But if you're calculating month over month increases the way I was... not quite since diff is a function of hashrate over the last 2016 blocks, I was taking a 4PH hashrate, increasing it at 30% every two weeks and seeing what the resulting difficulty would be 3 months out, and that gives us around 5 billion.

10 billion diff in march means a roughly 20x increase in network hashing power by then.  I simply don't think there's enough next-gen hardware in the pipeline to support that kind of growth by then, nor will the vast majority of delivery targets be met.