I co-founded and sold a company that did SaaS/outsourced natural gas and electricity billing for energy retailers in deregulated markets, and cost consistency was a major reason why corporations signed fixed price energy contracts. Even though they might pay more than simply taking the spot market price, having known costs was preferable.
If I offer a service for a flat fee, I should keep my price higher than if you were to accept a variable budget item. If you start going over the flat fee, I'm eventually going to be taking a loss on the electricity/hardware. So naturally I might have to raise the price more quickly since I'm not likely to be a massive electricity/natural gas company. The price for SaaS/Cloud would not be as sticky as gas/electric and so the contracts would be for a shorter length.
Where-as, had you done some napkin math for cloud costs and tried your best to overestimate, you can add some additional overhead and see if you run over/under that year and adjust. You still treat it like a flat fee, but there's just a re-adjustment if you're over/under at the end of the term.
If you paid $10-$100/month for every component just for simplicity, a lot of things would never get off the ground. One of the big differences is Amazon has also unbundled support from their pricing.
A lot of startups could run on a single $20 virtual server for their first few years if they so desired.