[killjoywashere]

I submit the central problem is that a lot of things don't scale smoothly, and figuring out what doesn't scale costs money. Discovering how to transition from 10 to 100 is easy. 100 to 10,000 might be a 9 figure problem, but everything after 10,000 is vertical growth until you hit 10M.

The other issue is that some great ideas need $10M or $100M or $1B up front. Google's TPUs are probably a 9-10 figure investment, with zero dollars income until the whole system is deployed at scale. But it becomes a money printing machine after that.