Conceptually, if very smart people did it, and other very smart people say it can't be done... that's what you're looking for for a deep moat. Usually that's going to be some sort of new mathematical principle in software.

It's excruciatingly hard to get funding if you do come up with that sort of idea, because the experts all say its impossible. ;)