I don't think you quite understand it.
Yes, BTC is proof of work. Yes, complexity increases to control the rate. But that's it.
Complexity has nothing to do with scaling. BTC doesn't scale. It's fixed at 1 block per 10 minutes, 1MB per block. No amount of horsepower is going to give it more capacity, so BTC doesn't scale up by wasting more power. It stays the same.
You can see this in forks like Bitcoin Cash, which took the exact same algorithm, and forked the exact same chain, but gained less miners. After some time it readjusted, and BCH is perfectly functional with less miners than BTC.
I agree that BTC and its ilk suck in many, many ways. But saying that power usage must increase is incorrect. It can remain static or go down without changing the functionality or the capacity of the network to process transactions, because those things don't depend in any way on the hashrate.