You really have to define "a bit of time." I defined it clearly in my post, and my math is not off based on the assumptions I posted. Please don't accuse me of being wrong when I provide all of the assumptions I used for the calculations, unless you've done the math with those assumptions yourself (and you can prove I did them wrong).

My math is correct according to the assumptions I gave and I stand by it. A 50 GH single will only ever produce about 28 BTC ($2800) if delivered in September and my assumptions prove correct.

If you want to argue about 18% increases every retarget being too pessimistic from Sept '13 to March '14 (when all currently available BFL devices become unprofitable at $0.11/KWh) , that's fine, but don't just say "your math is off." However, I don't think 18% is at all unreasonable. When GPUs were rolling out, we saw difficulty go from 19 to 9000 in about 5 months. And that's an order of magnitude greater than the change we're talking about (19M to 900M). If ASICs come on line at that rate, we'll be at 9 billion by February (except that none are profitable there, so it won't happen). But I don't think 900M is unreasonable by Jan 1. Granted we will see some leveling off after that as less efficient ASICs become unprofitable. But if 28 nm products are real and rolling out, they will quickly obsolete everything less efficient. This is a pretty inevitable path we're going down.

My entire point is that delays are extremely costly going forward. Progress is relentless.