Just thought I'd try to contribute something.  Your model relies on a fixed 1 W/GHs which we know is not realistic, so I thought I'd try to make a model to predict future chip efficiency using past and current developments of both BitmainTech's and Avalon's chips, as they are 2 of the bigger players that have been around for a while.  Please feel free to interject where you see any of my assumptions as unrealistic:



Both Exponential formulas come out somewhat closely, so I feel it could be useful to predict future technology as far as efficiency goes.  We know there has to be a point of diminishing returns.