Can you list your assumptions? For starters, are you assuming constant power production and constant power price? Because neither of those will hold whatsoever on the timescales you're talking about.
It is not a perfect model by any means that is why I called it "rough 'back of the envelope' estimates" . All values static and based on "today's" numbers:
https://bitcointalk.org/index.php?topic=694401.0The rising worldwide power production will "help" us use a lower percentage of the overall power production.
The rising cost of power will "help" us use less power.
Both of which are why your "estimates" would better be described as "upper bounds" for the portion of energy production used.
As for some estimates I have gathered from people in the US power industry with knowledge of such things, the expect power costs to go up 2-5X in the next 20 years, barring a quick and tremendous surge in new nuclear construction. That is very unlikely to happen since no new nuclear power construction projects have been initialized since the 1980s and the coal, gas, and renewable industries can easily stir up the emotional argument against any proposed projects. All that said, the US is only 4% of the world's population, so maybe overall it won't be so bad.