Is China winning the robotaxi race?

Why does China look like the robotaxi leader right now?

China looks like a leader because it has three serious robotaxi companies, while the United States has one. That industrial depth is real, but in 2026 it did not turn into more riders.

Pony.ai reported 1,975 robotaxis worldwide on 30 June 2026. WeRide said it had more than 1,800 robotaxis on 31 July, inside a wider fleet of about 3,400 self-driving vehicles that includes buses and vans. Baidu runs Apollo Go in a footprint it puts at 28 cities. All three charge passengers and keep adding vehicles.

The weak spot is Baidu, China’s largest operator. Its rides dropped from 3.2 million in the first quarter to about 1 million in the second, after operations were cut back in some cities. One stumble at the national champion was enough to erase China’s claim to the most rides.

The result is a split. China leads in fleet building, manufacturing and the number of operators. Waymo leads in weekly paid rides, safety proof and steady execution.

Measure	Leader now	What we found
Current weekly ride volume	Waymo	More than 500,000 paid rides a week, far above any disclosed Chinese level
Cumulative public rides	Baidu Apollo Go	More than 23 million by June 2026, versus Waymo’s more than 20 million
Vehicle cost	China	Purpose-built cars around the price of a normal EV
Public safety evidence	Waymo	Detailed results across more than 270 million autonomous miles
Overseas expansion pipeline	China	More live foreign services, mostly in the Middle East
Overall profitability	Nobody	Commercial revenue remains far below total spending

If you want more recent data on this point, please see our latest autonomous vehicle market report.

Is China giving more robotaxi rides than the US after Baidu’s drop?

Probably not. Waymo now gives more driverless rides than any combination of Chinese numbers we can verify.

Baidu’s roughly 1 million second-quarter rides work out to about 77,000 a week, our own average over 13 weeks. Its peak week in March had passed 350,000. Waymo first said it reached 500,000 paid rides a week in March 2026, and Reuters reported in October that it was above that level. Waymo’s own October blog post simply says “hundreds of thousands” of rides a week.

Pony.ai and WeRide do not publish weekly ride totals. WeRide says its Chinese robotaxis averaged more than 21 rides per car per day in the second quarter, with a peak of 28. Even if every Pony.ai and WeRide robotaxi ran at that pace, the three Chinese companies together would only roughly match Waymo, and real use sits below that ceiling. A China lead on rides is now an assumption, not a finding.

Why did Baidu Apollo Go fall further behind Waymo in 2026?

Baidu fell behind because its ride volume collapsed for a quarter while Waymo kept growing. It still leads only on lifetime rides.

On Baidu’s August earnings call, chief executive Robin Li said Apollo Go delivered around 1 million fully driverless rides in the second quarter. He blamed “operational adjustments in certain domestic cities due to regulatory considerations” and said operations in those cities had begun to resume in August. He did not name the cities. The timing follows the April fleet failure in Wuhan and the permit pause that came after it.

Apollo Go has still provided more than 23 million public rides since launch, against Waymo’s more than 20 million paid trips. But current use is not close. Waymo carries several times more passengers each week than Baidu averaged in its latest quarter.

The definitions also differ. Waymo counts paid rides, while Baidu counts “fully driverless operational rides” and rides provided to the public. Baidu does not say how many were discounted or free.

Metric	Baidu Apollo Go	Waymo
Cumulative disclosed rides	More than 23 million (June 2026)	More than 20 million (October 2026)
Latest weekly level	About 77,000 on average in Q2 2026 (our calculation)	More than 500,000 paid rides (reported)
Public operating footprint	28 cities, including tests and pilots	More than 15 US cities, according to Reuters
Strongest advantage	Low-cost fleet expansion	Current usage and safety proof

Does China have the world’s biggest robotaxi fleet?

China probably still has the biggest combined robotaxi fleet, but the margin is thin and no Chinese company matches Waymo alone.

Pony.ai’s 1,975 robotaxis and WeRide’s more than 1,800 already add up to more than 3,775 vehicles. Some of them run abroad: WeRide says about 400 of its robotaxis are in the Middle East. Baidu has not given a current fleet count. Waymo has about 4,000 cars, according to a Semafor report relayed by Reuters in October.

Both sides plan big jumps. Pony.ai targets more than 3,500 robotaxis by the end of 2026. Waymo reportedly wants 8,000 by year-end and 20,000 by the end of 2027, after raising $5 billion of debt in October; Waymo has not confirmed those fleet targets. These are plans, not vehicles on the road.

If you want more recent data on this point, please see our latest autonomous vehicle market report.

Are China’s robotaxis really fully driverless?

Yes, the leading Chinese robotaxi services run with nobody in the driver’s seat, but only inside approved zones and rules.

Baidu says its fleets have covered more than 240 million fully driverless kilometers. Pony.ai runs driverless commercial Gen-7 vehicles in Guangzhou and Shenzhen. WeRide runs driverless services in Guangzhou and also in Abu Dhabi and Dubai.

The cars follow approved maps, roads, hours, weather limits and pickup rules. A robotaxi can drive itself across one district and still be unable to serve the next one. Many of the cities in company footprints host only tests.

Remote staff also help when a car gets confused. They usually approve a safe next step rather than steering. Those teams cost money, so removing the in-car driver does not remove every labor cost.

Is China expanding robotaxis across cities faster than Waymo?

China lists more locations, but Waymo is now opening full public services faster.

Baidu puts its footprint at 28 cities. WeRide says its self-driving businesses reach more than 60 cities in 13 countries. Those counts mix paid services with tests, mapping work, robobuses and small pilots.

Waymo’s count is narrower but deeper. Reuters puts it in more than 15 US cities. In October 2026, Waymo opened rides to anyone in San Antonio, San Diego and Tampa, and it now targets first riders in Madrid in 2027 and Toronto in 2028. Each Waymo city usually means a large public zone. Many Chinese overseas locations hold only a few trial cars.

Are Chinese robotaxis much cheaper to build?

Yes. China has the clearest cost advantage in robotaxis, and it comes from its car industry as much as from software.

Baidu priced the purpose-built Apollo RT6 at RMB250,000, about $37,000, at launch and later said the delivered cost had fallen below $30,000. Pony.ai said in April 2026 that it expects a full Gen-7 robotaxi, kit included, to cost below RMB230,000 by 2027. Neither company has published a newer achieved cost.

Waymo has never said what a complete robotaxi costs. The Ojai uses fewer sensors and a Zeekr body, with final integration in Mesa, Arizona, where production was ramping in October. Waymo is clearly cutting costs, but without a published price a direct comparison is impossible.

China’s supply chain makes its low figures believable. Baidu works with Jiangling Motors, Pony.ai with BAIC, GAC and Toyota, and WeRide with Geely’s Farizon. All use established car factories.

Robotaxi platform	Disclosed cost or capacity	What it tells us
Baidu Apollo RT6	RMB250,000 at launch; later reported below $30,000	China can build an L4 vehicle near mass-market EV prices
Pony.ai Gen-7	Target below RMB230,000 by 2027	The whole vehicle and driving kit are being costed together
WeRide GXR	Used for new launches such as the Madrid pilot	Manufacturing partners can add fleet capacity quickly
Waymo Ojai	Final cost undisclosed	Lower-cost design, now ramping in Arizona

If you want more recent data on this point, please see our latest autonomous vehicle market report.

Can China turn cheaper robotaxis into a profitable business?

Revenue is rising fast, but no Chinese robotaxi company is close to proving a profitable business.

Pony.ai’s robotaxi revenue reached $12.1 million in the second quarter of 2026, up 691% from a year earlier. Total revenue was $36.2 million, and the company still lost $45.4 million and used $44.0 million of operating cash. WeRide reported $34.2 million of total revenue and a $59.1 million loss; only $18.5 million came from all its self-driving businesses combined.

Pony.ai says its operations show “positive unit economics,” and Baidu earlier said Wuhan was nearing fleet-level breakeven. Those claims cover the direct cost of running cars in some cities. They leave out research, software and expansion spending. Baidu still gives no separate Apollo Go income statement.

Are low robotaxi fares proving demand or buying rides?

Riders are paying more often, but low fares still do part of the selling, and Baidu’s drop shows demand depends on keeping cars on the road.

Apollo Go has offered base fares as low as RMB4 on routes where a normal taxi might start near RMB18. A study of 336 Apollo Go users in Wuhan found price among the strongest reasons people chose the service.

Paid use is growing. Pony.ai said fare-charging revenue rose more than 800% year over year in the second quarter. WeRide said ride-hailing revenue grew about 140% from the first quarter, and rides per car rose 24%.

The harder test is still ahead. Demand looks durable when riders keep coming after discounts shrink. China has shown that people will ride; it has only partly shown what they will pay.

Who has the stronger robotaxi safety evidence, China or Waymo?

Waymo has the stronger public safety case by a wide margin.

Its September 2026 analysis covers more than 270 million fully autonomous miles through June. Against human drivers on the same roads, Waymo reported 95% fewer crashes causing serious or worse injuries and 82% fewer injury-causing crashes. It publishes its method and local benchmarks.

An independent study from the Insurance Institute for Highway Safety found that Waymo’s driverless cars had 68% fewer police-reported crashes per mile than human drivers in the same cities. That outside result makes the lead hard to dismiss as marketing.

Baidu said in August that its driverless cars averaged about one airbag deployment every 14.4 million kilometers through June. That is a useful number, but it is not a crash table matched against local human drivers. Pony.ai and WeRide reveal even less.

Waymo’s record includes mistakes. It has recalled software on 3,871 vehicles after cars entered closed freeway work zones, and another update tightened behavior around flooded roads. Those recalls show weak spots, but they also let outsiders see the failure and the fix.

Safety question	Waymo	Leading Chinese operators
Large driverless exposure base	More than 270 million autonomous miles	Baidu reports more than 240 million fully driverless kilometers
Detailed crash rates	Publicly available	Baidu airbag rate only; no full breakdown
Human-driver comparison	Adjusted for the same operating areas	Not published in matching form
Public recall trail	Detailed through NHTSA	Much less visible outside China
What we can say	Strong evidence of lower crash rates	Too little public data for the same confidence

If you want more recent data on this point, please see our latest autonomous vehicle market report.

Did the Wuhan outage expose a serious weakness in China’s robotaxi model?

Yes. The Wuhan outage exposed a fleet-level weakness, and the timing suggests it has shown up in Baidu’s numbers.

On 1 April 2026, more than 100 Apollo Go cars stopped during a system malfunction, according to Wuhan police and the Associated Press. Bloomberg described “dozens” of stalled cars, a looser figure. Some passengers were stranded in fast traffic, though no injuries were reported.

One bad engine disables one taxi. A failure in shared software or fleet control can stop many cars at once. The exact technical cause has not been disclosed. What followed is clear: a reported pause on new permits, operating changes in some cities and a second quarter in which Baidu’s rides fell by about two-thirds. Baidu has not tied that drop to Wuhan by name.

Waymo has had fleet-wide problems too, including cars stopping during a power outage. Any city that depends on robotaxis needs backup communications, fast passenger rescue and a way to clear many stopped cars at once.

Is China’s regulation helping robotaxis or slowing them down?

Both. China’s rules let companies scale fast, but in 2026 they also sharply cut Baidu’s rides.

Cities such as Beijing, Guangzhou, Shenzhen and Wuhan created large test zones and permits that let companies remove the safety driver step by step. Local governments want the factories, data centers and jobs that come with the industry.

After the Wuhan failure, Bloomberg reported in April that authorities had paused new autonomous-driving permits. Baidu then said regulatory considerations cut its rides in some cities, and that operations began to resume in August. No official end date for the pause has been published.

China’s approach is less permissive than it looks. Officials let companies grow quickly, then pull the brake hard when public confidence is at risk. That makes Chinese volumes less predictable than Waymo’s.

Can Chinese robotaxi companies beat Waymo outside China?

For now, yes: Chinese companies run more real driverless services abroad. But most foreign launches are still small, and Waymo is now entering Europe too.

WeRide has the strongest proof. It runs fully driverless rides in Abu Dhabi and Dubai, with about 400 robotaxis in the Middle East, and holds a Swiss driverless permit for Zurich, its fifth joint deployment with Uber. In September 2026, it received a Spanish permit with Uber to test in Madrid, with commercial service targeted for the end of 2026.

Baidu launched fully driverless commercial rides in Dubai in July 2026, bookable on its own app and on Uber. It is testing in London with Uber and Lyft and in Hong Kong. Pony.ai started driverless test rides with passengers in Zagreb in September and has an Uber deal for more than 2,000 planned robotaxis in Europe.

Many of these are tests or plans, not regular passenger volume. Waymo now targets Madrid in 2027, so the Middle East is China’s clearest foreign lead and Europe is an open race.

Is the United States closed to Chinese robotaxi companies?

In practice, yes. The American market is closed to Chinese robotaxi software, which makes a single global winner unlikely.

The US Commerce Department’s connected-vehicle rule bars autonomous-driving and connectivity software linked to China or Russia from model-year 2027 vehicles, and covered hardware from model-year 2030. Chinese-controlled carmakers also face wider limits on selling connected cars there.

Waymo’s Ojai shows how narrow the permitted route is. Zeekr builds the vehicle body in China, then Waymo installs its own sensors, computers and software in Arizona. Waymo says the imported shell carries no Chinese telematics.

Chinese companies can lead at home and win contracts in the Middle East, Asia or Europe without entering American cities. Waymo can dominate the United States while staying out of mainland China. Security policy is splitting the race into regional markets.

So, is China winning the robotaxi race?

Not overall. China is winning the race to industrialize robotaxis, but Waymo is winning the race that riders feel.

China has the broader field. Baidu, Pony.ai and WeRide all run paid driverless cars, order thousands more, keep vehicle costs low and sign foreign partners. Its car factories give them a path to much larger fleets.

Waymo is the better operator, and 2026 widened the gap. It carries several times more weekly passengers than Baidu averaged in the second quarter, runs the largest disclosed single fleet and backs its safety claims with detail no Chinese company matches. Baidu’s drop showed that a Chinese lead in rides can disappear when regulators step in.

Tesla could change the American side later but is far from the lead. In September 2026, it said it had passed 1 million unsupervised robotaxi miles across six cities, a small fraction of Waymo’s 270-million-mile record.

China still faces three problems: transparent safety data, company-wide profitability and fleet reliability. Waymo has to prove it can scale abroad and bring costs closer to Chinese levels.

China is ahead as a manufacturing system and in the number of serious operators. Waymo is ahead as a service, by a wider margin than at the start of the year. Calling China the winner today goes too far; ignoring the industrial ground it holds would be a mistake.

If you want more recent data on this point, please see our latest autonomous vehicle market report.