Yup! Big beautiful burning pile of cash. Though there’s some nuance with subsidizing inference (which is supposed to make them money long term) vs training (which we’re supposed to ignore I guess).