he means that he's using stuff like Lambda, Cognito and other PaaS services provided by Azure and AWS - not considering the cloud as IaaS only providers.
I've been curious to learn some of this stuff.
Even if you try to avoid lock-in. It’s usually not worth the risk of regressions and downtime to change your underlying infrastructure once you build on top of it.
The chances of AWS or Azure going out of business in the grand scheme of things is not worth the trade off.
1. Being able to do this would mean lots of work to abstract and polyfill the discrepancies between providers.
2. If AWS or Azure goes down globally, everyone else would be too busy freaking out about their own problems than be worried about the downtime of your SaaS.