Tesla promised a robotaxi revolution. In Texas, that revolution currently runs on roughly 20 cars.
The company’s unsupervised autonomous fleet, vehicles operating on public roads without a human safety driver behind the wheel, has contracted to about 20 across the state, according to figures circulating among investors and industry trackers. Elon Musk has pointed to safety validation as the primary bottleneck, telling investors that the technology needs more proving before the fleet can grow. The admission is notable not just for what it says about Tesla’s timeline, but for how sharply it contrasts with years of ambitious promises from the company’s CEO.
It was back in 2019 that Musk predicted one million robotaxis on the road by 2020. In October 2024, he unveiled the dedicated Cybercab concept at a Hollywood studio event. By June 2025, Tesla had launched a supervised autonomous ride-hailing service in Austin, with safety drivers still in the front seat. The jump to fully unsupervised operation was supposed to follow quickly. Instead, the fleet appears to have shrunk.
What the filings actually say
Tesla’s most recent quarterly disclosure, a Form 10-Q covering the quarter ended March 31, 2026, identifies regulatory and legal risks tied to its autonomy programs as material to the business. The filing highlights potential safety issues, evolving regulations, and litigation exposure related to both advanced driver-assistance and automated driving systems. It does not break out a specific vehicle count for unsupervised operations in Texas, but the tone is cautious: the path from prototype fleet to commercial scale is framed as uncertain and subject to regulatory restriction, potential recalls, and even outright bans on certain automated features.
The roughly 20-vehicle figure itself does not appear in any public regulatory filing or Tesla document currently available. It has been referenced in investor discussions and industry analysis but lacks a single authoritative source. Readers should treat it as approximate until Tesla or a state agency confirms it in a public record.
Similarly, Musk’s remarks about safety validation acting as the constraint have been attributed to investor communications, but no specific earnings-call transcript, shareholder letter, or social media post has been publicly linked to the exact phrasing. The 10-Q acknowledges autonomy risks broadly without quoting Musk directly on validation timelines or numerical safety thresholds. The attribution functions as a paraphrase of management sentiment rather than a verbatim, citable claim.
The regulatory squeeze from two directions
Whatever Tesla’s internal reasons for keeping the fleet small, the external regulatory environment is not making expansion easy.
At the state level, the Texas Department of Motor Vehicles operates an automated-vehicle program that sets the authorization framework for commercial automated motor vehicles. The program includes enforcement provisions tied to May 28, 2026, establishing processes for operators to add or remove vehicles from approved lists and detailing complaint and investigation mechanisms. Any fleet change in Texas is a regulated action with a paper trail, not an informal business decision. Operators that reduce their disclosed fleet size after that enforcement date would logically face fewer crash-report submissions in subsequent quarters, simply because fewer vehicles are generating data on public roads.
At the federal level, the National Highway Traffic Safety Administration maintains Standing General Order 2021-01, which requires mandatory crash reporting for vehicles equipped with automated driving systems. Manufacturers and operators must submit data on certain crashes involving Level 2 and higher automated systems, with incident summaries made publicly available. Every unsupervised vehicle on the road generates data obligations and potential public scrutiny. A fleet of 20 produces far fewer reportable events than a fleet of 200, but each incident carries outsized attention when the total population is that small.
NHTSA also runs the AV TEST Initiative, a voluntary tracking tool where companies and states can disclose testing locations, approximate vehicle counts, and monitoring details. Participation is optional and disclosures can lag real-world operations. Whether Tesla actively updates its entries, and how its reported activity compares with competitors, remains unclear from publicly available data.
The Waymo gap keeps widening
The competitive context makes Tesla’s position harder to ignore. Alphabet’s Waymo has been operating a commercial, fully driverless ride-hailing service since 2020 and has expanded to multiple cities including San Francisco, Los Angeles, Phoenix, and Austin itself. Waymo reported serving over 150,000 paid rides per week as of late 2024, a figure that has continued to grow. Its fleet uses dedicated hardware, including lidar sensors, that Tesla has rejected in favor of a camera-only approach.
The philosophical divide between the two companies is well-documented: Tesla bets that vision-based AI trained on billions of miles of customer driving data will eventually surpass sensor-heavy systems at a fraction of the hardware cost. Musk has called lidar a “crutch.” But Waymo’s fleet is carrying paying passengers today at scale, while Tesla’s unsupervised program in Texas appears to be contracting. For investors weighing Tesla’s robotaxi narrative, that gap is the most concrete benchmark available.
General Motors shut down its Cruise autonomous vehicle unit in December 2024 after a pedestrian-dragging incident in San Francisco and subsequent regulatory fallout, removing one competitor from the field but also underscoring how quickly safety failures can derail an entire program. Amazon’s Zoox continues testing but has not launched commercial service. The competitive landscape is thin, but the companies still operating are doing so at scales that dwarf Tesla’s current Texas footprint.
What the fleet is actually doing
Details about the day-to-day operations of Tesla’s unsupervised vehicles in Texas remain sparse. It is not publicly clear whether these cars are carrying passengers, running empty validation loops to accumulate safety data, or some combination of both. Tesla’s supervised ride-hailing service in Austin, which launched in June 2025 with safety drivers present, operates separately and at a larger scale. The unsupervised program appears to be a distinct, smaller effort focused on proving the technology can function without any human in the driver’s seat.
The distinction matters because safety validation for unsupervised driving requires demonstrating reliability across a vast range of scenarios: construction zones, emergency vehicles, pedestrians behaving unpredictably, severe weather, sensor degradation, and software edge cases. Each scenario must be handled without the fallback of a human ready to grab the wheel. Accumulating enough data to satisfy both internal engineering standards and external regulatory scrutiny takes time, and doing it with 20 vehicles takes considerably more time than doing it with 200.
That math creates a feedback loop that could slow Tesla’s progress further. A small fleet generates less data. Less data means slower validation. Slower validation means regulators have less evidence to approve expansion. And without expansion, the fleet stays small.
What to watch through the rest of 2026
For anyone tracking whether Tesla’s Texas fleet will grow or shrink in the coming quarters, the most reliable signals will come from three places. First, changes in the TxDMV authorization records, which should reflect any vehicles added to or removed from Tesla’s approved list. Second, updates to NHTSA crash-report filings under the Standing General Order, which will show whether incident volume is rising or falling. Third, Tesla’s next quarterly disclosure: the 10-Q for the period ending June 30, 2026, should reflect any material changes in fleet size or regulatory status through revised risk language, new descriptions of pilot programs, or references to investigations.
Until those documents arrive, the gap between Musk’s public optimism about unsupervised robotaxis and the cautious tone of Tesla’s own filings will remain the central tension in this story. Twenty vehicles is not zero. But it is a long way from the million robotaxis Musk once promised, and the company’s own CEO is now saying the technology is not yet proven safe enough to scale. That candor, however belated, may be the most important data point of all.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
