This is a common fallacy, usually caused by people either not calculating or excusing downtime for self-hosted systems. A big one is time-to-recover: I’ve seen people disparage cloud options for years and then eat days or even weeks of downtime after the wrong server fails, and attempt to handwave reasons why that shouldn’t count the same as a 15 minute outage which was resolved before most users noticed.

One contributing factor may be the difference between surprise and planned outages: if you know things are going to be down and how the work is progressing, it doesn’t seem as dramatic as an outage of unknown duration.