The only things I'm aware of that I consider actual problems it solves are "it breaks classical encryption" and "you may be able to use it to directly model other quantum systems like for protein folding and such".

Everything else I consider pretty silly. "It can improve logistics" - I'm fairly sure computers are already as good as they can be, what dominates logistics calculations isn't an inability to optimize but the fact the real world can only conform so closely to any model you build. "It can improve finance" - same deal, really. All the other examples I see cited are problem where we've probably already got running code that is at the noise floor imposed by reality and its stubborn unwillingness to completely conform to plans.

If I had $1 to invest between AI and quantum computing I'd end up rounding the fraction of a cent that should rationally go to quantum computing and put the whole dollar in AI.

By far the most exciting possibility is one that Scott Aaronson has cited, which is, what if quantum computers fail somehow? To put it in simple and unsophisticated terms, what if we could prove that you can't entangle more than 1024 qubits and do a certain amount of calculation with them? What if the universe actually refuses to factor a thousand-digit prime number? The way in which it fails would inevitably be incredibly interesting.