As a storage admin for a decent size (1000+ employee) company, I get this argument from management every day.
We can go out and buy a terabyte for $50! How hard is it to justify 10TB for project X?
Answer: Your USB drive (or internal SATA drive, or cheap single desktop RAID solution) has neither the performance, reliability, or feature set required by a modern datacenter.
A standard (7200rpm) USB drive can get around 320 IOPS. A single application in an enterprise environment, serving multiple users, can easily require 20,000+ IOPS at the database level. An environment like the SSA could easily have dozens of apps serving the same number of users (employees, not even counting customers). How many USB drives do you plan on connecting? How are you going to maintain, monitor, and expand your 1000's of daisy chained USB drives? How many millions are you going to spend designing, implementing, and maintaining an interface to control them? How much to train people to use it?
Enterprise storage solutions from EMC, NetApp, 3Par etc. help control the issues above, but they don't come cheap. A terabyte of space for a NetApp filer, if you count licensing, training, power, cooling, disaster recovery etc., will run you easily $10,000 / TB. EMC storage can be double that.
I can't guess how much data they need to store, but knowing that they have 106,000 employees, and knowing that my company has around a thousand, even if they needed only a tenth as much storage as us, they're looking at 10 petabytes, or $100,000,000 in storage. If you budget roughly the same amount for network and server hardware, then about as much as both combined for application development, support, transition, and staff, then throw in a final $100,000,000 for government waste and bureaucracy, you're pretty much right on target.