD the numbers aren't known yet. But lets say 150K to be safe. You've have be able to support 350K instances at once globally just for Steam, and say 2.5x that to include both consoles. So about 800K at max peak levels.
Sony has been shipping over 1M units a month in PS5s alone and that's a much tougher process than building out 1M units a month for cloud. You don't need packaging, you don't need cooling, you have a big allotment of memory and access to a different kind of hard drive space, and a single blade is likely holding several chips. And a large number of blades can fit into a single cabinet making transportation issues a non issue.
You don't need to ship controllers or power cables, no DVD drives.
I think a company like MS can easily handle shipping Series S, X and making server blades all at the same time and keep it up. It will always take some time to shift your entire cloud base over to new hardware, since so many blades need to be replaced, but if that's all that is required, then shifting to new hardware may not have the growing pains as a net new premise install.
Lastly; if you look at AWS, Azure, and GCP, and others; they have arguably done much more work than the potential lift required for games, as they did it for all industry computer jobs ranging from web, storage, VM, compute, ML/AI, etc.
The list of available services on Azure, AWS, and GCP is massive and cheap.
