This article actually reflects a really really common misconception.
"As bitcoin grows, the math problems computers must solve to make more bitcoin get more and more difficult," which means more processing power is needed, Holthaus wrote.
This part isn't actually correct. The difficulty of the math problems computers need to solve to advance the blockchain change in response to how many and how powerful all the computer trying to advance the blockchain are. It doesn't change in response to either the price of bitcoin, the number of bitcoin users, or the number of bitcoin transactions.
So the statement reverses causes and effect: it's not that as the math problems because harder more processing power is needed, it is that as more processing power is used, the math problems become harder.
If 90% of the CPUs/GPUs/and ASICs currently working on the bitcoin blockchain disappeared tomorrow the exact same number of transactions could be processed as today for 1/10th as much energy, without changing the code or math of bitcoin at all.
And he made a startling forecast: Without a significant change in how transactions are processed, bitcoin could be consuming enough electricity to power the U.S. by the middle of 2019.
Six months later, that demand could equal the world's power consumption.
I would be happily to take the other side of the bet if this guy would like to wager the entire world's energy supply will be devoted to mining bitcoin in 2020.