Bitcoin is not trying to replace banking - the set of services that it's trying to replace is a (not so big) part of payments processing, and a tiny part of whole banking (for example, lending is a much larger function than payments).
If I look at the headcount structure of a typical bank or at the tasks where the wider financial industry employs people, the settlement of payments takes 1%, perhaps 2% depending on how widely you interpret the area. So total worldwide adoption of Bitcoin (or some newcoin) might replace at most 1%-2% of the current banking employment and energy cost. As far as I understand, this energy cost is already smaller than Bitcoin's despite processing multiple orders of magnitude more transactions.
First, what bitcoin tries to replace in the banking industry is not so clear cut. It could be payments, but also investment (value store), funding, and even credit. This very much depends on how successful and flexible the technology turns out to be. It is very early to say.
Second, the resources used by the banking industry to provide the services they do are not limited to the direct associated costs: we also have technology development, personnel costs (including energy used by the banking personnel to reach their workplace, the costs of turning on the lights in all the bank offices, and the fuel consumed by the bank managers to attend the party on the other side of the country without which the industry can not function, since, you know, you need some fun in order to sign the required contracts), the fuel consumed by the customers to reach the branch where, for the third time, they are going to evaluate their mortgage application, ... A long list.
In summary, yes, you are probably right that bitcoin is currently inefficient, and that currently it is not providing much value. But the sector is ripe for disruption, it is easy to disrupt, and any change in the current status quo will be an improvement, since the banking industry is so hugely inefficient for society as a whole.
At the core, the banking industry is just information processing, and that is the cake that technology is eating at the moment.
If I look at the headcount structure of a typical bank or at the tasks where the wider financial industry employs people, the settlement of payments takes 1%, perhaps 2% depending on how widely you interpret the area. So total worldwide adoption of Bitcoin (or some newcoin) might replace at most 1%-2% of the current banking employment and energy cost. As far as I understand, this energy cost is already smaller than Bitcoin's despite processing multiple orders of magnitude more transactions.