I'm not sure that Waymo and Zoox represent strong competition. Are they still relying on high-resolution maps of a few selected cities? If so, that seems like it will have scalability issues.
Tesla FSD supervised has reached an awkward stage where it is not capable enough to completely remove the human, but capable enough that human interventions are becoming infrequent.
Even with a global fleet numbering in the millions of vehicles, Tesla is running into difficulties in collecting enough data to reliably improve the system.
Will Waymo and Zoox be able to overcome the same challenges with a much smaller fleet?