I am not aware of an AWS outage in their 15-16 years of existence, that an architecture, built according to the recommended best practices, of distribution across availability zones and regions, would not be able to withstand. I am willing to be proven wrong. Can you provide one example?
Of course, these come with increased cost, but I am thinking web retail on a large scale, or airline companies, where for example a downtime of a few hours will easily wipe out any savings made by relying on a local data center. It might not be the solution for a smaller company.