Robotaxi Statistics 2026: Rides, Fleets, Safety

Half a million paid rides a week. That is how often people use Waymo robotaxis, according to a recent analysis. Baidu's Apollo Go most recently reported 3.2 million fully driverless rides in one quarter. Tesla, meanwhile, has a wide gap between registered vehicles and the fleet actually in service.

The figures sound clear. Their definitions often are not. I separate paid rides from tests, vehicle registrations from active fleets, and company reports from independent analysis. The research cutoff is October 7, 2026.

1. The market at a glance

This is not a ranking. Waymo's paid weekly rides, Baidu's fully driverless quarterly rides, and the fleet figures have different denominators. Even company-reported metrics are not automatically comparable.

2. Waymo: half a million rides a week

TechCrunch reported an average of 500,000 paid robotaxi rides per week in March 2026. Waymo's own March 31 post said it was serving more than half a million trips a week. TechCrunch reported 500,000 paid rides per week again on September 24. These reports do not establish further growth between March and September. Waymo's December 10 year-in-review initially said it had served more than 14 million trips so far. A year-end note on the same page puts the 2025 total at 15 million, more than triple 2024.

Sep '26:500K+0.0%vs. Mar '26

Sources: TechCrunch et al.

PNG|SVG

CC BY 4.0

gradually.ai

Sources: TechCrunch, Mar 27, 2026, Waymo, Mar 31, 2026, and TechCrunch, Sep 24, 2026

The 50,000 rides in May 2024 and the 500,000 in March 2026 come from TechCrunch reports based on Waymo statements. The September report repeats the same weekly figure. It does not establish additional growth between March and September. Waymo's year in review distinguishes more than 14 million trips by December 10 from the later year-end figure of 15 million.

Where does Waymo operate?

Waymo counts 15 US markets, but access differs. Its FAQ lists 12 cities where rides are available right away through the Waymo app. In San Antonio, Waymo says it is welcoming its first riders. Rides in Austin and Atlanta are booked through Uber. Las Vegas opened to public rides on September 14, 2026. Waymo's launch post names Boulder Junction, the Strip up to Sahara Avenue, and connections between Spring Valley and Winchester. It says access will expand gradually.

In February 2026, Waymo announced a $16 billion funding round at a $126 billion post-money valuation. That is a company valuation, not revenue or profit.

3. Safety: what the data says

Waymo's Safety Hub reports 271.3 million driverless miles through June 2026. Its comparison measures crashes over the same distance in Waymo operating areas against an adjusted human-driver benchmark. The percentages are Waymo's own data.

Crashes with serious or worse injuries

−95%

Pedestrian injury crashes

−93%

Cyclist injury crashes

−86%

Crashes with airbag deployment in any vehicle involved

−82%

Crashes with reported injuries

−82%

Source: Waymo Safety Impact Hub

PNG|SVG

CC BY 4.0

gradually.ai

Source: Waymo Safety Impact Hub, data through June 2026

Waymo reports 95% fewer crashes with serious or worse injuries, 93% fewer crashes with pedestrian injuries, and 86% fewer with cyclist injuries. These are relative comparisons based on company data. The serious-injury category is not a separate fatality rate.

The reporting bases are not identical. Waymo analyzes vehicle data under NHTSA's Standing General Order, which captures physical contacts alleged to cause property damage, injury, or fatality. Human benchmarks rely mainly on police reports. Waymo says it applies a 32% underreporting adjustment to the human benchmark only for crashes with reported injuries. Its serious-injury-or-worse and airbag benchmarks do not receive that adjustment.

An independent IIHS study published July 23, 2026, measured police-reportable crashes per vehicle mile. In the combined Waymo dataset for San Francisco, Phoenix, Los Angeles, and Austin, the rate was 68% lower than for human drivers. By city, results ranged from 76% lower in Phoenix to 4% higher in Austin, where the sample was small. The data covered 2021 to 2024. IIHS itself notes that available data are insufficient for continuously monitoring a large-scale expansion. The study and Waymo's Safety Hub do not measure the same thing.

4. Tesla: registrations are not an active fleet

Tesla's Robotaxi support page lists limited service areas in Austin, Dallas, Houston, Miami, Orlando, and Tampa. In the sources checked, the company does not publish a total for vehicles carrying passengers each day.

A RoboTaxi Report daily sync of Texas DMV registration data counted 546 registrations for Tesla Robotaxi, LLC on September 26, 2026. That included 420 Model Ys and 126 Cybercabs. The report confirmed the same total through its daily sync on September 28. A community tracker cited by Electrek observed eight vehicles giving rides in Austin during the week after the Cybercab event on September 3. That time-limited observation is not an official fleet total and is not directly comparable with the later Texas registrations.

Tesla commercially deployed Cybercabs in Austin on September 3, 2026. The next day, the National Highway Traffic Safety Administration opened an audit query into the vehicle's self-certification. NHTSA said existing safety standards remain in force until changed. The agency is examining the certification. Its announcement does not determine that Tesla violated the standards.

5. Apollo Go: quarterly rides, not a cumulative extension

Baidu reports fully driverless operational rides for Apollo Go by quarter. This is a quarterly series, not a cumulative total, and it is not comparable to Waymo's paid weekly rides.

Q1 '26:3.2M-5.9%vs. Q4 '25

Sources: Baidu Q2 2025 et al.

PNG|SVG

CC BY 4.0

gradually.ai

Sources: Baidu Q2 2025, Q3 2025, Q4 2025, and Q1 2026

The series rose from 2.2 million rides in Q2 2025 to 3.1 million in Q3 and 3.4 million in Q4. Baidu reported 3.2 million for Q1 2026. Its Q2 2026 results did not include a new ride count. They put Apollo Go's global footprint at 28 cities and total driving distance above 350 million kilometers. More than 240 million kilometers were fully driverless, according to Baidu.

Baidu's Q2 report says Apollo Go had launched fully driverless commercial operations in Dubai, with rides also available in the Uber app. In Hong Kong, Baidu reported driverless testing on Airport Island. In Switzerland, it reported open-road testing with PostBus. These are company statements. They do not give a current total fleet or show per-vehicle profitability.

Baidu's Q1 report separately put cumulative rides provided to the public above 22 million as of April 2026. Its Q2 report did not update the total. This remains a dated cumulative figure, not a Q2 ride count.

6. Pony.ai, WeRide, Zoox, and Motional

Pony.ai reported a global robotaxi fleet of 1,975 vehicles and robotaxi revenue of $12.1 million in Q2 2026, up 691.2% year over year. The company said more than 2,000 vehicles were contracted with Uber for Europe. That is a contract, not a count of vehicles already deployed.

WeRide reported a global L4 fleet of about 3,400 vehicles as of July 31, including more than 1,800 robotaxis. Zoox began charging for rides in Las Vegas on August 10, 2026. A July 30 NHTSA exemption permits commercial deployment of up to 2,500 vehicles a year for two years. The cap is not Zoox's current fleet size. Motional started public rides in Las Vegas through Uber on March 13. Its launch announcement says the initial service had a vehicle operator and expects public driverless rides by the end of 2026. At the cutoff, that was still a target, not confirmed driverless service. Provider sources: Pony.ai Q2 2026, the WeRide report, and Zoox's Las Vegas fares.

Fleet figures need their own definitions:

Waymo's roughly 4,000 vehicles are a TechCrunch estimate from September 24, 2026. Pony.ai and WeRide reported their fleets themselves. Tesla's count is Texas registrations. Baidu and Zoox do not disclose a total fleet in the current sources checked. These figures do not make a direct ranking. Sources include TechCrunch, Pony.ai's quarterly results, the WeRide report, Baidu's Q2 results, the RoboTaxi Report TxDMV sync, and the NHTSA Zoox exemption.

7. Prices: a dated Bay Area comparison

Obi's analysis compares quoted app prices in the San Francisco Bay Area. It does not measure nationwide prices or completed rides.

For the first period, Obi simulated more than 94,000 searches from November 27, 2025, to January 1, 2026, as reported by TechCrunch. Obi's follow-up analysis used organic app searches from March 1 to April 14, 2026, and compared identical pickup and destination pairs. It reported average quote increases of about 18% for Waymo, 5% for Uber, 12% for Lyft, and 41% for Tesla. The earlier comparison did not publish a like-for-like average for Tesla. These percentage changes are not current ride prices.

8. Recalls and open reviews

The recall reports do not show whether every software remedy was complete by October 7. NHTSA documents the status when each filing was submitted. Read the standing-water recall, construction-zone recall, NTSB school-bus investigation, NHTSA's Tesla query, and GM's Cruise announcement.

9. Market forecasts are scenarios

Goldman Sachs Research estimates the global robotaxi market could reach about $415 billion in 2035. BCG publishes a range of 700,000 to 3 million vehicles in the global fleet for the same year. Revenue and fleet size are different measures. Neither figure is a market result already achieved. Both forecasts depend on assumptions about costs, regulation, and demand.

10. Europe and Germany

Pony.ai, Verne, and Uber began a commercial robotaxi service in Zagreb on April 8, 2026. Pony.ai called it Europe's first commercial robotaxi service. On September 10, the partners also began invited passenger tests without a driver on a 22-kilometer route. Pony.ai said the service had logged more than 200,000 operating kilometers and several thousand rides since its April launch. These are company figures. The driverless test rides are separate from the commercial phase that began first.

Supervised Uber rides powered by Wayve became available in London on September 3, 2026. A trained, licensed private hire driver remains onboard. Uber may offer a Wayve vehicle for a matching trip when one is available.

MOIA began a passenger trial in Hamburg on July 15, 2026, for selected pre-registered residents. The ALIKE project lists free rides for a closed user group, up to 20 prototype vehicles, and a safety driver or vehicle attendant. The mixed fleet includes VW ID. Buzz AD and HOLON Mover vehicles. This describes a trial, not public driverless taxi service.

In August 2026, Waymo announced preparations for a service in Munich. It plans to begin by manually mapping streets and testing with trained specialists behind the wheel. The company is aiming for public commercial robotaxi rides toward the end of 2027. That is a future target, not a launched service.

Sources for Zagreb include Pony.ai's launch announcement and its driverless test-ride update. Wayve and Uber document the London launch on September 3. MOIA documents the passenger-trial launch and ALIKE project conditions. Waymo describes the Munich plan.

11. Milestones from 2009 to 2026

The timeline separates company announcements, commercial rides, and tests. The NTSB report on the March 2018 Tempe crash identifies distraction by the safety driver as the probable cause and lists contributing organizational factors. GM did not shut down Cruise as a company in 2024. It announced that it would stop funding robotaxi development. Sources include the NTSB investigation, the Waymo FAQ on the 2009 project start, IIHS's Waymo history summary, Waymo's 2016 announcement and 2020 Phoenix launch, the California DMV statement on Cruise, GM's 2024 announcement, the Tesla Austin launch report, Waymo's year in review and funding announcement, NHTSA's Zoox decision, and the Tesla audit query.

12. Conclusion: growth with open data gaps

Waymo reports a high volume of paid weekly rides and offers ride-hailing in 15 US markets. Apollo Go reports millions of fully driverless rides per quarter and a presence in 28 cities. Pony.ai and WeRide publish fleet figures. Tesla's registration count shows how different a registered vehicle total can be from the fleet actually deployed.

Some gaps remain visible. Baidu gave no new Q2 2026 ride count. Tesla does not publish an active fleet total. Some safety data comes from providers, and pricing research uses app quotes from one region. Forecasts are scenarios. The available figures describe a fast-growing market, but do not yet form one consistent global leaderboard.

For figures on the AI industry, see my AI statistics. My article on AI agent statistics looks at how autonomous systems are scaling beyond the road.