I actually literally think that these small scripts, if widely applied, are far, far more bigger than that. Even solely in the US, let alone globally. I'm also relatively certain that, if they weren't sinking money into research etc, subscriptions on inference are probably already profitable. These companies are burning money because they are investing in research, not because $10/month doesn't cover the average inference costs. Although I'd love to find a better source than the speculative ones I've seen about it.