This is why when I contract for an early stage startup, I pose the question:

"What if your app went viral and you woke to a $20k cloud bill? $50k? $80k?"

If the answer is anything less than "Hell yeah, we'll throw it on a credit card and hit up investors with a growth chart" then I suggest a basic vps setup with a fixed cost that simply stops responding instead.

There is such a thing as getting killed by success and while it's possible to negotiate with AWS or Google to reduce a surprise bill, there's no guarantee and it's a lot to throw on a startup's already overwhelming plate.

The cloud made scaling easier in ways, but a simple vps is so wildly overpowered compared to 15 years ago, a lot of startups can go far with a handful of digitalocean droplets.