Google Robotaxi: Inside Waymo - the robotaxi revolution leader

The rehabbed mid-century warehouse on San Francisco’s southeastern fringe carries no obvious signage. Nestled among mundane businesses (a refrigerated truck company, a door installation service, a drywall distributor), it seems wholly unremarkable—at least until you notice that the white Jaguar I-Pace electric SUVs that keep pulling into its parking area don’t have anyone in their drivers’ seats. They’re Waymo autonomous cars—often called robotaxis—and the building is their depot.

On this quiet Friday morning, dozens are on the premises, a meaningful chunk of the city’s fleet of more than 300. Many are merely charging up for their next tour of duty. Others are cleaning their own sensors, emitting faintly comical squirting and huffing sounds as they do. And some, needing tire repair or replacement parts, are here to get attention from a Waymo mechanic. When each car is ready to hit the road again, it sees itself out, deftly veering around any humans who may cross its path.

Not long ago, this scene would have been the stuff of science fiction. Most places, it still is. But in the three cities where Waymo achieved full commercial deployment in 2024—San Francisco, Phoenix, and Los Angeles—summoning a robotaxi from your smartphone is rapidly losing its futuristic tinge. Yes, it’s still novel. But it’s also workaday transportation.

It was hardly preordained that this day would come. The company’s origins date to 2009, when Google hatched an improbable self-driving car venture. It quickly became the flagship initiative at Google X, an entire arm devoted to launching audacious “moonshots.” And in 2016, the business was spun into Waymo, a stand-alone unit of Google’s parent company, Alphabet.

Waymo didn’t begin offering autonomous rides to members of the public until 2020. By the end of 2023, it had logged more than 1 million of them. In 2024 alone, its ride-hailing service, branded as Waymo One, completed more than 4 million more. With 700-plus cars on the road, it’s currently surpassing 200,000 rides a week, up from 100,000 last October and 10,000 in August 2023. What was long regarded as one of Google’s most idiosyncratic science projects finally feels real. Especially in tech-enamored San Francisco, where Waymos are suddenly taking on cable cars as symbols of the city.

As remarkable as Waymo’s technological underpinnings are, the logistics of its real-world deployment over the past couple of years have demanded just as much ingenuity. “What we’ve done in San Francisco is prove to ourselves—and to the world—that not only does autonomy work, but it works at scale in a market and can be a viable commercial product,” says Dmitri Dolgov, who was named Waymo’s co-CEO in 2021, sharing the title and oversight of the company’s strategy with Tekedra Mawakana. A member of Google’s original self-driving car engineering team, Dolgov now oversees development and deployment of Waymo’s technology; Mawakana, formerly Waymo’s COO, joined the company in 2017 as head of policy and is responsible for its commercialization.

Now, Dolgov says, “The question is, How quickly can we scale it to many more markets?” Anything resembling the ubiquity of Uber—which is available in more than 10,000 cities—won’t be in the offing anytime soon, if ever. But Waymo brought service to Austin earlier this month as well as a 27-mile chunk of Silicon Valley. It plans to launch in Atlanta later this year, with Miami following in 2026. The company says it’s sending small squads of cars to Las Vegas, San Diego, and eight other cities for early tests with human drivers behind the wheel, and it has announced its first foray outside the U.S.—a Tokyo “road trip” in which it will start collecting mapping data on the city’s streets, also with a driver on board.

The investment required to reach this point has been extraordinary. After receiving its first outside funding in 2020 and raising another round in 2021, Waymo announced its largest infusion of cash last October: $5.6 billion from Alphabet, Andreessen Horowitz, Fidelity, Tiger Capital, and other firms. The new round boosted its valuation past $45 billion—far short of the potential $175 billion floated in a 2018 Morgan Stanley note to clients but still evidence that backers see the possibility of profit after years of spending.

Not that profit is the only overarching motive. “Every year, 1.4 million lives are lost to traffic crashes,” read a sobering display in Waymo’s booth at January’s CES, the annual techfest in Las Vegas. “The status quo is not acceptable.” By wresting driving out of the hands of humans who might be distracted, intoxicated, or otherwise impaired (or, just, you know, human), the Waymo Driver—the company’s name for its autonomous platform—could someday put a dent in that terrible reality.

“Building a culture of safety, a culture of trust, that’s been a big part of what we’ve been focusing on for the last two years,” says Mawakana. Making cars drive themselves, as amazing as it is, isn’t as important as successfully commercializing it at scale. What comes next could define Waymo’s legacy.

Scientists have long been fascinated by the challenge of teaching inanimate objects to navigate the world on their own. Back in the 1960s and 1970s, researchers at Stanford University hacked together multiple versions of a rolling robot they called the Stanford Cart. Eventually, it could painstakingly steer its way through a large room cluttered with chairs.

In the mid-1990s, autonomous technology captured the imagination of a Stanford graduate student named Larry Page. He didn’t pursue it himself—along with classmate Sergey Brin, he was too busy figuring out how to make search engine results more relevant. But Page was still paying attention in 2005, when the U.S. Department of Defense’s DARPA innovation lab sponsored a 132-mile self-driving race across the Mojave Desert. A modified Volkswagen Touareg nicknamed “Stanley” completed the route in 6 hours and 54 seconds, netting the $2 million prize for the Stanford team.

By then, Page and Brin had turned their own Stanford research into Google, the world’s most popular and profitable search company. Hungry for additional brainpower to throw at even bolder undertakings, Page hired Sebastian Thrun, one of Stanley’s masterminds, in 2007. Thrun’s first project was Street View, Google Maps’ panoramic photo feature. Then Page charged him with a new task: Build autonomous cars that could take on cities.

That was a radically more difficult goal than racing them in the desert, where they didn’t need to account for other vehicles, pedestrians, stop signs, and additional urban realities. But in a 2013 interview, Page told me that, if anything, the project had been overdue. “We said, ‘We can actually drive on public roads,’” he remembered. “‘We’re going to do it safely. We’re going to test it. We’re going to prove these things are possible.’ That all could have been done 10 years ago.”

Initially, Google’s effort was fueled by classic Silicon Valley overoptimism. At a 2012 bill-signing ceremony making self-driving vehicles legal in California, Brin said they would become a reality for “ordinary people” within five years. They didn’t. And for a while, it felt like autonomous cars might be one of those breakthroughs that would remain permanently five years in the future. (By the time Brin made his prediction, Thrun had left full-time work at Google to focus on Udacity, the online learning startup he’d founded.)

For all the wildly advanced technology involved, the basic idea for self-driving is straightforward enough. By decking out a car’s exterior with a bevy of sensors—lidar, radar, and cameras—Google could give it a 360-degree view of the surrounding traffic, safety signage and signals, and other elements, expected or otherwise. AI and mapping software, running on a powerful computer in the trunk, would let it steer through a route while complying with the rules of the road. Over the years, Google, and then Waymo, autonomized production-model cars from Toyota, Audi, Lexus, and Chrysler. In 2014, Google even showed off its own cute-as-a-button self-driving microcar, which looked like it had escaped from a Disneyland ride. (One is still parked in the lobby of Waymo headquarters.)

But as Waymo plugged away at building out its self-driving platform, it discovered that every step forward tended to reveal additional work ahead—helping to explain why it took the company until 2024 to attain some semblance of Brin’s prediction for 2017. “You think you’ve solved a certain layer of the problem, but then you find that there’s another layer past it,” says Chris Ludwick, who joined Google’s self-driving car project in 2013 as its second product manager and is now Waymo’s director of product management for ride-hailing commercialization.

If it was any consolation, the many other teams that had sprung up to build self-driving cars weren’t having an easier time of it. Uber and Lyft, both of which had tried to jump-start their own autonomous technology, ended up abandoning their efforts and selling the remains to self-driving trucking company Aurora and Toyota, respectively. Meanwhile, Elon Musk’s 2019 promise that a software update to Tesla’s so-called Full Self-Driving feature for existing cars would create a million robotaxis within a year still hasn’t amounted to anything, though he now says that the company will launch some sort of FSD robotaxi service in Austin in June.

“First we watched companies combine, and then we watched companies go away,” says Mawakana. “We knew that this takes a lot of resources and also a lot of fortitude to see it through.”

Indomitability, not speed, turned out to be the secret weapon that got Waymo to the point where it could go beyond testing autonomous cars and begin the work of making them commercially available. “They have been like a tortoise moving toward this moment,” says Alex Roy, a venture capitalist, cross-country car racer, and alumnus of Argo AI, a self-driving startup backed by Ford and Volkswagen that was valued at a reported $12.4 billion in 2021—and folded the following year.

The prescience and patience that got Waymo this far haven’t just helped it pull ahead of the rest of the autonomous field. In an era when many forms of artificial intelligence are still struggling to live up to the hype, the company has successfully applied the technology to one of everyday life’s most pervasive activities in a way that remains rare. As Dolgov puts it, “The Waymo Driver is the most mature manifestation of AI in the physical world today.”

Waymo’s technology may be exceptional, but in a world that’s already so familiar with ride-hailing services, the company needed to create an experience that felt ordinary, only nicer. Summoning a car in the Waymo app works similarly to calling an Uber or Lyft, and the price for a ride, which is set dynamically based on demand, is often comparable once you factor in the lack of tipping. Ask fans what they like about the service, and they will likely cite its comfort, consistency, and privacy. Maybe there’s theoretical bonhomie to be had from chatting with random Uber drivers and listening to their favorite radio stations in whatever vehicles they happen to own. But the cushy isolation of a Waymo autonomous Jaguar “is like a little office on wheels that I can take a call from,” says Factory CEO Matan Grinberg, who uses the cars to commute to and from the San Francisco AI coding startup he cofounded. (In case you’re wondering: Waymos have interior cameras and microphones, but Google says the mics only switch on if you’re talking to a remote customer-service rep.)

There are some notable limitations. First, there’s Waymo Driver’s inability to take highways, except in test runs. (Waymo says it expects to offer highway driving later this year.) It’s also programmed to drive carefully and deferentially, which can lead to slower travel times than in an Uber or a Lyft. Rather than attempting to idle on a congested thoroughfare, it may also make you walk a bit out of your way for pickup and drop-off. With a fixed quantity of cars on the road, pickup times can be on the slow side in some instances, though enthusiasts say the experience is worth it: “I plan ahead—I’ll wait extra eight minutes to get in a Waymo,” says venture capitalist Villi Iltchev of Category Ventures.

Despite their caution, Waymos still make dumb mistakes, some of which are fascinating case studies in AI’s limitations: Last December, one blithely rolled into a patch of wet cement outside a San Francisco hospital. (The Washington Post’s Lisa Bonos reported that Waymos received 589 San Francisco parking tickets in 2024.), Overall, though, customers seem delighted. The Waymo One app has a five-star rating in the iPhone App Store and 4.9 on Google Play.

Of course, most people haven’t even seen an autonomous vehicle in person yet. Studies have long shown that the whole idea remains jarring to the uninitiated. In 2024, for example, an American Automobile Association survey found that 66% of respondents feared self-driving cars, while only 9% trusted them. Getting more folks into Waymos might be the best way to conquer that trepidation: “It’s night and day between people hearing about it and experiencing it,” says Dolgov.

Hard evidence that a world teeming with autonomous vehicles will be safer than the current state of affairs could also help. Waymo’s website is dense with information on the subject, including downloadable files of raw, incident-by-incident data from the miles its cars have traveled so far. It’s still early to come to sweeping conclusions, but overall, the picture is promising: Over 25.3 million miles, Waymo cars were involved in only two bodily injury claims and nine property damage claims. According to stats from Swiss Re, a major reinsurer, human drivers would have accounted for 26 bodily injury claims and 78 property damage claims across the same distance.

Others in the autonomous-driving industry have had a checkered past when it comes to taking safety seriously. GM’s Cruise service, for instance, was once Waymo One’s most direct competitor. But in October 2023, one of its cars struck a San Francisco pedestrian who’d already been knocked down by another vehicle. Rather than noticing anything was amiss, it dragged her 20 feet. She survived, but Cruise didn’t: GM suspended its operations, paid more than $2 million in fines to federal and state regulators for failing to cooperate with their investigations of the incident, and ultimately exited the robotaxi business last December.

Waymo, by contrast, is “making the case for [its] safety in a more principled and credible way than I think a lot of others have done,” says Edward Niedermeyer, the cohost of Autonocast, a podcast on the future of transportation. “And that’s one of the really important lessons of this: [Autonomous vehicles] are not just a technological challenge.”

Still, not everyone buys the proposition that self-driving cars will be a boon to the cities that get them. In the end, they remain cars, a mode of transportation that will always be more traffic-clogging than any form of mass transit. They may not even be a singular solution to unsafe streets.

“We could be installing speed-limiting technology on our [existing] cars,” says David Zipper, a senior fellow at the MIT Mobility Initiative (and Fast Company contributor). “We could be creating dedicated bike lanes. We could be doing daylighting at intersections to improve visibility. There’s a long, long list of things we could be doing, and there’s really no reason to jump straight ahead to robotaxis as the only way to reduce the deaths on our roads.”

Nonetheless, many places seem poised to make that jump. “My hypothesis is that we’re going to see more and more cities that really want automated solutions,” says William Riggs, the director of the University of San Francisco’s Autonomous Vehicles and the City Initiative. “They see them as synergistic with existing transit operations and potentially filling in gaps, and [believe] that the benefits will outweigh the costs of this technology.” Data from Waymo’s six existing and announced markets could do as much as anything to prove out those theories—or show that they were misplaced.

In Phoenix, where Waymo has been driving since 2016, Mayor Kate Gallego says some of the upsides have been unexpected. HIV/AIDS patients are more likely to show up for appointments: “People like taking the Waymo to the clinic, because there’s no judgment from the driver.” She also notes that the cars’ restrained driving has a calming effect on other vehicles: “If there’s two lanes of traffic and two Waymos going the speed limit, everyone goes the speed limit.”

Listening to cities’ needs has been critical to Waymo’s recent ramp-up. It’s already trained more than 15,000 first responders on how to best interact with its robotaxis, which are programmed to pull over safely if they detect a police vehicle flashing its lights behind them and then roll down the windows and establish a call to a remote representative. But in San Francisco, when the company presented the rules it had formulated to help cars safely drive through emergency situations such as accident sites, “Law enforcement gave us the feedback of, ‘That’s too complicated,’ ” says Lety Cavalcante, Waymo’s director of operations and head of operations center. That led to a revised strategy for dealing with danger zones: Whenever possible, avoid them altogether.

For Waymo, this kind of engagement goes beyond working with government agencies. “From the very beginning, they have brought our community to the table,” says Sharon Giovinazzo, the CEO of San Francisco-based LightHouse for the Blind and Visually Impaired. According to Giovinazzo, ride-hailing vehicles driven by humans have a poor record for reliably accommodating passengers with guide dogs— regardless of the law—making Waymo’s driver-free service a major advance for blind riders. The organization she leads has also consulted with the company on features such as an accessibility mode with spoken cues about the route a car is taking.

For all of self-driving’s potential to rewire the world, there’s another, largely invisible story here. Now that Waymo has demonstrated that a critical mass of people will pay to be driven around by an AI-infused vehicle, it needs to tackle the unglamorous details of transitioning from money pit to moneymaker. Alphabet isn’t done shoveling money into the effort: On an earnings call last July, CFO Ruth Porat said it had committed another $5 billion. But Waymo’s future viability is contingent on it offering rides at a price that leaves room for profit, and that requires getting serious about spending.

That goal is reflected in the sixth-generation version of its Waymo Driver hardware and software suite, which it announced last August and will roll out in minivans manufactured by China’s Zeekr (a corporate cousin of Volvo) that are currently undergoing testing. Google says the update includes improvements for scenarios such as winter driving. Yet it also uses fewer sensors than the fifth-generation system: “It really helps us drive down costs, and all of it is without compromising safety,” says Mawakana. (Chinese automakers such as Zeekr produce vehicles at far lower costs than their U.S. rivals, presumably another element in Waymo’s calculus; how President Trump’s new tariffs might impact the math of the deal remains to be seen.)

There’s also plenty of opportunity to make Waymos more truly self-sufficient, which might both lower costs and improve rider experience. The company says it’s reducing reliance on the remote technicians who sometimes step in to help cars out of sticky situations: “Our team is working on getting you moving” read the screen in my Waymo during one recent trip, when the car was flummoxed by the challenge of turning onto a busy street. It’s also automating more processes at its depots, where the cars can now handle routine jobs such as seeking out an unoccupied charging station on their own. “We—or I, at least—didn’t fully appreciate how important it is to do that kind of stuff well in order to run the service well,” says Ludwick.

Even at this early stage in Waymo’s expansion, it’s increasingly turning to partners to help shoulder the responsibility of scaling up. In Austin and Atlanta, Uber will incorporate Waymo-hailing into its app and manage the cities’ fleet operations. Nigerian mobility startup Moove will take over Phoenix’s fleet and wrangle Miami’s cars from the start. The Tokyo mapping expedition is a collaboration with two locals: taxi company Nihon Kotsu and ride-hailing service Go.

If Waymo can grind down expenses over time, the benefits might redound beyond its own bottom line. By 2035, a McKinsey study estimates, the average cost of operating a robotaxi could tumble from $8.18 (assuming a 1,000-car fleet in a city) to $1.35 per mile. While McKinsey sees that involving everything from lower manufacturing costs at higher volumes to reduced ongoing R&D investment over time, it expects the biggest reductions would come from applying more automation to operational tasks such as charging and cleaning vehicles, as well as leveraging partnerships to scale up operations more efficiently—areas where Waymo is already making progress.

Passing those savings on to customers in the form of cheaper rides might tempt people into giving up their cars more quickly than they have in the era of human-driven ride-hailing services. In aggregate, the time they redirect from driving to activities that are more productive and pleasurable could be meaningful to consumers: “There might be a real economic value proposition to this,” says Philipp Kampshoff, global co-lead of McKinsey’s automobile practice.

Kampshoff isn’t talking about Waymo in particular, and the company could eventually face robust competition from multiple players. Amazon-owned Zoox is testing its robotaxi—which ditches the driver’s seat in favor of a comfier passenger experience—in San Francisco and on the Las Vegas Strip, though The Washington Post recently reported that the National Highway Traffic Safety Administration concluded last year that Zoox’s vehicles’ lack of items such as a brake pedal violated current safety standards. Lyft says it will offer robotaxis in Dallas as soon as 2026. Others in the game include London-based Wayve and Motional, in which Hyundai has a controlling stake. (The Korean carmaker is also a partner of Waymo, which plans to begin testing self-driving Ioniq 5 SUVs this year.) In China, Baidu, Didi, and WeRide have robotaxis on the road and EV-maker BYD is promising to roll out self-driving technology to its cars starting this year, though geopolitics may keep these companies from entering the U.S.

Then there’s the world’s most-watched carmaker: Tesla. Last October, Elon Musk presided over an extravaganza on the Warner Bros. lot where he unveiled the Cybercab, Tesla’s first purely autonomous vehicle. Resembling a drastically shrunk-down Cybertruck, the $30,000 two-seater will recharge by parking on a giant wireless pad, eliminating the need to plug in a charging cable—a duty still handled by a human at Waymo’s depots.

The Cybercab will supposedly go into production in 2026, though Musk himself acknowledged that timeline might be overly optimistic. Even if it turns out to be the real deal, autonomous fleets need to be charged, repaired, and otherwise managed. “The operational component does not exist inside Tesla, [so] it either must be built or it must be outsourced,” says Roy. (The company has shared a video showing a Cybercab’s interior being cleansed by a bespoke robot.)

By the time other robotaxis have caught up, Waymo may be exploring new territory. There’s meal delivery, for starters. (Some Uber Eats orders in Phoenix are already ferried via Waymo; Mawakana was relieved to discover that people are willing to come outside to retrieve their food.) And long-haul trucking. (In 2023, the company paused work on Waymo Via, its autonomous truck platform—a joint venture with Daimler—but Mawakana is already talking about it resuming “when we’re ready.”) And the tantalizing possibility of Waymo technology being built into cars people can buy for themselves. (“We’re in some really interesting conversations there,” she says.)

The goal, Mawakana says, is to become “the world’s most trusted driver across a host of applications.” A year or two ago, that might have felt out of whack with autonomy’s seemingly diminished prospects. Now it’s just wildly ambitious—and if any autonomy company has earned the right to be that, it’s Waymo.

FASTCOMPANY