There Are Two Futures for Self-Driving Cars. You Won’t Like Either One.

Eric Thayer/Getty Images

Sign up for the Slatest to get the most insightful analysis, criticism, and advice out there, delivered to your inbox daily.

Earlier this year, I tried to convince my mother-in-law to ride in one of Waymo’s driverless taxis in San Francisco. It was a tough sell at first, but was helped along by a series of careening Uber rides, which prompted the question: How much worse could a robot drive? The answer, at least for us, as our Waymo carried us gently down the city’s famous hills from stop sign to stop sign, was clear. We felt safer with no one behind the wheel.

This might be the summer America asks and answers the same question. Waymo is now clocking more than 250,000 paid rides a week, on track to double last year’s total. It currently operates in San Francisco, Los Angeles, Phoenix, and Austin, and plans to expand to Atlanta, Miami, and Washington, D.C.—with Boston, Nashville, New Orleans, Dallas, Las Vegas, San Diego, Orlando, Houston, and San Antonio in the pipeline.

Meanwhile, Tesla says it will deploy “full self-driving” taxis in Austin this month, the long-awaited debut of a service that Elon Musk has been promising for a decade—and one that, to his most bullish investors, will soon represent 90 percent of Tesla’s value as a company.
The two approaches could not be more different, in ways that suggest different possible futures for what autonomous vehicles mean for society.

Waymo has been patiently plugging away for years at the problem of navigating complex urban environments with computers, outlasting rivals who have stepped back after hitting and dragging (GM’s Cruise) or killing (Uber) pedestrians. While Waymo’s sample set is limited to a handful of big cities, its safety record—more than 56 million miles of robot driving—is impressive. According to a peer-reviewed study in the journal Traffic Injury Prevention, Waymo recorded 85 percent fewer crashes likely to cause serious injuries—like broken bones, unconsciousness, or skull fractures—compared to human drivers.

That kind of progress hasn’t come cheap. Google parent company Alphabet has sunk billions into the technology, and the business may lose nearly $2 billion a year. (The financials are buried in Alphabet’s Other Bets division.) It costs as much as $100,000 to equip Waymo’s cars with their array of lidar sensors, and further expenses include mapping, maintenance, electricity, insurance, parking, and on-call staff to help pilot the car if it gets stuck. Or, presumably, set on fire—as happened this week at anti-ICE protests in Los Angeles.

Tesla is trying something different: A low-touch approach based on cameras and A.I. “The issue with Waymo’s cars,” Musk said this spring, “is they cost way-mo money.” Tesla’s “full self-driving” hardware may cost as little as $400. But its safety is in dispute, to put it mildly. On Friday, Bloomberg published a video of a Tesla in “full self-driving mode” killing a woman on an Arizona highway in 2023. The National Highway Traffic Safety Administration announced an investigation into the company last fall, but U.S. Department of Transportation Secretary Sean Duffy has seemed less than eager to press automakers on crash reporting, and the version of the GOP tax bill that passed the House would suspend states’ abilities to regulate artificial intelligence.

Certainly, Tesla’s tech is cheap. The question is whether the technology works well enough to avoid the kinds of gruesome accidents that halted its competitors. (The cars must also avoid construction sites, tractor trailers, snow banks, wet concrete, and other hazards of the urban environment.) For Waymo, it’s the opposite: The tech clearly works. The question is whether the cost can come down to make robotaxis profitable—and competitive with Ubers, public transit, and even private cars.

To fundamentally reshape urban transportation, robotaxis have to be both functional and cheap. Currently, ordering a Waymo costs as much as an Uber or more—and the company is largely competing for the same clientele (some data from San Francisco suggests Waymo is eating up Uber and Lyft trips). But without human drivers to pay, it is easy to imagine the service being substantially cheaper. If that happens, we might see a supercharged “Uber effect” like what happened in cities in the 2010s—increased mobility for people who do not or cannot drive themselves, including children, seniors, and people with disabilities. One estimate concluded that we might see 14 percent more traffic if those nondriving groups traveled at the same rates as current drivers. Every trip for which a robot replaces a human driver will come with a lower risk of crashing.

If Waymo’s higher-cost model proves to be the path for self-driving tech, we might see a divide between rich, deep markets that justify its presence—and those that don’t. Lewis Lehe, an assistant professor of transportation systems at the University of Illinois, notes on his Substack that this service will have different economics than today’s ride-hail. You can ride an Uber anywhere someone is willing to drive one, and drivers have different hours (and different wages) in low-demand places. Not so for a Waymo, which will cost the same to deploy in a rich city and a poor town—and see much less demand in the latter. “The divide will beget tropes about the ‘two Americas.’ ” Lehe writes. “Country musicians will never shut up about driving. President Vance will speak for the drivers whom global elites have left behind.”

As with the rise of Uber, this boom will decrease transit ridership and increase traffic congestion. Tens of thousands of full-time ride-hail drivers will struggle with auto debt, while cities try to wield congestion pricing against an adversary with little obligation to share data and considerable power to withhold service for leverage. Unlike Ubers, whose drivers flood the roads at rush hour and sleep at night, Waymo fleets could struggle to meet peak demand—but offer low-priced, off-hours service when there are more vehicles than riders. And while public transit agencies are required to offer paratransit services to people who use wheelchairs, the expectations for ride-hail companies are much murkier—and many wheelchair users report being stranded by Uber and Lyft.

That is one scenario you can contemplate as you watch Waymos arrive in your city. The other road will be represented by Tesla’s apparently imminent experiment in Austin this summer.
What happens if autonomous vehicles are cheap? Cathie Wood’s ARK, an investment management firm that has been one of Tesla’s most bullish investors, outlined the maximal case for AVs in a February research note. ARK predicts global robotaxi services costing customers just 25 cents a mile by 2035—a price that would be about eight times cheaper than an Uber today, and about one-third the price of AAA’s estimate of what a new car costs per mile.

That assumption is predicated on a low-cost model similar to the one that Tesla says it can deliver. It imagines the price of an autonomous vehicle dropping to $15,000, which seems far-fetched until you know that the Chinese auto giant Baidu said in November that its Apollo Go robotaxis cost less than $30,000 to produce, and rides cost one-fifth as much as a conventional cab. Baidu will start testing its service in Switzerland by the end of the year.

If robot-driven cars cost just 25 cents a mile to use, you could get a cab from Chicago O’Hare to the Loop for $4.25. Most big-city trips would cost less than a candy bar. The mobility and traffic boom described above would occur on steroids, and some travelers would find car ownership is no longer sensible or necessary. Parking lots and garages would vanish from cities as robotaxis ferried people and freight around all day long in a wave of endless traffic that occupied road space with an efficiency that error-prone humans could never handle. Streets could be closed to traffic at the push of a button. Drivers will be reduced, as John Ruskin once said of railroad passengers, from active travelers to “human parcels.” Rural mobility would explode as onerous drives were replaced by low-cost robotrips. (Note that this world of limitless, low-cost automobility relies on some very sunny assumptions about the decreasing costs of maintenance, energy, insurance, operations, and taxes.)

In most cities, however, there are two iron laws that autonomous vehicles cannot circumvent no matter how little they cost or how well they drive. The first is space: There is a limited amount of roadway available for travel, and in most cities, it is pretty much occupied with current, human-driven cars. Perhaps robotaxis can squeeze in a bit better and raise our tolerance for traffic, but their utility will be limited by urban geometry. The second is time: Everybody wants to travel at the same time of day, a need that can only be met by a fleet of cars that sits idle for most of its life.

During the rise of the traditional automobile, we did our best to overcome these challenges. We built highways and parking lots to accommodate the maximum moment of demand, at enormous expense, with marginal use during off hours. We bought cars that are vital to get us from place to place but spend 95 percent of their time parked. Those choices were expensive. A carmaker trying to make a buck probably wouldn’t want to follow in our footsteps. But without new and expensive infrastructure or a supersized fleet, a new kind of car can only produce so much new mobility in the metropolis.