Tesla's Robotaxi service just gained one hour of nighttime operation in Austin, but the reason it needed that hour back reveals a deeper constraint: the camera-only perception stack that Elon Musk has championed for years struggles with the exact low-light, low-contrast scenarios that lidar was designed to solve. ANALYSIS Fifteen months after launch, the service's operating window is narrower than it was on day one, the Cybercab fleet numbers a fraction of its chief rival's, and the technical limitation Musk himself described points directly at the architectural choice he has refused to revisit.
Why it matters
Tesla's valuation thesis increasingly rests on Robotaxi. The company's Q2 operating profit fell 57% year-over-year in its core automotive business2. Musk told the World Economic Forum in Davos in January that Tesla's robotaxis would be "very, very widespread" across the U.S. by the end of 20269. ◆ The gap between that forecast and the current operational reality in Austin is the central tension investors face.
The big picture
Tesla launched Robotaxi in Austin in June 2025, initially accepting ride requests from 6 AM to midnight3,8. The service later extended to 2 AM in September 2025, then pulled back to a 10 PM close last August across Austin, Dallas, Houston, Miami, Orlando, and Tampa. On October 3, 2026, Tesla pushed the Austin cutoff to 11 PM6. Musk explained the constraint: "The main thing we're trying to solve is making sure that we don't run over pets when they're hard to see at night. Literally trying to avoid grey kittens on grey tarmac in the dark".
That is, as Electrek's Fred Lambert noted, "a low-light, low-contrast detection problem, which is the textbook weakness of cameras and the exact thing lidar and radar handle". Musk has spent years dismissing those sensors. In April 2019 he declared, "Lidar is a fool's errand. Anyone relying on lidar is doomed". In August 2025, he claimed that "lidar and radar reduce safety due to sensor contention". Tesla removed radar from its cars in May 2021 and dropped ultrasonic sensors in 2022.
Meanwhile, the Cybercab fleet remains small. Tesla launched commercial Cybercab operations on September 3, 2026, with 45 vehicles initially approved, rising to 169 by early October. Waymo has over 4,000 driverless vehicles in commercial operations in the U.S. and 1,154 autonomous vehicles authorized for commercial use in Texas alone. Waymo has completed 270 million fully autonomous miles of commercial driving domestically. Reviews of the Cybercab have been mixed, with paying passengers citing long wait times, pickups and drop-offs in the wrong place, and technical difficulties with butterfly doors or the trunk closing improperly.
ANALYSIS Musk's own framing of the nighttime problem is striking. He did not describe a software tuning issue or a temporary calibration gap; he described a fundamental perception challenge, one where a camera's dependence on reflected light and contrast leaves it unable to reliably distinguish a low-contrast object from its background. Lidar, which uses laser pulses to determine distance and shape independent of ambient light, handles this by design4. Radar operates on radio waves with similar light-independence.
Tesla's AI chief Ashok Elluswamy said in September that Tesla would launch 24-hour driverless rides "around next month" and that Robotaxis will get FSD v15 this month. ◆ That timeline now sits alongside Musk's own admission that the pet-detection problem remains unsolved, creating an internal tension between the engineering leadership's public schedule and the CEO's description of a still-open technical challenge.
Regulatory pressure compounds the picture. NHTSA initiated an audit query after the Cybercab launch to ensure the vehicles comply with federal safety standards. The agency has also escalated its investigation of roughly 3.2 million Tesla vehicles to the engineering analysis phase, citing multiple fatal crashes where FSD failed to detect pedestrians in low-visibility conditions. Tesla obtained an extension to respond to NHTSA's safety questions, which were originally due September 30. The company still has not obtained a permit to operate without safety drivers on public roads in California.
What's next
The near-term test is whether FSD v15 can close the nighttime gap without a hardware change. Elluswamy's timeline puts that update in October. NHTSA's extended deadline for Tesla's safety responses will set the regulatory tempo for any expansion beyond Texas and Florida. ANALYSIS If the camera-only stack cannot solve the low-contrast detection problem Musk described, Tesla faces a choice it has deferred for years: add sensors it has publicly rejected, or accept permanent operational limits that its lidar-equipped competitors do not share. Rivian said this week that lidar now costs it "a few hundred dollars" per vehicle.