By that definition, it seems like no services are critical online, as there is always some alternative. The reality is that doing all the changes you mentioned could literally cost some companies many months or even years of time.
I find that unlikely. AWS hasn't been a major provider for that long. They weren't PCIDSS certified until just 5 months ago, so no big companies were even doing ecommerce independently on their platform yet.
If it took a year to build your service on AWS products, it shouldn't take years to do it again with another provider now that you've done it before. A server is a server, whatever you've programmed to run on those virtual servers will run on identical virtual servers at another company. It's only the parts of your system that interact with Amazon's APIs, and your human systems/policies, that have to be reworked.
And if you're big enough that your systems take years to build, you have a disaster plan already. If Amazon blinked out of existence tomorrow, Netflix wouldn't be standing there with no idea what they'd do to get back online.
