> Still, in the end, open-weight models deter further AI capex.

Good.