Would you want to run a $100MM/yr business that costs $120MM/yr to run?
What if you have a product that brings a great deal of value to many people, but for various reasons you can't collect much revenue on it. You could be operating at a loss, but still producing great value. In that case, if there were a way to cover that $20MM/year gap (corporation sinks profits from other divisions into it, for instance) it could be worthwhile.
Or perhaps that business that's losing money brings in customers who go to the other divisions and make them profitable. That is, its return is indirect and hard to measure (though not necessarily hard to verify).