Yes... but there were concerns previously that inference was so costly that the subscriptions/API billing weren't covering basic operating expenses. That's clearly not the case. People are willing to pay them enough that they can afford to run the models. That's a really positive sign.
