Yes, that is the assumption.  Actually I do a quick depreciation to %65, then linear depreciate to 0%.  

I do consider 2 years to be conservative, but not without a realistic basis.  If a card craps out at month 16, my depreciation cost estimate would instantly become 33% too low.  Also I'm unsure how far factors such as ASIC development and a flood of GPUs into the market following a confidence crash could lower GPU prices.