Is there evidence they are unambiguously making profit off inference when all the supply costs of that inference are considered? It does not seem likely.