I had already applied all the best practices in terms of reducing the cardinality, removing unwanted columns and making sure that only the data required is being brought into the dataset. Even at this point the dataset size was consuming 90GB of memory in Azure Analysis Services. With the steps below I got my dataset size down to a whopping 37GB of memory!
I used the awesome tools from SQLBI.COM and DAX Studio to see which columns were consuming the most space, and because my dataset had currency converted values, this meant that the cardinality was very high. (The reason that I decided to store the currency conversion values, is when trying to do it on the fly in a large dataset it is very slow)
Two simple tricks led to a pretty nice reduction in size.