Tesla's robotaxi program has now been running in Austin, Texas for nearly a year. U.S. regulators have just received the full, unredacted account of every incident that has happened during that time. Tesla had previously asked for those details to be hidden, citing business confidentiality. Now the picture is clearer, and it is more complicated than either critics or supporters have claimed.
Out of 17 reported incidents total, most are minor and many were not Tesla's fault at all. Several crashes happened because other drivers rear-ended the robotaxi while it was stopped at a red light. That is a pattern Waymo sees too. But two crashes stand out in a way that matters structurally. In both cases, the car's software got stuck and could not move forward. A safety monitor in the front seat called for remote help. A human operator, sitting somewhere at a control center, took over the car remotely and drove it into a metal fence in one case and a construction barricade in the other.
The remote operator is supposed to be the ultimate backup. When the car cannot handle a situation, a human takes control. The fact that the human also crashed, in controlled low-speed situations with a safety monitor present and no passengers on board, points to something worth paying attention to. Driving a car through a video feed with a slight delay and limited camera angles is genuinely difficult. Waymo uses a different model: their remote staff advise the car on a new route, and the car's software carries out the instructions. Tesla's operators physically steer and accelerate, which is a much harder task.
The scale gap between Tesla and Waymo makes this even more significant. In Austin, Tesla is running roughly 50 cars. Waymo has more than 250 in the same city. Waymo's national fleet is around 3,000 vehicles across 10 U.S. cities, completing about 400,000 rides per week. Tesla is in three Texas cities. In Austin specifically, reporters tracked availability over three weeks in April and found that no car was available at all in about one in four checks, and waits exceeded 15 minutes more than half the time.
The service is also cheap: Tesla is charging around 60 percent less per mile than Uber. That suggests the company is intentionally subsidising rides to collect driving data and build volume, not yet trying to run a real business. That is a reasonable strategy at this stage, but it also means the commercial case is still entirely unproven.
Musk predicted in July 2025 that Tesla robotaxis would cover half the U.S. population by the end of that year. He also predicted 500 vehicles in Austin alone by December 2025. The actual number at that point was closer to 35. The service is now in three cities, with plans for seven by mid-2026.
Waymo has its own problems. It has issued software recalls, faced investigations over school bus incidents, and still loses more than two billion dollars a year. But Waymo has driven over 170 million fully driverless miles and has peer-reviewed safety data showing it causes serious injuries at a fraction of the rate of human drivers. Tesla has driven around 800,000 miles in robotaxi mode, and its crash rate per mile is currently worse than average human drivers, based on Tesla's own benchmarks.
For industries that rely on transportation, logistics, or mobility, the practical question is simple: which of these services could you actually build a business dependency on today? The answer is not Tesla's. The gap is not primarily about crashes or safety incidents. It is about whether the service is dense enough, reliable enough, and consistent enough to be useful. Right now it is not, and the crash data helps explain why the company itself is moving slowly.