CI has essentially become a commodity with a need for reliability.

For us, we moved from Travis to Github because Travis just wasn't reliable enough (and had terrible security). We'd be blocked from deploying for hours at a time because Travis had a random outage. I know it's way more complex than "boxes that run some stuff", but CI/CD isn't exactly rocket science.

We also noticed that by using a dedicated CI provider, we doubled our deployment downtime. We'd be blocked from deploying when either Travis or Github had issues. Whereas, we're only blocked now when Github has issues.