This point particularly applies to breaking cryptography. A non-NISQ quantum computer with a slow runtime is quite possibly the first place we get. The runtime would constrain other use cases but not, notably, recovering a private key from exposed Bitcoin pubkey using Shor's

$4T is a pretty big incentive. I agree that outside of cryptanalysis, there aren't many use cases that look lucrative. But then again, early computers of the 50's and 60's didn't have a whole lot of use cases relative to today either. The innovation flywheel happens when breakthroughs in hardware lead to new algorithms/approaches that weren't possible, leading to more investment in hardware, etc Are you arguing that's not a *possibility*? Or are you just arguing it's unlikely? Either way, what data are you basing your assessment on?