Tesla Robotaxi Timeline 2016–2026: From Musk’s 1 Million Promise to Model Y Rides and Cybercab

In 2019, Elon Musk predicted Tesla would have more than one million robotaxis on the road by 2020. That did not happen. Seven years later, Tesla’s Robotaxi service is finally real — running in six U.S. metro areas — but the car that shows up is a Model Y, not the purpose-built Cybercab. Cybercab has entered production, is undergoing public-road engineering tests, and has carried its first employee riders at Gigafactory Texas. But as of this update, Tesla has not yet put a paying customer inside one. This page tracks exactly where each piece of that story stands, and keeps updating as it moves.

These are two different things wearing similar branding. Tesla Robotaxi is the ride-hailing service — open the app in one of six metros and hail a car. Cybercab is a specific, purpose-built vehicle — a two-seat autonomous car with no steering wheel or pedals, designed to eventually become a major part of that fleet. As of this update, the two haven’t fully merged: someone taking a Tesla Robotaxi today gets a Model Y, a normal production car adapted for driverless ride-hailing. Cybercab is real, in production, and being tested on public roads — but it is not yet the car that shows up when a customer hails a ride.

Six metros is not nationwide. Tesla also runs a separate ride-hailing pilot in the San Francisco Bay Area that California’s own regulator says is not an autonomous vehicle service — see the regulation section below.

Tesla’s own hardware roadmap complicates this claim. Musk has said Tesla taped out the AI5 inference chip in April 2026 — nearly two years later than Tesla’s original schedule — with sample and limited production possible within 2026, but high-volume AI5 production targeted for 2027. The next-generation AI6 chip is aimed at 2028, with some reporting citing a late-2026 tape-out and mass production as early as the second half of 2027 for AI6 specifically; timelines here have moved before and should be treated as targets. Because a 2026 production Cybercab predates AI5’s high-volume manufacturing, it should not be described by default as running production AI5 hardware — Tesla’s vehicle-specific documentation, not assumption, is the source to check before making that claim.

Tesla’s argument is that a camera-first platform, trained on data from millions of consumer vehicles, can scale at lower hardware cost than sensor-heavy competitors. Waymo’s approach adds sensor redundancy and dense mapping, betting that layered perception matters more than deployment speed in the near term. Neither approach is established as objectively superior by independent data; this page presents the strategies, not a verdict.

On July 30, 2026, Amazon-owned Zoox received the first-ever NHTSA commercial exemption, under the agency’s Part 555 process, for a purpose-built robotaxi with no steering wheel or pedals — capped at 2,500 vehicles annually over two years, and barred from private consumer sale. Zoox plans to begin charging fares in Las Vegas first. This matters for Cybercab because it shows the federal approval pathway a no-manual-controls vehicle needs to reach paid public service exists and has been used once — but Tesla has not received the same authorization, and Cybercab’s own regulatory approval should be tracked separately rather than assumed.

NHTSA regulates vehicle safety, defect investigations and any exemptions from federal manual-control standards; states and cities separately govern ride-hailing permits, insurance and road operations. Cybercab’s lack of a steering wheel and pedals raises Federal Motor Vehicle Safety Standards questions distinct from Model Y Robotaxi’s, which is why the Zoox exemption above is a relevant precedent to watch, not a guarantee Tesla receives the same treatment.

Tesla’s Robotaxi launch drew real, documented early problems: rider and bystander videos in the first weeks after the June 2025 Austin launch showed wrong-lane movements, abrupt braking and stops, speeding concerns, and awkward drop-off locations. NHTSA contacted Tesla afterward seeking more information, and Reuters reported Tesla sought confidentiality for parts of what it submitted — NHTSA did not ban or suspend the service. Separately, and on a different system entirely, NHTSA escalated a probe in March 2026 into consumer FSD (Supervised) covering roughly 3.2 million vehicles, over concerns the camera-based system fails to detect or warn about poor-visibility conditions; nine incidents including one fatality are under review, and this moved to an engineering analysis. That investigation is about personal vehicles running a driver-supervised feature, not about the Robotaxi fleet, and the two crash datasets should never be combined into one number.

Tesla vehicles are not always unattended by any human when something goes wrong. A car can have no one inside while still receiving remote assistance from a human operator for specific situations — that is different from a remote operator actively driving every mile. Where NHTSA documentation confirms remote-control involvement in a specific low-speed incident, it’s worth noting explicitly, because “driverless” does not mean “no human support anywhere in the system.”