Comment by WarmWash

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.