Therefore, if 328500 BTC is mined each year (6.25/block) and 26 PWh of electricity is produced each year with a value of $2.6 trillion (at $0.10/kWh), then the cost of mining will approach the value of all the electricity produced in the world at about $8 million per bitcoin. That implies that at that price, all of the electricity in the world will be devoted to mining bitcoins.
The halving will take care of that, it will be nearly impossible for it to reach that level before 2024, at which point it will need to grow again by 2x and 2x every 4 years. But since you're talking in a hypothetical scenario there is something else to be noted, not only is this energy consumption depends on the price but also on the available gear that could mine.
Assuming last month's numbers, and rounding them pretty good we were doing around 125 exahash at 10k/coin, taking the best gear available on the market and you need 1.1 million of them. Now for 8 million a piece, you would need around 1 billion in mining gear, which will take ages to manufacture, and as the profitability goes down so will the incentive to produce gear to the limit, again not doable in this small interval.
It seems that the only way out is to increase energy production faster than the value of a bitcoin.
And again theoretically over a long period, it can be done.
According to IEA, the increased production of electricity in non-OECD averaged 5% per year, and we're talking about 80% of the world population here, with China, India , Russia , Brazil, and many more, who have enough potential to drive the growth up.
If we can't fight the apocalypse scenario energy wise, do you think that we can defeat it by breaking through the computing plateau, one promising future technology is quantum computing. Having a faster computing speed means that blocks are mined faster.
It doesn't matter with what you mine or how energy efficient that is!!!
It's a matter of what you can afford to pay in electricity regarding what you get in profit.
An S19pro does 110th/s while an S3 had 0.4 Th/s , that's a 300x increase in hashing power with just barely 10x in power consumption, from around 0.8J/GH to 0.03J/GH. Did overall power consumption drop? No!