Yes it is, especially at scale. This is most obvious in the cloud where zero interest rate era companies blow ludicrous amounts of money running crap code at great expense.

Even then llama.cpp requires special hardware (a gpu) to run at an acceptable product level of speed.

I hate hearing this but I know it's true because I've tried to build my career as a software engineer around minimalism, simplicity, maintainability and efficiency at scale but it has constantly felt like a struggle; that there is no market for that. You're totally right in pointing out that a lot of valuable software is comically inefficient. Sometimes it almost seems like only bad, ridiculously inefficient software is allowed to succeed financially in this system. Freaky. I've literally seen projects with horrible code be worth billions and witnessed their market cap drop over time as the code quality improved. I don't know WTF is going on in this industry but it's weird.

You’re conflating constraint with bottleneck. Of course its a constraint.