Apple is probably also waiting for increased RAM supply to start making high-quality local inference a viable mainstream option.