No, because compute, not model ability, is the moat.

The second moat is convenience, which all the big labs make it (comparatively) easy to glide into their models.