I'm sure the existing banking system takes energy+other resources to maintain. But the incentives in the existing banking system are to look for ways to use LESS of these. Each Kilowatt-hour a bank doesn't use shows up in their bottom line.
Proof of works is DESIGNED to waste energy. If tomorrow, you'd make hash computations 1000x as energy efficient, the hashing difficult is designed to adjust to make sure that 1000x as many computations are performed and the same amount of energy is wasted.
Think of it this way: Currently, 900 bitcoins are mined per day. At a price of $40K per bitcoin, whoever mines those bitcoins earns a combined $36M. So miners are incentivized to invest approximately that much money into energy & equipment per day (minus some profit margin). If they invested substantially less, somebody could invest more and pick off their coins.
(And that's counting only mining profits. There are also transaction rewards, and, if you were to control a majority of mining, more nefarious sources of profit).
This means that PoW has some truly perverse incentives:
* Industry wide energy efficiency does not matter, because the difficulty will adjust accordingly.
* The higher the bitcoin price rises, the MORE energy gets wasted (That's a big difference with e.g. Gold. Sure, rising prices lead to more marginal deposits being exploited, but it's not like pickaxe prices for a given prices rose proportionally with gold prices).