... except that it still pretty much requires Nvidia hardware. Maybe not for edge inference, but even inference at scale (ie. say at companies, or governments) will still require it.