The theory is that if you burn through a few (many) kilodollars doing it the inefficient way then you can at least prove the existence of a market for doing it properly ... and have a head start in actually having subscribers.

I see the theory but $100k/mo in AWS fees is madness. They could probably knock a zero off just by deploying in containers.