Tesla Starts Cybercab Production. The Hard Part Comes Next.
Tesla says it has started making its robotaxi. The cars still cannot drive themselves without a human watching, they crash at 4-8x the human rate, and the team that built them is largely gone.
Tesla Starts Cybercab Production. The Hard Part Comes Next.
Tesla has begun manufacturing its Cybercab robotaxi at Giga Texas, CEO Elon Musk confirmed Friday on X, fulfilling a production deadline he has promised and pushed back multiple times since 2024. The news landed alongside corroboration from Bloomberg and a wave of drone footage showing dozens of Cybercabs lined up at the Austin factory. For Musk, it was a moment to claim vindication. For anyone watching the actual state of the program, it is a stranger story.
Because the production start does not mean the product is working.
Tesla's robotaxi service has been operating intermittently in Austin and San Francisco since late 2025 with a fleet of roughly 200 vehicles, according to Electrek. Those cars are still supervised. A human sits behind the wheel, ready to take over. That is not a robotaxi in any meaningful sense. It is a beta test with a safety driver, and the data from that test is not flattering.
According to figures Tesla disclosed to NHTSA, the fleet has been involved in 14 crashes since the service launched, roughly one every 57,000 miles, Fortune reported. The national average for human drivers is one police-reported crash every 200,000 to 500,000 miles depending on how you count. By the most conservative federal measure, Tesla's robotaxis crash at four times the human rate. By NHTSA's full-count methodology, the figure is closer to eight times. Tesla is also the only autonomous vehicle operator that has fully redacted its crash data from public disclosure, a distinction that has drawn scrutiny from safety researchers and regulators alike.
The production ramp itself is also worth examining. Musk pledged at the 2024 Cybercab launch that full production would reach 2 million units per year, roughly 38,000 vehicles per week. That number was always aspirational. What Tesla has started now is initial low-volume assembly, not mass production. The company is targeting a Q2 2026 commercial launch, with unsupervised Full Self-Driving targeted for gradual rollout late this year, and meaningful revenue not expected until 2027, according to Tesla's Q1 2026 earnings call summary.
Three senior leaders responsible for the Cybercab program have left Tesla since February. Victor Nechita, the program manager, Thomas Dmytryk, who built the OTA and ride-hailing infrastructure, and Mark Lupkey, who led assembly, are all gone. Electrek reported that Tesla now has no original program managers remaining for any of its production vehicles. That is a staffing fact, not a verdict on competence, but it is not a detail that belongs in a story about a smooth production ramp.
There is also a regulatory structure worth explaining. The NHTSA typically caps autonomous vehicle exemptions at 2,500 vehicles per manufacturer. Tesla appears to have sidestepped this ceiling by designing the Cybercab to comply with all existing Federal Motor Vehicle Safety Standards independently. It is a clever argument, and it may hold. But it also means the Cybercab is being certified under rules written for human-driven cars, which safety advocates argue are not adequate for evaluating autonomous systems that make real-time decisions at speed.
Tesla is not alone in having a rough launch. Waymo, widely considered the leader in commercial robotaxi operations, recalled more than 1,200 of its driverless vehicles last year after faulty software caused them to collide with chains, gates, and barriers. The AV industry has a learning curve, and it is steep. But Waymo's problems emerged after it was already operating driverless at scale. Tesla's problems are emerging before it has gotten there.
The honest version of this story is not "Tesla starts making a robotaxi." It is: Tesla has begun manufacturing a car that does not yet do the one thing that makes it a robotaxi, the existing version of that car crashes far more often than a human driver, the team that built it is largely gone, and the commercial launch the company is promising for Q2 is still months away from the unsupervised capability it requires.
That is still a story. It is just a different one than Musk's tweet suggests.
