Just because something is available for free doesn't mean it's better for all use cases. There are cases where Oracle will perform better than Postgres (and vice versa of course).

And there's a business case for finger pointing

security issue with your open source DB install? It's either your fault (configuration), or the fault of some possibly volunteer engineer (bug). But if you pay enough, the whole thing is Oracle's problem, and you can tell investors with a straight face that it's not your fault. And Oracle are big enough that it's an easy decision to defend should something go wrong (which is something of a self fulfilling prophecy, but that's the way it is).

But yeah, whenever I need a database it's Postgres :)

Even a database with no licensing fees costs money in terms of wages/salaried employee time to use, so while that cost advantage is real, there are costs on both sides. If MS has products you want to use that are much easier (read: cheaper) to use with their paid database than some free alternative, that's certainly a good reason to consider it.

The longer you use it, the less likely it is to pay off, but execs focused on short term profits don't weigh that very highly.

Who would pay for a database when you can get ~~a free one~~ Postgres?

Typically support is cheaper and many times infrastructure and high availability will be baked into the contract. Unless you are planning the service as a core business capability it's typically cheaper to support.