On September 4, Tesla opened its purpose-built robotaxi, the Cybercab, to public rides across Austin's entire service geofence. The two-seat vehicle has no steering wheel and no pedals — just butterfly doors and a screen — and it's the first time Tesla has put a car designed from the ground up for autonomous-only operation in front of paying riders.
Tesla paired the launch with a safety milestone: Ashok Elluswamy, the company's VP of AI Software, announced that Tesla's "Unsupervised" robotaxi fleet had passed one million cumulative miles without a human safety driver behind the wheel. Elon Musk framed the moment as the start of a manufacturing ramp, telling attendees Tesla plans to eventually build "several times more Cybercabs per year than all of our other vehicles combined."
The context Tesla didn't put on stage
A million driverless miles is a real milestone. It's also a fraction of what the category leader has already logged. Waymo, which has been running fully driverless rides since 2020, has passed roughly 200 million autonomous miles and now operates in more than ten U.S. cities with a fleet in the thousands. Tesla's launch is a single city, a geofenced service area, and a fleet size the company hasn't disclosed.
None of that makes the Cybercab launch meaningless — getting a purpose-built, steering-wheel-free vehicle into public service at all is a genuine engineering and regulatory achievement, and Tesla's approach (camera-only perception, no lidar) is a materially different bet than Waymo's sensor-heavy stack. But "1 million unsupervised miles" and "200 million autonomous miles" are not the same order of scale, and coverage that treats the Austin launch as Tesla "catching up" to Waymo is skipping the math.
What actually matters from here
The real signal to watch isn't the launch-day numbers, it's the expansion rate. Waymo built its lead over roughly six years of gradual, city-by-city rollout with a heavier, more expensive sensor suite. Tesla is betting that a cheaper, camera-only approach can scale faster once the software is proven — which means the next few quarters, not the launch event itself, are where this comparison actually gets decided.