Storing text data isn't nearly that costly. Audio as well. I think they have the capacity to easily store the audio contents of every phone call made in America. Video is another question.
Much of what they capture can be stored temporarily for filtering, saving what looks interesting and discarding what isn't.
Sure, we can concede that they aren't storing absolutely everything. The point is: they're storing a lot, and they're trying to store as much as they possibly can.
