https://en.wikipedia.org/wiki/Utah_Data_Center
The lower bound on Forbes' estimate of its capacity is 3 million TB, and it consumes 65MW.
What precisely do you think they are storing there?
> Binney has said that the facility was built to store recordings and other content of communications, not only for metadata.
> The project had been designed for foreign signals intelligence (SIGINT) collection but, Binney alleged, after the September 11, 2001, terrorist attacks, controls that limited unintentional collection of data pertaining to U.S. citizens were removed, prompting concerns by him and others that the actions were illegal and unconstitutional.
More importantly, sorting through that amount of data to find anything is nearly impossible, even if you have 65MW of servers running. Even the entire Amazon cloud would have difficulty, simply because of the scale of the problem.
So yeah, they're storing a lot of data compared to the average corporation, but not compared to the amount that exists nor are they storing all data that gets transferred. It's just too much.
From a certain point of view, there's less surveillance going on than you might think, simply because collecting everyone's information all the time is impractical to do and pretty useless.
https://en.wikipedia.org/wiki/XKeyscore#Workings
https://daserste.ndr.de/panorama/xkeyscorerules100.txt
https://www.wired.com/2014/07/nsa-targets-users-of-privacy-s...
https://www.theverge.com/2013/7/31/4574532/xkeyscore-data-tr...
https://www.theregister.co.uk/2013/07/31/prism_put_in_the_sh...
> The cover has been blown on an NSA program which collects data on “nearly everything a user does on the internet” even as the debate rages over the secretive US agency's mass surveillance of innocent people.
> The XKeyscore program covers emails, social media activity and browsing history
That's more than enough to store emails, unencrypted chats/messaging, gsm and voip phone records (metadata, and voice where available), GSM HLRs, and SMSs for all time; that's also the lower bound. Presently I'm sure it's expanded to parse and store as much metadata as possible that's ingested/processed by private data brokers as well, which as we now know are receiving copies of timestamped, precise GSM location data from many consumer apps.
https://techcrunch.com/2018/09/07/a-dozen-popular-iphone-app...
That original capacity estimate ranges up to 12EB and also was dated 2013; I would anticipate it to be at least 2-3x that by now, to be more like in the 10-40EB range.
They don't store everything in case it's needed, nor do they have programs capable of detecting everything people suspect is monitored.
It's still an invasion of privacy and probably breaks various laws, but it's not the complete surveillance people tend to think of, and it's not as useful as you'd expect given the money invested and the effort put in to it.
I'm pretty sure "every email, private message, text, search query, hostname/ip connected to, and call log of phone call you make" is pretty much "anything" for the vast majority of people.