So yes a drug-resistant bacteria might be outcompeted for limited resources in a normal environment by bacteria without that adaptation. However, if there is an antibiotic introduced into the environment, the drug-resistant bacteria will probably have proportionally more resources than the non-adapted bacteria did before the drug was introduced, due to a complete elimination of their competition.
This is assuming that the adaptations bacteria have that enables them to be drug resistant harm their fitness in a drug-free environment in the first place. It is entirely possible that the drug was targeting a trait that actually required more resources to produce, but that wasn't selected against because it was basically vestigial.
I think it's a pretty big misconception that the force of natural selection is directed towards absolute efficiency. I'd classify the process of natural selection as a 'good enough' mechanism.
But if you completely remove the antibiotic from the environment, the resources required to maintain the defense offer a small selective disadvantage. Resistant strains of bacteria do not remain resistant indefinitely. Repeated exposure to the antibiotic is needed to maintain resistance. But even very low exposure levels will suffice for that.
Just like software doesn't usually face sudden new performance issues when vulnerabilities are patched. Sure, openSSL is technically slower once they patched heartbleed, but not much slower. Certainly not enough to be a factor in anyone's decision of whether to use it.
I'm also a complete layperson to the topic, as you may have guessed already.