Tesla’s long-promised robotaxi network is far smaller than its rhetoric suggests. State filings made public on May 28 show the company has 42 autonomous vehicles registered in Texas, while a separate third-party tracker counts as few as 20 cars actually carrying passengers without a human safety driver. Both numbers land hard against Elon Musk’s prediction of 1,000 robotaxis on the road within months of launch.

The two figures measure different things, and the gap between them matters. One counts vehicles a company has registered with regulators. The other counts vehicles that have actually given rides in the past week. When I pulled up the Texas Department of Motor Vehicles database after the new law took effect, the contrast with Waymo was the part that stopped me: Alphabet’s unit has 577 vehicles registered in the same state, more than 13 times Tesla’s count.

This is the first time an official count for Tesla’s Texas fleet has existed in public records. Until now, every fleet-size claim came from Musk or from estimates. The state forced the disclosure, and the disclosure undercuts a year of optimistic guidance.

Texas DMV Filings Put the Official Count at 42

The 42-vehicle figure comes from registration data Tesla submitted to the Texas Department of Motor Vehicles under a new state authorization program that took effect May 28. It is the official, company-reported number, and it covers vehicles Tesla has registered to operate as automated vehicles in the state.

Bloomberg’s Kara Carlson, who first reported the count, noted that the same database shows Waymo with 577 registered automated vehicles in Texas. CNBC’s review of the records found Tesla also trails Avride, which registered 317 vehicles, and sits roughly level with Nuro at 47 and Amazon’s Zoox at 35. For a company that has positioned driverless cars as the core of its future valuation, ranking behind Avride in its home state is an uncomfortable data point.

Tesla launched its robotaxi pilot in Austin in June 2025 and later added Dallas and Houston. The 42 registered vehicles represent the entire Texas footprint nearly a year in.

Tracker Data Shows Only 20 Cars Actually Driving Unsupervised

The smaller, more revealing number is operational. Robotaxi Tracker, a third-party service that monitors Tesla’s active vehicles, counted just 20 unsupervised cars carrying passengers in the seven days leading up to May 26: 14 in Austin, 3 in Dallas, and 3 in Houston. That is down from a mid-May peak the tracker put at 39.

Registration and active operation are not the same thing. A company can register a vehicle and leave it parked. The tracker measures cars that actually moved riders, which is why its count runs below the official 42. The direction is what concerns me. In late April, the unsupervised fleet had reached 25 cumulative vehicles, and Tesla had added five cars in a single burst. That looked like the start of a ramp. Weeks later, the active count went backward.

The broader picture is worse. Counting Tesla’s supervised Full Self-Driving cars in the Bay Area, which run with a driver behind the wheel under a California charter-party permit, the total active ride-hailing fleet sits at 34 vehicles. The Bay Area operation, once the bulk of the program, has shrunk to single digits. I covered the regulatory reason for that California setup in March, when the state’s utilities regulator confirmed Tesla holds only a limousine permit there, not an autonomous vehicle permit. You can read that breakdown in EVXL’s coverage of the CPUC’s limo-permit confirmation.

Safety Validation Remains the Bottleneck on Growth

Tesla’s own executives have named the constraint. On the company’s Q1 2026 earnings call, Musk told investors that safety validation is the limiting factor for scaling the fleet and that robotaxi revenue would not be material to 2026 results. That is a notable climbdown from the launch-era framing of self-driving as nearly solved.

The safety record gives the caution some context. Tesla has reported 15 crashes to the National Highway Traffic Safety Administration since the Austin launch. Using the roughly 800,000 miles Tesla disclosed through early 2026, that works out to about one crash every 57,000 miles. Comparisons to human drivers vary by source and by which baseline you use, ranging from roughly four to nine times the human rate, so the multiplier should be read as a range rather than a single hard figure. Every one of those crash narratives remains redacted in the federal database as confidential business information, while Waymo, Zoox, and Aurora file readable descriptions. I walked through that transparency gap in detail in EVXL’s February analysis of the 57,000-mile crash rate.

A safety monitor still sits in many of these vehicles, presumably stopping incidents that never reach the log. The reported rate may understate how often the system needs help.

FSD v15 Is the Software Tesla Says Will Unlock Scale

Tesla has told investors that large-scale commercial deployment waits on Full Self-Driving v15, a complete architectural rewrite Musk has targeted for late 2026 or early 2027. The new software jumps to roughly 10 billion parameters from about 1 billion in the current build, a tenfold increase Musk first teased in August 2025.

In the meantime, Tesla is preparing launches in Phoenix, Miami, Orlando, Tampa, and Las Vegas on the current v14.3 software base, though the company expects those deployments to stay small. Morgan Stanley has forecast 1,000 Tesla robotaxis operating by the end of 2026. With the active unsupervised count currently at 20 and moving in the wrong direction, that target requires a fiftyfold expansion in roughly seven months.

Tesla self-certified its commercial robotaxi software as SAE Level 4 under the new Texas framework on May 28, and Musk shared video of Cybercab units leaving Giga Texas with no one aboard. The self-certification applies only to the purpose-built Cybercab fleet. Consumer cars running FSD Supervised remain Level 2 systems that require driver attention.

EVXL’s Take

Here’s the pattern. Every time Tesla’s robotaxi program gets measured by an outside party rather than described by Musk, the real number comes in a fraction of the promise. In February I wrote about the crash data the company tried to keep redacted. In March I covered California’s regulator stripping the “robotaxi” label down to a limo permit. Now Texas has forced out a fleet count, and it’s 42 registered cars against Waymo’s 577, with maybe 20 actually working.

I’ve been tracking this specific program since the Austin launch in June 2025, and the through-line is consistent: the deficit is operational, not just rhetorical. Waymo runs thousands of cars across more cities, with wider geofences and higher utilization, while Tesla’s active unsupervised count went backward this month.

Tesla will not reach 1,000 active unsupervised robotaxis in Texas by December 31, 2026. The fleet would need to grow fiftyfold from its current 20-car active count in seven months, and Tesla’s own executives have already told investors that safety validation, not ambition, sets the ceiling.

Sources: Bloomberg, CNBC, Robotaxi Tracker.

EVXL uses automated tools to support research and source retrieval. All reporting and editorial perspectives are by Haye Kesteloo.