In other words, just because you might get 10000000000gigs of "free" inbound, doesn't mean you can actually get that much inbound out of the pipe.
For the most part that's roughly comparable to how hard it would be to generate that much data, parsing and datamining are probably a little bit more cpu intensive than generating data. So if typically you find that you can push 300 Mbit out of a box you can expect to get around that much inbound or a bit less, depending on your application.