You seem very confident of your input parameters. What are you basing 0.2-1EB on? It seems like there is a lot unknown about the utah data center, so all their "expert estimates" sounded more like guesses. Hence my conclusion "nobody knows for sure".
Even if, as you say, the estimate is 100hours per person. How much time does the average person spend on the phone? My current cell plan has 200 minutes per month, and I never go over. Say the average person spends 1000 minutes (16.7 hours), that means you could store a month of rolling phone conversations, and be only at 16.7% capacity. That leaves a fair bit of extra room for persons of interest internationally.
Your estimate is correct, somewhere on HN I've read that average young American talks 20 hours per month, older somewhat less. You're also correct that as soon as they decide that not everybody is actually important, the capacity to store communications of "persons of interest" is enough for "the whole life of the persons of interest."
Regarding my estimates, somebody here on HN also sent a link to the highest capacity automated tape libraries, and using them instead of hard disks the estimation of capacity still fits in the same order of magnitude. I still assume hard disks are more convenient than tapes for anything that needs to be accessed "when needed" (who would accept to say "will get the data bout the terrorists regarding the attack tomorrow, but it will take two days." So I can even bet that the main storage there are hard disks.
Even if, as you say, the estimate is 100hours per person. How much time does the average person spend on the phone? My current cell plan has 200 minutes per month, and I never go over. Say the average person spends 1000 minutes (16.7 hours), that means you could store a month of rolling phone conversations, and be only at 16.7% capacity. That leaves a fair bit of extra room for persons of interest internationally.