On one hand the labs say that they can't keep up with demand for tokens.
On the other hand there is an entire ecosystem built around figuring out which magic words will make LLMs output fewer tokens.