Even if one does have a single binary and config file that one can just copy to a server and run, there's more to non-trivial deployments than that. For example, how do you do a zero-downtime deployment where you copy over a new binary, start it up, switch new requests over to the new version, but let the old one keep running until either it finishes handling all requests that it already received or a timeout is reached? One reason why Kubernetes is popular is that it provides a standard, cross-vendor solution to this and other problems.

Then you need to add management of storage for it, management of logs, integration of monitoring, healthchecks, maybe some multiple environment case because UAT is good thing to have, etc. etc.