Work there is published if and only if it is in Microsoft's interest to do so.
Do you have any evidence for that claim?
Often a bare outline is published: enough to impress people but not enough for an outsider to duplicate the work. This isn't academia.
Actually, there are plenty of research papers that describe a system in some detail, but not nearly enough detail to allow an "outsider to duplicate the work". Both academics and industrial researchers sometimes accompany a research paper with a working prototype; in both cases, they often do not.
This is where MSR is something different. Work there is published if and only if it is in Microsoft's interest to do so. Often a bare outline is published: enough to impress people but not enough for an outsider to duplicate the work. This isn't academia.
Xerox PARC and Bell Labs back in their respective days did function like academia.