Self driving cars are one of the first “real” robots people will see. And by “real” robot, I mean something acting out in the real world, doing something that most people thought of as a human-specific skill.
Tesla is planning an invite-only launch event for its robotaxi service Sept 3, 2026. They seem to be preparing a great number of vehicles for the launch in Austin, TX. In addition, there have been sightings all over the United States, in cities like New York and Bellevue, Washington.
It’s an interesting moment, and hopefully marks the beginning of an explosion of genuine self-driving-car competition and expansion across many different cities. Time for a short blog post ahead of the Robotaxi launch.
Expansion Plans
Tesla’s robotaxi service right now is only in a few cities, but it seems like at the very least they’re thinking about rapidly scaling. Here are 25 robotaxis in Tampa, Florida; here are 17 in Bellevue, Washington. Anecdotally, I saw one in Pittsburgh this weekend, joining Pittsburgh’s occasional exploratory Waymo Ojai trials.
Seeing vehicles all over the country is promising. Tesla is clearly betting that their end-to-end-learning focused strategy will allow them to roll out much more quickly and to more cities compared to Waymo.
There’s some evidence this will work. Tesla Autopilot already deploys in a lot of countries; Wayve, which has a similar learning-focused strategy, once drove through 90 cities in 90 days. The question has largely been one of reliability, but Tesla has been focused on collecting driver-in-the-loop data with its own supervised robotaxis for a while now.
The new vehicle has Tesla’s brand-new HW4 self-driving compute, with 20 CPU cores at ~2.35 GHz, and upgraded cameras. It also seems like it will have a Starlink antenna, to allow data upload and remote interventions.
The Competition with Waymo
Unsurprisingly, Waymo has gone on the offensive, talking about how multiple sensors are required for safety. This makes sense; Waymo’s vehicle is almost certainly much more expensive than the Cybercab, and Tesla is aggressively vertically integrated. Waymo will be in a tough position if Tesla actually proves its ability to out-scale the Google spinoff.
Waymo runs its successful robotaxi service in 14 cities, having recently added San Diego, Denver, and Tampa as of the time of writing. The two still operate on wildly different scales, with Tesla starting its pilot service in only five cities in Texas and Florida:
The question is rate of progress. Tesla, again, is aggressively vertically integrated; they can iterate on vehicle production, design, and software much faster than Waymo. Their stack is simpler, and they have a lot less specialized sensors.
Will Tesla run out of steam, or will their much more minimalist, economical stack allow them to drive success rates up while scaling much more aggressively than Waymo? Or will Waymo continue their steady expansion, offering a high-quality and relatively intervention-free ride experience to more and more people?
It remains to be seen. We will find out more on Sept 3rd.
Meanwhile, let's hope Tesla provides more transparency into their safety incidents during the FSD scale up. It does feel like we are all lab rats in this experiment... it won't matter how fast they can scale their stack if riders and communities don't trust them to operate safely. https://electrek.co/2026/09/01/tesla-driver-assist-left-turn-batavia/
Finally, we’re there. I’m still thinking about buying a couple of these and just letting them go out and earn their keep. 😄