What the Numbers Actually Say About Robotaxis in 2026
If you have been following autonomous vehicles, you have likely seen contradictory claims: robotaxis are either imminently ubiquitous or fundamentally unsafe. The truth, as usual, lives in the data. This guide compiles the most current statistics on driverless cars — miles driven, collision rates, service availability, pricing, and technology limitations — so you can evaluate the state of Level 4 autonomy without relying on marketing copy or fear-driven headlines.
We focus on the two services that have accumulated meaningful real-world mileage with fully driverless (no safety operator) vehicles in the United States: Waymo One (Alphabet/Google) and Cruise (GM-backed, currently in phased relaunch after its October 2023 suspension). Zoox (Amazon) and Tesla's announced robotaxi efforts remain either geographically restricted or not yet in paid public service as of early 2026.
Robotaxis like Waymo One operate at SAE Level 4: the vehicle handles all driving tasks within a defined operational design domain (geofenced area, specific conditions) with no human driver required. This is fundamentally different from Level 2 systems like Tesla Autopilot, Ford BlueCruise, or GM Super Cruise, which require continuous driver supervision. Level 5 — full autonomy everywhere, in all conditions — does not exist in production today. When you read "self-driving" in consumer car marketing, it almost always means Level 2 driver-assist, not Level 4 driverless.
Where Can You Actually Use a Driverless Robotaxi Right Now?
Despite years of expansion announcements, fully driverless robotaxi service remains limited to a handful of U.S. metro areas. Here is the current operational map as of early 2026:
| Service | City / Region | Service Area | Hours | Status (2026) |
|---|---|---|---|---|
| Waymo One | Phoenix, AZ (metro) | ~225 sq mi (Chandler, Tempe, Mesa, Scottsdale, parts of Phoenix) | 24/7 | Fully operational, open to public |
| Waymo One | San Francisco, CA | ~65 sq mi (most of SF proper) | 24/7 | Fully operational, open to public |
| Waymo One | Los Angeles, CA | ~80 sq mi (Santa Monica, Hollywood, Downtown, Mid-City) | Extended hours | Operational, expanding zones |
| Waymo One | Austin, TX | Limited initial zone (South Austin) | Limited hours | Launched late 2025, ramping |
| Cruise | Phoenix, AZ | Limited geofence | Daytime only | Phased relaunch with safety operators; driverless TBD |
| Zoox | Las Vegas, NV / Foster City, CA | Small test zones | Limited | Employee/invite-only; no paid public service |
The key takeaway: if you live outside Phoenix, San Francisco, Los Angeles, or Austin, you cannot currently hail a driverless robotaxi. Announced expansions to cities like Miami, Atlanta, and Washington D.C. are in testing phases — not live paid service.
Driverless Car Safety Statistics: What the Data Shows
Safety is the most contested question in autonomous vehicles, and the one where statistics are most frequently cherry-picked. Here is what the verified data actually shows.
- Total fully driverless miles driven: Over 15 million miles (cumulative across Phoenix, SF, LA)
- Police-reported collisions: Approximately 30-35 incidents across the fleet through late 2025
- Injury rate comparison: Waymo's 2024 IIHS-HLDI study found a ~57% reduction in police-reported crash rate per million miles vs. human benchmarks in comparable urban environments
- At-fault rate: The majority of Waymo collisions involved other human drivers rear-ending or sideswiping the robotaxi; Waymo was determined at-fault in a minority of incidents
- Most serious incident context: No confirmed passenger fatalities in Waymo driverless service as of early 2026
For comparison, the NHTSA FARS database records approximately 1.35 fatalities per 100 million vehicle miles traveled for human-driven vehicles in the U.S. Robotaxis operate almost exclusively in dense urban environments where crash rates per mile are higher but fatality rates per crash are lower (due to lower speeds). Direct statistical comparison requires adjusting for road type, speed, and urban density — something neither robotaxi companies nor their critics consistently do.
Cruise: The October 2023 Incident and Aftermath
Any honest safety discussion must address Cruise's October 2023 incident in San Francisco, where a Cruise vehicle struck and then dragged a pedestrian who had been thrown into its path by a human-driven car. The California DMV suspended Cruise's permit, the company pulled all vehicles nationwide, and GM eventually wrote down over $8 billion in Cruise-related losses. Cruise has since relaunched in limited capacity with safety operators, but fully driverless service has not resumed as of early 2026. The incident underscored a specific edge case: post-collision behavior when the vehicle's perception system does not fully understand the surrounding context after an initial impact.
Intervention and Disengagement Rates
California requires autonomous vehicle operators to report "disengagements" — moments when a safety driver takes control or the system alerts that human intervention is needed. For fully driverless operations (Waymo One in Phoenix and SF), this metric is less relevant because there is no driver to intervene; instead, the vehicle is designed to pull over safely (a "minimal risk condition") if it encounters uncertainty. Waymo has reported that its vehicles initiate precautionary pullovers roughly once every 5,000–8,000 miles in complex urban environments. These pullovers are not failures — they are the system recognizing its own limitations, which is arguably safer than a human driver pushing through uncertainty.
How Much Does a Robotaxi Cost vs. Uber and Lyft?
Pricing for Waymo One is comparable to UberX and Lyft in the same markets, with some notable nuances:
| Metric | Waymo One | UberX | Lyft |
|---|---|---|---|
| Base fare (typical 3-mile trip, SF) | $11–$16 | $12–$18 | $11–$17 |
| Surge pricing | Less aggressive; capped in most markets | Can exceed 2–3x base | Can exceed 2x base |
| Tipping | Optional in-app (no driver to expect it) | Expected / prompted | Expected / prompted |
| Wait time | 3–7 min average (geofence-limited fleet) | 2–5 min average | 2–6 min average |
| Route flexibility | Geofenced; cannot go outside service zone | Anywhere driver will go | Anywhere driver will go |
To use Waymo One, download the Waymo app, create an account, enter your pickup and dropoff within the service zone, and confirm. The vehicle arrives without a driver, you unlock it via the app, and it drives you to your destination. The experience is remarkably uneventful — which is the point. No conversation, no driver route preferences, no tipping pressure. For airport runs or trips outside the geofence, you still need Uber, Lyft, or a personal vehicle.
Technology Stack: How Robotaxis Actually See the Road
Understanding the hardware helps explain both the capabilities and the limitations.
Waymo 6th-Generation Driver System (Jaguar I-PACE platform)
- LiDAR: Multiple solid-state and spinning LiDAR units providing 360° 3D point-cloud mapping up to ~500 meters. LiDAR is the core sensor for object detection and distance measurement independent of lighting conditions.
- Cameras: 29+ cameras covering all angles, used for lane detection, traffic light state, sign recognition, and semantic understanding of the scene.
- Radar: 6+ radar units for velocity detection of moving objects, particularly useful in rain and fog where LiDAR and cameras degrade.
- Audio sensors: External microphones to detect emergency vehicle sirens.
- Compute: Custom-designed Waymo compute platform (not consumer GPU) processing sensor fusion in real time.
- HD Maps: Centimeter-precision pre-mapped routes within the geofence. The vehicle does not navigate unmapped roads.
Tesla's Camera-Only Approach (for contrast)
Tesla's FSD (Supervised) relies exclusively on cameras and neural network processing — no LiDAR, no radar (removed in 2021+ vehicles). Tesla's approach is designed for scalability across all roads, but it remains a Level 2 system requiring driver supervision. Tesla's announced "Cybercab" robotaxi has not entered paid public service as of 2026. The camera-only approach faces particular challenges in low-light, heavy rain, and sun-glare conditions where LiDAR-equipped vehicles maintain performance.
The fundamental trade-off: LiDAR-equipped systems (Waymo, Zoox, Cruise) are more robust in edge cases but more expensive per vehicle (estimated $100,000–$150,000+ in sensor and compute hardware). Camera-only systems are cheaper to scale but less reliable in challenging perception conditions. For Level 4 driverless operation, every current commercial service uses LiDAR.
Regulatory Landscape: Who Approves Robotaxi Operations?
Autonomous vehicle regulation in the U.S. is a patchwork of federal and state authority:
- Federal (NHTSA): Sets vehicle safety standards (FMVSS) but has not yet issued comprehensive regulations specific to Level 4 vehicles without steering wheels or pedals. NHTSA issues exemptions on a case-by-case basis (e.g., GM's petition for the Cruise Origin, which was ultimately withdrawn).
- State DMVs and PUCs: California's DMV issues testing permits; the California Public Utilities Commission (CPUC) issues commercial deployment permits. This dual-approval process is why Cruise's suspension came from the DMV while operational restrictions came from the CPUC.
- Arizona: The least restrictive regulatory environment. Waymo has operated commercially in Phoenix since 2020 under a permissive executive-order framework with minimal state oversight.
- Texas: Emerging framework allowing commercial deployment with local coordination. Waymo's Austin launch proceeded under a combination of state law and city agreements.
- Other states: Most states either have no AV-specific legislation or require case-by-case approval. There is no federal preemption — a robotaxi legal in Phoenix has no automatic authorization in Miami.
For consumers, this means service expansion will remain slow and city-by-city. There is no "flip a switch and go national" path for robotaxi companies.
Current Limitations: What Robotaxis Still Cannot Do
Operational Limitations (as of 2026)
- Geofencing: Vehicles will not operate outside pre-mapped zones. If your destination is 500 feet outside the boundary, the ride cannot be booked.
- Weather: Heavy rain, fog, and snow degrade LiDAR and camera performance. Waymo reduces service or pauses operations during severe weather events. Phoenix's dry climate is not a coincidence.
- Construction zones: Unmapped lane shifts, temporary barriers, and flaggers remain challenging. Vehicles may initiate precautionary pullovers or route around construction.
- Highway driving: Most robotaxi services avoid highways or limit highway segments. Waymo has begun limited highway-capable routes in Phoenix, but the core use case remains urban point-to-point at speeds under 45 mph.
- Precise pickup/dropoff: The vehicle will not pull into a driveway, parking garage, or private lot. Pickups occur on public streets, which can mean a short walk from your actual origin.
- Complex multi-point routing: No intermediate stops. You cannot ask the robotaxi to wait while you run into a store.
- Accessibility: Current Waymo vehicles (Jaguar I-PACE) are not wheelchair-accessible. Purpose-built vehicles like the Zeekr RT (Waymo's next-gen platform) and the Zoox vehicle are designed with level-floor entry but are not yet in widespread service.
- Night operations: Waymo operates 24/7 in Phoenix and SF, but performance in unlit areas with poor lane markings degrades. Some services restrict nighttime zones.
Frequently Asked Questions
Is this truly self-driving or just driver-assist?
Waymo One in Phoenix, San Francisco, and Los Angeles is genuinely driverless — SAE Level 4. There is no human in the driver's seat, no remote operator controlling the vehicle in real time (though remote assistance can provide high-level guidance in edge cases). This is categorically different from Tesla Autopilot, FSD (Supervised), BlueCruise, or Super Cruise, all of which are Level 2 systems requiring a licensed, attentive driver behind the wheel. If a system has a steering wheel and expects you to take over, it is not self-driving.
How many driverless cars are actually on the road?
Waymo's fleet is estimated at approximately 1,000–1,500 vehicles across all operational cities as of late 2025. Cruise's active fleet is significantly smaller during its relaunch phase. Zoox has a limited number of its purpose-built vehicles in testing. Combined, there are likely fewer than 2,000 fully driverless robotaxis operating commercially in the United States — a rounding error compared to the ~280 million registered vehicles nationwide.
Have there been any fatalities in robotaxi service?
As of early 2026, there have been no confirmed passenger fatalities in fully driverless Waymo One service. Cruise's October 2023 incident involved a serious pedestrian injury (not a fatality) but was the catalyst for the company's nationwide suspension. For context, human-driven vehicles kill approximately 40,000+ people annually in the U.S. (NHTSA FARS data). Robotaxi sample sizes remain too small to make definitive per-mile fatality comparisons, but the available data suggests urban collision rates at or below human benchmarks.
Can I use a robotaxi for airport trips?
In Phoenix, Waymo One serves Phoenix Sky Harbor Airport (PHX) with designated pickup and dropoff zones. In San Francisco, SFO service launched in 2024 with specific pickup areas. LAX service in Los Angeles remains limited. Airport trips are among the most popular robotaxi use cases because they are predictable, well-mapped routes.
Will robotaxis replace Uber and Lyft drivers?
Not in the near term. Robotaxis are constrained to geofenced urban zones, cannot handle complex routing or highway trips in most markets, and have fleet sizes measured in hundreds rather than millions. Uber and Lyft each facilitate over 20 million rides per day globally. Robotaxis will serve as a complement in specific high-density urban corridors for years to come, not a wholesale replacement. Uber has acknowledged this by partnering with Waymo to surface Waymo rides within the Uber app in Phoenix.
What happens if the robotaxi gets confused or stuck?
The vehicle is programmed to enter a "minimal risk condition" — typically a controlled pullover to the side of the road. Remote assistance (not remote driving) can then evaluate the situation and provide high-level guidance, such as confirming it is safe to proceed around an obstacle. The vehicle does not have a remote human "joystick" operator. If the vehicle cannot resolve the situation, it remains stopped and the rider can contact support or exit the vehicle.
Sources: NHTSA Automated Vehicle Safety, IIHS-HLDI, Waymo safety reports, California DMV autonomous vehicle program filings.
