This model is fundamentally flawed. It assumes (highly) exponential increases in hash rate with time but constant energy efficiency. However it is nearly entire

This model is fundamentally flawed. It assumes (highly) exponential increases in hash rate with time but constant energy efficiency. However it is nearly entirely increases in efficiency that have driven the higher hash rates. If those efficiencies stop increasing (and we've probably picked the low hanging fruit already with ASICs), then hash rate growth rate will dramatically slow.