It's funny. The other day I received a campus-wide email asking to conserve power to help avoid brownouts. So I wandered around the building, closed a bunch of shades to keep the sun out and add that extra tiny bit of insulation, and nudged up a bunch of thermostats a degree or two. Then I realized that the elephant in the room was the 6-13 GW of power consumed by the computer room next door. My little bit of conservation was like peeing in an ocean compared to the racks and racks of computers and the cooling needed to keep them running.
Truth be told, a data center will not immediately crater during elevated temperatures. The heat tends to shorten the life of the components - but it's not immediate.
As an example, there was an IBM mainframe running OUTDOORS
throughout the entire 1968 Hemisfair (South Texas summer heat and all) in San Antonio. I use to work with a guy (long since retire now) that was on the team that supported the system. No hiccups at all. They did, however, have to vacuum out the dust on a daily basis.
Right now, our racked-server battery backup system (half-cycle fail-over) will hold everything up for a little over an hour without the help of the diesel generator (~15-seconds startup) - with room cooling shut down. Should the generator fail, we have a shutdown list of non-essential servers to vary off to extend that time. The mainframes are on a different battery backup system. Both connect to the generator on separate circuits.
As far as I know, our data center is not on the same circuit that participates in the "Conservation Events" that I mentioned earlier. That may, primarily, be because the mainframes are water-cooled.