As a user of both, the lack of a true semantic layer is a massive hole. I
ve got approximations of similar functionality (i.e. publishing blank reports with the dataset built and then using the dataset as a datasource) but its not the same thing. Nothing can be done in SSAS to help either with a large backend database as SSAS OLAP doesn`t suit and SSAS Tabular is relatively simplistic.
Try working with PowerBI over a large (80+) table OLTP database and it
s in-memory model is also painful, its not like you can bring it all in and model it in the same way as a Universe as it would have to bring all the data into memory which is massively inefficient. Then add to that the fact that PBI is limited when it comes to joins. Then using directquery isn`t great in this scenario as it then limits some of the funky thinks you can do with DAX and Q&A.
3rd thing, PBI isn`t truly 3 tier yet, to use the tool properly you have to install and maintain a client due to some of the current limitations (e.g. only desktop can consume a PBI dataflow).
That said, I`m a massive fan, mainly due to the power of DAX and the M language.
What I`ve found though that that power sometimes gets put in the hands of the end user with their datasources all over the place, so you just end up with a data governance nightmare.
I could actually see a lot of usecases for supporting both in a business to be fair!
ABILtd (BOB member since 2006-02-08)