Tesla widened its unsupervised robotaxi service to cover the entire Austin metropolitan area on Wednesday, more than doubling the operating zone and opening up highway driving on Interstate 35 along with suburbs like Pflugerville and Manor. The company’s official Robotaxi account announced the change on X, and Ashok Elluswamy, Tesla’s vice president of AI software, confirmed that driverless rides are now available across greater Austin. The expansion went live at 10:00 AM local time.

The map got bigger. The fleet did not. Independent tracking data puts Tesla’s active unsupervised fleet at roughly 20 vehicles, a number that has been falling, not rising, over the past two months. That is the tension buried inside an announcement built to look like momentum: Tesla is asking 20 or so driverless Model Ys to serve a metro area of more than 2.5 million people, an operating zone now larger than the patch Waymo covers in the same city with a far bigger fleet.

I have been tracking the gap between Tesla’s robotaxi promises and its deployed reality since the Austin service launched in June 2025. This is the fifth or sixth geofence expansion in a year. It is also the one metric where the program has moved fast, because drawing a bigger boundary on a map costs nothing close to what putting more certified driverless cars on the road does.

The Geofence Doubled While the Driverless Car Count Fell

Tesla’s unsupervised zone now matches its full Austin service area for the first time, after roughly a year in which driverless rides were confined to a small slice of South Austin. According to data from the Robotaxi Tracker cited by Electrek, the active unsupervised fleet sits near 20 vehicles, down from a peak of about 25 in late April.

Tesla has never published an official fleet count. The numbers that exist come from trackers watching vehicles on the street and from city records. A presentation by Austin officials, cited by Reuters, put Tesla at roughly 50 vehicles in the city, while Alphabet’s Waymo runs more than 250 in the same area. The Austin figure likely counts both supervised and unsupervised cars; the 20-vehicle number tracks only the driverless ones available to the public. Either way, the direction of travel for the fully autonomous fleet has been down.

Wider geography with fewer cars means one predictable outcome for riders: longer waits. The service has been running in Austin for close to a year, and customers already report wait times above 30 minutes during busy periods. Spreading the same small fleet across highways and outer suburbs does not shorten that.

Musk Has Pinned Real Scaling on an FSD Software Rewrite

The fleet is small by design, not by accident. Elon Musk has told investors that Tesla is holding back aggressive expansion until the FSD v15 software rewrite reaches the cars, a release he has framed as a major architectural change. That timeline pushes any meaningful ramp to late 2026 or early 2027.

Safety validation is the stated bottleneck. Musk has acknowledged to investors that it is the limiting factor on fleet size, and the math explains the caution. More cars mean more miles, and more miles mean more chances for an incident on a system Tesla is still refining. The company appears to have chosen to keep the driverless fleet small rather than scale into that risk, which is defensible on safety grounds even as it undercuts the growth story attached to Tesla’s valuation.

The deadlines have a history. Musk said last year that robotaxis would cover most of the United States by the end of 2025. By that deadline, Austin was the only city with a public deployment. Dallas and Houston launched in April 2026 with tiny geofences and, by tracker counts, have barely grown since.

Tesla Reversed Course and Unredacted Its Crash Narratives

For nearly a year Tesla was the only autonomous operator that hid the narrative section of every crash report it filed with the National Highway Traffic Safety Administration (NHTSA), labeling each one confidential business information. In May 2026 the company quietly reversed that policy and made the descriptions public.

The unredacted filings cover the 17 to 19 incidents logged in Austin since the launch, and they tell a more measured story than the raw count once implied. Most were low-speed contact events, and a meaningful share were not the fault of Tesla’s Full Self-Driving system at all, including several cases of the stationary Tesla being rear-ended by inattentive human drivers. No crash involved a serious at-fault impact, and the injuries on record are minor. As I detailed in EVXL’s earlier analysis of Tesla’s own crash data, context changes the picture that bare numbers paint.

Two incidents stand out, and not in Tesla’s favor. In both, the FSD system stalled, a safety monitor called for remote help, and the human teleoperator who took control then drove the car into a fixed object, a metal fence in one case and a construction barricade in the other. Both happened at low speed with no passengers aboard. They point at the part of the operation Tesla talks about least: the remote humans who step in when the software gives up.

Waymo Is Scaling Hardware While Tesla Scales Boundaries

The contrast with Waymo frames the whole story. Alphabet’s robotaxi unit now runs roughly 3,000 vehicles across several US cities, completes hundreds of thousands of paid driverless trips each week, and recently raised about 16 billion dollars to fund expansion into London, Tokyo, and additional American markets. It operates without safety monitors in the seats.

Tesla’s pitch has always been the inverse: a cheaper, camera-only approach that scales through software and existing vehicles rather than expensive lidar-equipped cars. That thesis can still pay off if the v15 rewrite delivers. The scoreboard right now favors the company that built the hardware and the operations to match its claims.

EVXL’s Take

A geofence is the easiest thing in this entire business to expand. You change a polygon. You do not certify a single additional car, hire a technician, or validate one new mile of software behavior. So when Tesla announces that driverless rides now cover all of Austin on the same day independent trackers show the active fleet sitting at 20 cars and falling, I read it as exactly what it is: a map update dressed as a scaling milestone.

This fits a pattern I have been writing about for over a year at EVXL. Back in December I covered Waymo’s roughly 100 billion dollar valuation as the market pricing the difference between autonomy that works and autonomy that is coming soon. In January I broke down the fatal flaw in Musk’s “5x value” self-driving math. The Austin expansion is the same story in a new costume. The boundary grows, the fleet does not, and the gap with Waymo widens.

To be fair to Tesla, the unredacted crash data did vindicate part of its case. The driverless system is not the menace the early redacted numbers let critics imagine, and shrinking the fleet rather than risking more crashes is the responsible call. Credit where it is due.

Here is my prediction. Tesla’s active unsupervised fleet in Austin will still be under 50 vehicles on January 1, 2027, regardless of how many more suburbs get added to the map before then. The geofence will keep growing because it is free. The car count will not, because it is not. Watch the fleet, not the boundary.

Source: Reuters (reporting by Juby Babu, editing by Pooja Desai).

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