On September 3, 2026, a two-seat vehicle with no steering wheel, no accelerator or brake pedals, and no rearview mirrors rolled onto the streets of Austin, Texas. The Cybercab, as it’s called, was designed from the outset to leave no entry point for human takeover. Elon Musk posted a single line on social media: “The first car purpose-built for unsupervised full self-driving hit the road today.”<\/p>\n
A month later, the gap between Cybercab’s commercial operating data in Texas and Musk’s grand vision is impossible to ignore. Public records from the Texas Department of Motor Vehicles show that on Cybercab’s launch day, Tesla had 420 autonomous vehicles authorized for commercial operation in Texas\u2014only 45 of which were Cybercabs, with the remaining 375 being converted Model Ys. By early October, Cybercab authorizations had risen to 169, converted Model Ys to 420, bringing Tesla’s Texas fleet to roughly 589 vehicles. But against Waymo’s operational scale of more than 4,000 vehicles, that’s still a rounding error.<\/p>\n
The first month of operations was hardly smooth either. Multiple paying passengers reported on social media that wait times initially exceeded 45 minutes, and issues with inaccurate pickup\/drop-off locations and occasional malfunctions of the butterfly doors and trunk were repeatedly mentioned.<\/p>\n
A decade-long bet on a technology path<\/p>\n
To understand Tesla’s catch-up logic in the Robotaxi race, one must go back to a pivotal decision Musk made ten years ago. In 2015, while testing Autopilot on Interstate 405, faded lane markings caused the vehicle to swerve suddenly, nearly colliding with oncoming traffic. After his anger subsided, he rejected engineers’ recommendations to add lidar, insisting that Tesla needed only optical sensors.<\/p>\n
According to Electrek, Tesla’s internal engineering staff had explicitly opposed removing radar, concerned that cameras would be compromised in rain, direct sunlight, and other conditions, increasing collision risk. But Musk wasn’t betting that cameras were better than lidar\u2014he was betting that scale could solve what technology couldn’t. In 2021 and 2022, Tesla removed radar and then ultrasonic sensors, fully pivoting to its “Tesla Vision” camera-only approach. At Autonomy Day in 2019, he put it even more bluntly: “Lidar is a fool’s errand.”<\/p>\n
While Musk was placing his bet on cameras, the rest of the global auto industry was making a different choice. Volkswagen sought external funding for its autonomous mobility unit MOIA, and after investing approximately \u20ac2 billion (approximately $2.3 billion), Uber and Lyft had exited negotiations. Hyundai Motor, meanwhile, is manufacturing Robotaxis for Waymo, with Hyundai CEO Jos\u00e9 Mu\u00f1oz stating that the vehicle manufacturing business is already profitable. Traditional automakers are betting on division of labor; Musk is betting on vertical integration. The former generates profits faster; the latter is harder to replicate.<\/p>\n
In 2024, Tesla released FSD V12. Horizon Robotics VP and Chief Architect Su Qing likened the moment to “Oppenheimer proving the atomic bomb could be built.” FSD V12 demonstrated that the end-to-end autonomous driving paradigm could actually work. In April 2026, Tesla pushed FSD V14.3, removing the final 300,000-plus lines of hand-written C++ code from the vehicle control layer and switching to a “full-domain single-stage pure end-to-end” architecture where raw camera pixel signals directly output steering, braking, and throttle commands.<\/p>\n
In January 2026, Nvidia CEO Jensen Huang unveiled Alpamayo, an end-to-end autonomous driving AI model, at CES. Asked about Tesla’s inference-based driving, Huang said: “I wouldn’t be surprised\u2014I think Tesla’s autonomous driving stack is the most advanced in the world.” Musk responded in writing: “They’ll find that achieving 99% of the technical goal is trivially easy, but cracking the remaining long-tail problems is extraordinarily difficult.” He also judged that Nvidia’s new model “probably won’t pose competitive pressure to Tesla for another five or six years.”<\/p>\n
The key milestones of this decade-long gamble can be condensed into a timeline:<\/p>\n
Dual pressure from scale and valuation<\/p>\n
Musk is betting on Robotaxi not because he enjoys risk, but because Tesla’s valuation logic demands it. In 2025, Tesla delivered 1.636 million vehicles globally, down approximately 8.6% year-over-year\u2014the first annual delivery decline in the company’s history. In Q2 2026, Tesla delivered 480,100 vehicles, up 25% year-over-year and a record for the quarter, with total revenue of $28.236 billion, up 26%. But operating profit was just $398 million, down 57% year-over-year, and operating margin fell from 4.1% a year earlier to 1.4%.<\/p>\n
Sales are growing; profits are vanishing. R&D expenses jumped from $1.589 billion to $2.371 billion, up 49% year-over-year. That money went to autonomous driving, Robotaxi, chips, and training compute. The growth curve of the car-selling business has peaked, and Tesla needs a second growth curve. If FSD is merely “assisted driving,” it can only be sold for 64,000 yuan (approximately $9,500) per unit, with a take rate below 2% in China. If FSD becomes “unmanned driving,” it can be monetized per mile, per subscription, and per ride-hailing trip. The same technology, under different business models, can differ in value by orders of magnitude\u2014and Robotaxi is the vehicle for that transformation.<\/p>\n
The deeper reason is that Tesla’s valuation logic is largely built on the imagined potential of Robotaxi. If Tesla were merely an EV company, its valuation should track Toyota or Volkswagen. But Tesla’s market cap far exceeds traditional automakers because capital markets price it as an “autonomous driving platform.” So Musk isn’t betting on a business\u2014he’s betting on Tesla’s entire valuation thesis. Car sales have peaked, FSD needs to monetize, and Robotaxi is the only exit.<\/p>\n
Capital markets’ patience is also wearing thin. Tesla’s stock has fallen about 18% since the start of 2026, making it the only mega-cap tech stock that has failed to deliver positive returns to investors this year. Q3 delivery data released on October 2 beat expectations\u2014486,500 vehicles delivered versus the 462,000 consensus\u2014but still declined 2.1% year-over-year. The stock rose nearly 5% in a single day following the announcement.<\/p>\n
Tesla’s actual Robotaxi fleet operating data in 2026 is far removed from the “rapid expansion” narrative. The fleet reached 200 vehicles in January 2026, but by late May, third-party monitoring platform data showed that driverless Robotaxis had dropped to roughly 20 vehicles. That meant on Austin’s streets, a Tesla Robotaxi passed by only once every 20 minutes on average, while Waymo’s fleet in the same city was 20 times larger.<\/p>\n
However, by early September when Cybercab launched, crowdsourced tracking platform Robotaxi Tracker showed that driverless operations across Austin, Dallas, and Houston had rebounded to roughly 200 vehicles.<\/p>\n
Even as Tesla’s Robotaxi fleet shrank from 200 to 20, other global automakers were turning Robotaxi into a manufacturing business. Under the current plan, Hyundai will deliver its first IONIQ 5 Robotaxis to Waymo in Q4 2026, with plans to produce “tens of thousands” of vehicles for Waymo going forward. Hyundai CEO Mu\u00f1oz put it bluntly: “Competitors chose to write down assets; we decided to maximize the use of existing assets.”<\/p>\n
As of May 2026, Tesla’s FSD Supervised cumulative driving mileage surpassed 10 billion miles. Tesla claims FSD Supervised’s major collision rate is one per 5.3 million miles, significantly better than the human driver rate of one per 660,000 miles. But Tesla’s 10 billion miles are “supervised” miles, with a human driver in the seat the entire time. Waymo’s 271 million miles are “unsupervised” miles\u2014fully driverless. Data measured under different definitions cannot be directly compared.<\/p>\n
Uncertainty under the regulatory shadow<\/p>\n
Musk can decide what sensors a car uses and when to push a new software version, but in Washington, there’s an agency with more authority than him\u2014the National Highway Traffic Safety Administration (NHTSA). NHTSA doesn’t care about Tesla’s technology path or the theoretical advantages of the end-to-end paradigm. It cares about one thing: whether FSD can detect pedestrians ahead in low-visibility conditions.<\/p>\n
Since March 2026, the answer to that question has pointed in one direction. On March 18, NHTSA escalated its investigation into approximately 3.2 million Tesla vehicles to engineering analysis (case number EA26002), covering 2016\u20132026 Model S and Model X, 2017\u20132026 Model 3, 2020\u20132026 Model Y, and 2023\u20132026 Cybertruck. The investigation involves nine crashes where FSD was engaged in low-visibility conditions, including one in Arizona where a 71-year-old woman helping direct traffic was struck and killed by a Model Y traveling at high speed in FSD mode. At the time of the crash, direct sunlight caused camera glare, and the system failed to detect the pedestrian ahead.<\/p>\n
NHTSA’s core findings point directly at the soft underbelly of the camera-only approach. In the crashes it reviewed, FSD “failed to detect common road conditions that impair camera visibility” and only issued alerts moments before impact. In May 2026, Tesla submitted 207 crash reports to NHTSA\u2014a single-month record. Since 2019, the cumulative total has reached 3,763, accounting for roughly 85% of all ADAS crash reports industry-wide. Tesla redacted 99.9% of the descriptions in those 3,763 crash reports, citing trade secrets.<\/p>\n
NHTSA also holds a document more explosive than the crash data. Tesla engineers once wrote an internal memo with a two-word title: “Radar Saves Us.” NHTSA has issued 24 information requests to Tesla. The core of U.S. product liability law isn’t whether a defect exists\u2014it’s what the manufacturer knew and when they knew it. If the memo shows that Tesla engineers had concluded before radar removal that the radar-camera combination was safer in low-visibility conditions, it ceases to be a technical footnote and becomes an evidentiary chain.<\/p>\n
After Cybercab launched in Austin, NHTSA promptly initiated what it called an audit inquiry to ensure the vehicle complies with federal safety standards. NHTSA had originally required Tesla to respond to a long list of safety questions by September 30, but according to CNBC, the company received an extension.<\/p>\n
Public sentiment is also building. A September 2026 survey by polling firm Slingshot Strategies found that 50% of American respondents said they would be unwilling to ride in a robotaxi without a steering wheel or pedals; after hearing about NHTSA’s audit of Cybercab, 70% said they wanted Tesla to suspend passenger operations.<\/p>\n
Regulatory pressure isn’t limited to the federal level. Austin Fire Department Captain Matt McElearney said that while Tesla has provided detailed vehicle handling guides to first responders, that isn’t enough. He argues that regulations should require autonomous vehicles without steering wheels and pedals to provide alternative means for public safety personnel to take control of the vehicle\u2014to move, steer, or shut it down\u2014in both emergency and non-emergency situations; vehicles should also undergo continuous testing throughout their full product lifecycle.<\/p>\n
In California, Waymo vehicles are a common sight across multiple cities, but Tesla has yet to obtain a permit for driverless operations on public roads. Musk said at the World Economic Forum in Davos in January 2026 that Tesla’s Robotaxi would be “very, very widespread” in the U.S. by the end of 2026\u2014a goal that has yet to materialize.<\/p>\n
Waymo has never missed an opportunity to needle Musk on the technology path. In August 2026, Waymo co-CEO Dmitri Dolgov played a comparison video at Y Combinator: in a Phoenix dust storm, lidar clearly identified pedestrians on the roadside while the camera feed was nearly a wall of chaos. Dolgov said one thing: relying solely on “weak perception” solutions like cameras will quickly hit a safety ceiling. He also emphasized: “L2 to L4 is a qualitative leap\u2014they’re not the same thing.”<\/p>\n
Waymo’s operational data does indeed constitute pressure. As of October 2026, Waymo operates in 15 U.S. markets, with another 15 cities in preparation, and has announced plans for international expansion to London, Tokyo, and Munich. In Texas, Waymo has 1,154 autonomous vehicles authorized for commercial use, including 359 of its new Ojai model. Waymo provides more than 500,000 paid rides per week and has accumulated 270 million miles of fully driverless commercial driving in the U.S.<\/p>\n
The Tokyo plan has already entered a concrete phase. On September 15, 2026, Waymo signed a strategic partnership with GO, Japan’s most-downloaded taxi-hailing app, and Nihon Kotsu, Tokyo’s largest taxi operator, aiming to launch Japan’s first Level 4 fully driverless commercial taxi service by 2027, gradually scaling to 100 vehicles. The three companies have been conducting road tests in seven of Tokyo’s wards since 2025. However, Japan’s regulatory bar is not low: since the revised Road Traffic Act took effect in April 2023, creating the “specified autonomous operation” permit system, the Ministry of Land, Infrastructure, Transport and Tourism reported only 11 locations and 12 permits as of June 2026\u2014still far from the government’s target of 50 locations.<\/p>\n
A comparison of the two companies’ operations as of early October 2026:<\/p>\n
MetricTesla RobotaxiWaymoTexas commercial operating vehicles589 (420 converted Model Y + 169 Cybercab)1,154 (including 359 Ojai)U.S. commercial fleetNot disclosedOver 4,000Weekly paid ridesNot disclosedOver 500,000Cumulative miles10 billion miles (Supervised, driver in seat)270 million miles (fully driverless)Operating marketsThree Texas cities; Cybercab limited to Austin15 U.S. markets, 15 more in preparationCalifornia driverless permitNot obtainedOperating in multiple citiesInternational expansionNoneLondon, Tokyo, Munich<\/p>\n
The two mileage figures measure different things: Tesla’s 10 billion miles come from “FSD Supervised,” with a human always in the driver’s seat; Waymo’s 270 million miles are entirely driverless.<\/p>\n
On the July 2026 earnings call, Musk introduced the concept of the “five-nines reliability journey.” Achieving 99.999% reliability is the sole constraint on scaled deployment\u2014a statement that is itself an admission of the bet. At the Tel Aviv Smart Mobility Summit, he said: “Ten years from now, about 90% of all miles people travel will be completed autonomously. Driving yourself will become a very niche activity.”<\/p>\n
If that prediction holds, the gap between Cybercab’s initial 45 vehicles and Waymo’s 4,000 is merely a moment in history. But between Austin’s 45 vehicles and city streets around the world lies not just technology, but time, regulation, and the extreme edge cases that end-to-end models have yet to learn to handle. From that 2015 decision to reject lidar, every step Musk has taken has been a bet on the same thing: that scale can solve what technology cannot. The 10-billion-mile data threshold, the five-nines reliability standard\u2014these are wagers, not answers.<\/p>\n","protected":false},"excerpt":{"rendered":"On September 3, 2026, a two-seat vehicle with no steering wheel, no accelerator or brake pedals, and no…\n","protected":false},"author":2,"featured_media":456148,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[30513,132,134,133,46960,8892,115870,156385,106910,30584,483,27,10020],"class_list":["post-456147","post","type-post","status-publish","format-standard","has-post-thumbnail","category-austin","tag-alphabet","tag-austin","tag-austin-headlines","tag-austin-news","tag-cybercab","tag-elon-musk","tag-fsd","tag-hyundai-motor","tag-jensen-huang","tag-nhtsa","tag-tesla","tag-texas","tag-waymo"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/posts\/456147","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/comments?post=456147"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/posts\/456147\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/media\/456148"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/media?parent=456147"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/categories?post=456147"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-tx\/wp-json\/wp\/v2\/tags?post=456147"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}