Every week, it seems, Waymos pop up in another U.S. city. And as these Alphabet-backed robotaxis spread, so do the fights over them, at just about every level of government. Last week, Washington D.C. Councilmember Charles Allen introduced a bill to operationalize self-driving taxis there after the council seemed poised to stall a prospective rollout earlier this month. In New York City, when Mayor Zohran Mamdani recently let a permit allowing Waymo testing expire, Governor Kathy Hochul reversed course on the issue by withdrawing an earlier proposal that would have allowed tech companies to start picking up passengers with driverless cars outside the city limits.
In reaction to the political pushback, robotaxi enthusiasts present a straightforward moral arithmetic: that an irrational tech backlash and provincial concerns about protecting taxi and rideshare drivers’ jobs are holding back a revolution in road safety that will save tens of thousands of lives per year. To be sure, early data from Waymo looks promising. Compared to the average human driver in places where it operates, Waymo reports that its vehicles are 92 percent less likely to be involved in crashes with serious injuries or fatalities, and 92 percent less likely to be involved in crashes where pedestrians are injured. Extrapolating from these and other numbers, maximalist visions for what self-driving cars could mean for road safety promise that they could be the key to “eliminating traffic deaths as a leading cause of mortality in the United States,” as one New York Times op-ed suggested. The Argument’s Kelsey Piper dubbed them “a pretty good cure for car accidents,” before claiming (provocatively) that “autonomous vehicles will save more people than a cure for most cancers.”
This is an attractive story in a country where, last year, 36,640 people died in traffic fatalities, and where—despite recent improvements—roads remain far more dangerous than in other wealthy countries. Could Waymo be the solution? And if so, who would be stupid enough to stop it?
Experts, however, caution against overinterpreting Waymo’s one-to-one comparisons as definitive proof that its cars are currently safer than human drivers. “No one has enough data to know anything about fatalities yet,” said Philip Koopman, a longtime researcher of self-driving vehicle safety and emeritus professor of engineering at Carnegie Mellon. Waymo’s fleet of roughly 3,000 vehicles has by now traveled at least 200 million miles. “They’re going to have to go a billion miles before we know how fatalities turn out,” Koopman told me. “If they want to claim they’re safer than people, they don’t have the data yet.” Even Waymo has walked back claims that its technology is “already reducing traffic injuries and fatalities,” acknowledging that there simply isn’t enough data to make that case.
So far, Koopman added, most of Waymo’s data comes from places with relatively good driving conditions, like Phoenix, the Bay Area, and Austin. Taxis in these locations largely haven’t had to grapple with snowstorms and hurricanes, and when they have run into inclement weather, they’ve had trouble. Waymo temporarily suspended operations in San Antonio this week when one of its cars was swept onto a flooded road. Large-scale electrical or network outages present challenges for systems that rely on connectivity. So do rural areas with bad cell phone and internet service. These may well be easily surmountable challenges; it’s just too early to tell how large numbers of Waymos will handle problems that small numbers of Waymos have yet to encounter.
The idea that Waymos and other robotaxis are on the verge of preventing mass death, moreover, is premised on an assumption that human driving would be eliminated, if not completely, then at least on a gargantuan scale. This is a plausibly defensible and extraordinarily radical vision. Yet convincing even a modest portion of the 92 percent of U.S. households with at least one car to give up their wheels will be much harder than convincing a handful of city governments to let Waymos operate. Waymo, Tesla, and Zoox currently have fewer than 4,000 robotaxis on the road nationwide; there are 285 million private and commercial vehicles registered in the U.S.
As E.V. and public transit enthusiasts here have learned, changing U.S. drivers’ behavior in even minor ways is extremely difficult. The kind of behavioral revolution required to keep millions of people from driving would likely entail muscular government policies and open conflict with—at a bare minimum—the U.S. auto sector. Waymo seems no more likely to achieve such large-scale transformations than Uber, which never fulfilled its heady promises to “make car ownership a thing of the past.” While robotaxi companies are certainly eager to expand, they’re a long way from bringing about a mostly driverless future: S&P Global Mobility’s 2025 Autonomy Forecast projects autonomous vehicles to account for just 0.5 percent of the U.S. light vehicle market by 2035, or around 85,000 units.
China’s robotaxi companies, by contrast, seem poised for more rapid growth. That’s thanks in part to a much more favorable environment for self-driving vehicles, including long-running industrial policy aimed at growing that country’s auto and tech sectors. Supportive regulations have helped Chinese robotaxi companies take off, and start to expand their businesses abroad. City-level governments in Beijing and Shanghai, for instance, have offered pilot zones, expanded digital infrastructure, and developed licensing systems that allowed Baidu and WeRide to quickly transition to commercial deployment. Just two weeks after launching its seventh-generation vehicle in Gaunghzho, Pony.ai was able to achieve citywide profitability on its rides there.
Product development cycles for new vehicles in China are 20 to 30 percent shorter than those of their Western counterparts, said Shreyas Sirsi, a partner in the automotive and industrial practice at the consultancy AlixPartners. This relative speed allows companies to iterate new software and software-designed features quickly within a radically competitive sector. In its recent survey of 1,000-plus senior executives at auto suppliers, tech companies, and automakers oriented toward self-driving vehicles, or SDVs, AlixPartners found that 36 percent of Chinese automakers devote more than half of their R&D budgets to SDVs, as compared to just 21 percent in the U.S., Europe, and the U.K.
Even if autonomous vehicles aren’t likely to replace human drivers, their proliferation in the U.S. and abroad could entail massive implications for the American companies that have struggled to develop workable business models. AlixPartners warns that Western firms face a “potentially existential” challenge as they lose out on driving across the board and bringing down costs. Ford and GM both ditched their self-driving ventures after hemorrhaging cash on them. Uber, which gave up on trying to develop its own custom-made robotaxis in 2020, is now aggressively pursuing partnerships with companies like Nvidia, Stellantis, U.S. E.V.-maker Lucid, and the Chinese robotaxi leader Baidu.
Economies of scale on higher-tech cars in China have dramatically brought down costs for LiDAR sensors, reducing per-unit costs to as little as $200 per sensor from upward of $50,000 just a decade ago. “Now you have multiple LiDAR being used on a robotaxi,” said Lei Xing, an independent analyst focused on the Chinese auto sector. “They’re only a 10th of the cost.” Chinese automakers may start equipping E.V.s that retail for less than $9,000 total with LiDAR-based intelligent driving systems. The cost of just the hardware suite on Waymo’s new sixth-generation vehicles—including cameras and LiDAR sensors—is estimated at $20,000. They’ll be fitted onto powder-blue vans from Zeekr, a Chinese firm.
Still, despite these advances, autonomous vehicles aren’t expected to dominate China’s roads anytime soon, either: S&P predicts they’ll account for just 1.3 percent of new light vehicle sales in a decade, or 825,000 units. As of today, Xing estimates that the “Big 3” Chinese robotaxi firms (Baidu, WeRide, and Pony.ai) have a combined 3,000 vehicles in cities around the country. “Compared to annual sales of 25 million vehicles,” Xing said, “those are a drop in the ocean.”
Where they do operate in the U.S., Waymos haven’t made a meaningful dent in either road safety or personal car ownership. StreetsblogNYC reported this week that, in San Francisco—where Waymos have been since June 2024—traffic injuries increased by 2.6 percent, from 2,896 in 2023 to 2,907 in 2025. There was one more crash death there in 2023 than in 2025, when 27 people died. Car ownership in the city has remained relatively steady since the turn of the century, at around 70 percent. In Phoenix, where Waymo has operated since 2020, pedestrian deaths have increased by 36.3 percent between 2019 and 2023.
Those numbers aren’t Waymo’s fault, of course, but they should signal that solving America’s epidemic of traffic deaths and injuries involves more systemic solutions. Installing roundabouts, adding more street lights, increasing enforcement of speed limits, and reducing the amount of space between crosswalks are relatively inexpensive, proven ways to make roads safer for drivers and pedestrians alike. “There’s no magic bullet,” said Ann Carlson, the Shirley Shapiro professor of environmental law at the UCLA School of Law, who served as acting administrator of the National Highway Traffic Safety Administration, or NHTSA, during the Biden administration. “Increasingly, Waymos’ safety data seem impressive,” she told me. “But we’re talking about a scale that is tiny compared to what you would actually need if you were going to drive down traffic fatalities and traffic crashes. I take the maximalist safety claims with a big grain of salt.”
Next-generation vehicle features, such as automatic emergency braking and ignition interlock devices—breathalyzers that prevent cars from starting if the driver’s blood alcohol content exceeds the legal limit—hold a lot of promise but are exceedingly difficult for NHTSA to mandate. Even if automakers are eventually required to make those features standard, the rules would apply only to new cars that are increasingly out of reach for poor and middle-class buyers. With the total cost of new car purchases now hovering around $50,000, many more people are either buying used or opting to keep older vehicles. “People hang onto their cars a lot longer than they used to,” Carlson said, so “it’d probably take 25 years to fully penetrate the fleet.”
That might be a best-case scenario. Comparatively ancient vehicle safety “technologies” still aren’t mandatory. NHTSA finds that lap and shoulder seat belts reduce passenger car occupant deaths by 45 percent, and 60 percent for light trucks—a legal classification that includes pickups and SUVs. Despite decades of clear evidence showing the effectiveness of seat belts and the dangers of not using them, there is still no national law mandating seat belt usage.
None of the above amounts to a case against Waymo, robotaxis, or autonomous vehicles. There’s no reason why more traditional solutions for improving America’s abysmal road safety statistics—including public transit investments and less car-centric planning—can’t coexist with driverless vehicles. But as governments consider legitimate questions about safety, liability, and labor dislocation, it’s worth being honest about the stakes.
