Earlier quoted context omitted.
This statement is probably all we're going to get out of a private company that's a long way from making a profit. Personally, I'd like to see data, because I don't really trust whatever twisted definition of "profitable" Altman prefers. Until then I'll take advantage of the $20 price tag, but remain skeptical about its sustainability.
Lets assume that it is not profitable now because inference is too costly. Do you think making inference 20% more efficient would make it profitable? Or 60%?
No idea, we don't even have ballpark numbers about that. If they ever truly & transparently become profitable, there will be billions to pay back to investors -- I don't think anyone treats OpenAI like a charity project. At that point I highly doubt the subscription will still be $20.