Google is cursed by its own success. Any product they release will always be on shaky ground when compared to their advertising and search money-printing monopolies.
At this point, they should give up and settle into their role as the 21st century IBM.
This is not a Google Project, it's an Alphabet 'Moonshot' designed to maybe shake things up, create a market, discover new solutions/opportunities/induce investment with a very high risk profile and no obviously commercial outcome i.e. quasi NGO stuff that VC's won't touch.
There's no way to compare Loon to a regular commercial venture.
It's just stuff the founders want to do and that's fine.
IBM builds their own line of processors and operates some of the biggest supercomputers (e.g. Summit at Oak Ridge). I know at least one guy who works on a pretty ambitious project at Google, and last I talked to him, he lusted after the insanely crazy things he used to get to do in an IBM-internal dev shop.
At this point, they should give up and settle into their role as the 21st century IBM.