I think a bit of "It depends" is what is needed here. Especially reading the other comments, you can see that large groups of users actually _do_ experience improvements and see the value of these tools. That said, the common theme is that it works well _after_ you have invested the time in producing the needed configuration files. And these are highly environment-specific.

Take the following scenario: I have an application that requires an up-to-date Linux environment, an up-to-date Java JRE/JDK, several specific subsidiary tools, and some networking and storage configuration. Additionally, I need a group of these apps deployed, with some additional out-of-the-box tools like a LoadBalancer, service registry, etc.

Now, ignoring if my target is GCE, Azure, or AWS, I can do this with Terraform or something else, and the result is a quick and painless deployment, even upgrades, and certainly a full rebuild. (disaster recovery scenario) But if I now say "Great! Works on GCE, do me a setup in Azure tomorrow!", we're almost back to square one because the structure of the infrastructure is completely different. Worse, the Terraform scripts and configuration files need to be redone because the variables have different names and different meanings. Abstract away the differences, and you've got a job that may require some modern AI involvement before it can promise optimal results on either platform.