The downside is

i) reproducibility. In the academic/research world there is a big move towards reproducible research, the nature of the Excel interface is not conducive to that. A standard scripting language is fine that plays with Excel but you don't want Excel necessarily as the main access point and the menu driven approach isn't conducive to keeping a permanent record of what buttons you click on. ii) In multi-cultural settings Excel defaults to the users system language, these days you simply jsut work with students laptops rather than using a standardized lab, so unlss you are polyglot working with Excel is just a really bad idea. Jupyter provides a standard interface and the default is always English, but supports multilingual content. iii) Python modules provide a huge range of solutions for many different domain applications, it's hard to see how Excel as an interface would provide support for every one of these modules. What I mean is how would Excel work as an I/O interface to a Python script that was doing something other than working with data? Notebook solutions already do this quite well.

Also for visualizing there are things like pandastable and DataExplore and other tools. You won't fix Excel by adding Python to it, but you do need the ability to process legacy xls and xlsx files so they need to at least continue that support.