From the abstract, its the inherent noise. So it's going to be some variance of monte carlo random approaches, or simulated annealing where "beneficial noise" helps find a useful optimum better than the others found so far.

(not a mathematician, and happy to be shot down in flames)

Since most systems looking at generalised QC want to minimise noise, and it's only the systems which do Quantum Annealing which "work" right now, I'd say this is a niche use case which will make a lot of money, right up to the point the cost of the machine drops so everyone does it, and then the market is back where it needs to be: equally available to all players, at least "all players who can afford time on a QC which does QA"

News release from HSBC

https://www.hsbc.com/news-and-views/news/media-releases/2025...