So now the question is just size?
That is pretty tough to answer without getting into specifics (how big is too big? do you need all the data you are including? etc)
You could look at working in ROLAP mode instead of MOLAP, which would save space in the cube, but make things slower (and increase space usage someone in the dw). Other than that, your only method I know of is getting rid of things you don't need.
I will do my best to answer this fully:
The issue is size. The data is correct, and the measures can be reported along with the dimensions in the same cube.
When I convert the cubes over from Analysis Services 2000 they are but 2 or 3 megs a piece. Unfortunately, when I create the cubes from scratch using Koen Verbeeck method, stated previously in this thread, the cube can easily climb to over 80 mb while only hooking up a few dimensions and I need to hook up many more.
Also, ROLAP option is not my first choice because this scenario will not be fast enough.