Let’s leave the networking aspect aside for a moment.
When a language is compiled, the source files go through a pipeline of parser -> preprocessor -> compiler -> assembler -> linker, to end up with an executable. With interpreted languages, the source code is instantly executed line by line by an interpreter software. With JIT languages, the program gets compiled and optimized into portable bytecode, which is run by the language’s runtime.
If I had to guess, web pages (i.e. HTML/CSS/JS) are most likely run by an interpreter that is a web browser, but isn’t that inefficient given that most of what people do on computers is browsing the web? What about browsers, what standard is there that specifies how each language should be run/rendered? What pipeline does a webpage go through to end up as a process in a computer?
It’s true that docx (also other Office Open XML and OpenDocument file types) is a zip file containing the files that make up the document, but those files have little in common with HTML, they are their own XML schemas.
Also, OpenDocument was designed by humans to be an XML schema well-suited for describing word processor documents, spreadsheets, presentations, etc., while Office Open XML is basically a memory dump of MS Office internal data structures wrapped in XML tags. They are very much not equivalent except for Microsoft bribing ECMA to make their shit a fake “standard.”