Waymo

Based on Wikipedia: Waymo

The Pizza Delivery That Launched a Revolution

In February 2008, a television producer called Anthony Levandowski with an unusual request. The Discovery Channel was filming a show called Prototype This! and wanted to borrow Levandowski's autonomous motorcycle—a self-driving two-wheeled machine he'd built for a Pentagon competition and later donated to the Smithsonian.

The motorcycle wasn't available. But Levandowski had a counterproposal.

Within weeks, he'd retrofitted a Toyota Prius with laser sensors, cameras, and the artificial intelligence code developed by Stanford's robotics team. The result was a car that could drive itself across the San Francisco Bay Bridge to deliver pizza, with a police escort trailing behind for the cameras. The episode aired in December 2008 to modest fanfare.

What viewers didn't know was that they were watching the seed of something that would fundamentally reshape transportation. That pizza-delivering Prius became the foundation for Google's self-driving car program, which launched the following month. Today, that program—now called Waymo—operates over 450,000 robotaxi rides every single week across American cities, with plans to expand to a million weekly rides by 2027.

From Pentagon Competitions to Silicon Valley Labs

The story actually begins earlier, in the deserts of the American Southwest. In 2004 and 2005, the Defense Advanced Research Projects Agency—the Pentagon's technology incubator, usually called DARPA—held competitions challenging teams to build vehicles that could navigate rugged terrain without human drivers. The military wanted autonomous trucks that could resupply troops without risking soldiers' lives to roadside bombs.

Most entries failed spectacularly. Vehicles crashed into rocks, drove in circles, or simply stopped moving. But a team from Stanford University, led by a German-born computer scientist named Sebastian Thrun, built a modified Volkswagen Touareg they called Stanley. In 2005, Stanley completed the 132-mile desert course in just under seven hours—the first vehicle ever to finish the DARPA Grand Challenge.

Thrun's team included fifteen engineers who would later form the nucleus of what became Waymo: Dmitri Dolgov, Mike Montemerlo, Hendrik Dahlkamp, and others. They refined their approach for the 2007 competition with a vehicle named Junior, which navigated through mock urban environments with traffic and intersections.

Google had been watching closely. The company's co-founder Sergey Brin had a particular fascination with ambitious, boundary-pushing technology—the kind of moonshot projects that might seem impossible until suddenly they weren't. In 2007, Google quietly acquired the entire team working on a Stanford mapping project called VuTool, ostensibly to improve Google Street View. But the real prize was the expertise in making machines understand and navigate the physical world.

The Secret Project Revealed

On January 17, 2009, Google officially launched its self-driving car project within Google X, the company's secretive research laboratory. Thrun led the effort, working alongside Levandowski, whose pizza-delivery prototype had proven the concept could work on real roads.

For nearly two years, Google kept the project hidden from public view. Seven vehicles—modified Toyota Priuses bristling with sensors—logged hundreds of thousands of miles on California roads. The cars had human safety drivers ready to take control, but the machines did the driving. They navigated highway traffic, city streets, and everything in between.

Then, on October 9, 2010, the New York Times broke the story. Google confirmed the project later that same day, and the world learned that self-driving cars had moved from science fiction to science fact.

The announcement triggered immediate questions about regulation. How do you license a car that drives itself? Who is responsible when something goes wrong? Nevada became the first state to create a legal framework, passing legislation that took effect in March 2012. A month later, Nevada's Department of Motor Vehicles issued the first-ever U.S. license for a self-driving vehicle—to a modified Prius with Chris Urmson behind the wheel (or rather, behind where the wheel would have been if he were actually driving) and Levandowski in the passenger seat.

What Self-Driving Cars Actually See

To understand how these vehicles work, imagine trying to drive while looking through a kaleidoscope that translates everything into mathematics. That's essentially what a self-driving car does, using multiple overlapping systems to perceive the world.

The most distinctive technology is lidar—light detection and ranging. A lidar sensor works like radar, but instead of radio waves, it shoots out laser beams and measures how long they take to bounce back. By firing millions of laser pulses per second in all directions, the system builds a detailed three-dimensional map of everything around the car: other vehicles, pedestrians, cyclists, traffic signs, lane markings, even potholes.

Early self-driving cars relied on lidar systems from a company called Velodyne, which cost around $75,000 each. That single component cost more than most cars. By 2017, Waymo had developed its own lidar systems that cost roughly 90 percent less—a crucial step toward making the technology economically viable.

But lidar alone isn't enough. The cars also use traditional cameras to read traffic lights, recognize brake lights, and interpret road signs. Radar sensors detect the speed and distance of nearby vehicles, even in rain or fog when cameras struggle. GPS provides general positioning, while detailed maps—updated constantly with data from the fleet—tell the car exactly where lanes are, where stop signs should be, and what the speed limit is at every point along every route.

All this sensor data feeds into artificial intelligence systems that must make predictions and decisions in real time. When a pedestrian steps toward a crosswalk, will they cross or stop? When a car ahead signals a lane change, will they actually move over? The AI draws on patterns learned from billions of miles of real and simulated driving to make these judgments dozens of times per second.

The First Truly Driverless Ride

By 2015, the technology had matured enough for a landmark demonstration. Google invited Steve Mahan, the former CEO of the Santa Clara Valley Blind Center, to take a ride through Austin, Texas. Mahan was legally blind—exactly the kind of person who stood to benefit most from cars that didn't require human drivers.

What made this ride different from every previous demonstration was what wasn't there. No steering wheel. No gas pedal. No brake pedal. No safety driver ready to take over. No police escort clearing the road ahead.

For the first time in history, a car drove itself on public roads with only a passenger aboard. The vehicle navigated traffic, stopped for pedestrians, and delivered Mahan safely to his destination. It was a quiet revolution—no drama, no near-misses, just a machine doing what it was designed to do.

Google had spent $1.1 billion on the project between 2009 and 2015. For context, when General Motors acquired self-driving startup Cruise Automation a few months later, it paid $500 million. Uber spent $680 million to buy Otto, another autonomous vehicle company. Google's investment dwarfed these competitors, and it showed in the results.

Becoming Waymo

In December 2016, Google's self-driving car project got a new name and a new corporate structure. Waymo—a portmanteau derived from "a new way forward in mobility"—became a separate subsidiary of Alphabet, Google's parent company. The spin-off signaled that this was no longer a research project but a business intended to stand on its own.

The newly independent company moved aggressively to build a fleet. It partnered with Stellantis (the company formed from the merger of Fiat Chrysler and Peugeot) to deploy hundreds of Chrysler Pacifica minivans. In 2018, Jaguar Land Rover announced that Waymo had ordered up to 20,000 of its I-Pace electric SUVs—a deal worth over a billion dollars. Alphabet followed up with plans to add 62,000 more Pacifica minivans.

The numbers were staggering, but they reflected Waymo's ambition. Self-driving technology isn't just about individual cars; it's about fleets large enough to provide reliable service across entire metropolitan areas. You need enough vehicles that customers don't wait too long, enough coverage that people can actually get where they want to go, and enough redundancy that the system keeps running even when individual cars need maintenance.

Waymo One: Hailing a Robot

In October 2020, Waymo crossed another threshold. In Phoenix, Arizona, the company began offering rides to the general public in vehicles with no human safety driver present. You could pull up the Waymo One app, request a ride, and a car would arrive and take you to your destination—all without any human ever touching the controls.

Phoenix wasn't chosen randomly. The city offered several advantages: relatively simple road layouts, predictable weather (rain confuses sensors), and regulatory officials willing to permit the service. Waymo had been testing there since 2017, first with safety drivers, then without, gradually expanding its operating area as it gained confidence.

The Phoenix suburbs of Chandler, Tempe, and Mesa became proving grounds for what a self-driving future might look like. By 2023, the service area had expanded by 45 square miles to include downtown Mesa, uptown Phoenix, and South Mountain Village. In September 2025, Waymo integrated with Chandler's public microtransit service, essentially becoming part of the city's public transportation network.

Spreading Across America

San Francisco presented a different challenge. Narrow streets, aggressive drivers, steep hills, frequent fog, and a population deeply skeptical of tech-company promises created a more hostile environment than Phoenix's sprawling suburbs. But Waymo launched there anyway, beginning consumer testing in August 2021 and gradually expanding.

Los Angeles followed in 2023, with service areas spreading throughout the massive metropolitan region. By November 2025, Waymo's California permit allowed autonomous driving across an enormous territory: all nine counties of the San Francisco Bay Area plus Sacramento and Yolo counties in the north, and in Southern California from Thousand Oaks to the Mexican border, stretching east to Interstate 15.

Atlanta and Austin joined the network, each presenting their own mix of traffic patterns, weather conditions, and road designs. As of December 2025, Waymo operates about 2,500 robotaxis across these cities, collectively providing 450,000 paid rides per week. The company aims to reach one million weekly rides by the end of 2026.

In November 2025, Waymo began allowing some rides on freeways—a significant expansion of capability. Highway driving is actually simpler than city streets in some ways (no pedestrians, no cross traffic) but harder in others (higher speeds mean less time to react, and a failure could be catastrophic).

The Global Ambition

Waymo's expansion plans extend far beyond American borders. The company has announced intentions to launch in Tokyo and London, which would make them the first foreign markets for the service. But the regulatory and lobbying groundwork suggests even broader ambitions.

In Australia, Waymo has contacted transport officials in New South Wales and hired lobbyists in Victoria. In Canada, the company has lobbied both the provincial government of Ontario and the city of Toronto, while also engaging lobbyists in British Columbia.

Within the United States, Waymo is positioning itself for rapid expansion once regulatory barriers fall. The company has hired lobbyists in Alabama, Delaware, New Mexico, Utah, and Virginia. It has registered with North Carolina's Secretary of State and lobbied Portland, Oregon. In Illinois, where robotaxis remain illegal as of late 2025, Waymo has run advertising campaigns encouraging residents to pressure legislators for legalization.

This strategy—building relationships with regulators and politicians before launching service—reflects lessons learned from earlier tech-industry mistakes. Uber and Lyft often launched first and fought regulations later, creating antagonistic relationships with local governments. Waymo appears to be taking a more methodical approach, though critics might call it slower.

The Road Not Taken: Trucking

For several years, Waymo pursued autonomous trucking alongside its robotaxi business. The logic seemed compelling: long-haul trucking involves predictable highway driving, faces a chronic driver shortage, and moves valuable cargo that could justify expensive technology.

In 2020, Waymo launched a division called Waymo Via dedicated to commercial trucking. The company tested Class 8 tractor-trailers—the eighteen-wheelers that move most of America's freight—across routes in Texas, New Mexico, Arizona, and California. It partnered with Daimler to integrate autonomous technology into Freightliner Cascadia trucks, tested with United Parcel Service for commercial deliveries, and worked with Stellantis on autonomous Ram ProMaster delivery vans.

Then, in July 2023, Waymo abruptly shut down the trucking program.

The company offered little public explanation beyond wanting to focus on ride-hailing. But the decision likely reflected both competitive pressures and strategic calculations. Other companies like Aurora Innovation and TuSimple had dedicated themselves entirely to trucking, while Waymo's attention was split. Meanwhile, the robotaxi business was growing faster than expected and demanding more resources. Sometimes focus means choosing what not to do.

Most Waymo Via employees transferred to other roles within the company rather than being laid off—a small mercy in an industry that often treats workers as disposable.

The Money Behind the Machines

Building self-driving technology requires almost unfathomable amounts of capital. The sensors, computers, and specialized hardware needed for each vehicle cost tens of thousands of dollars. The engineering teams command Silicon Valley salaries. The testing and validation required before deploying on public roads runs into billions.

Waymo has raised over $11 billion from investors, most of it from Alphabet itself. The company secured $2.25 billion in March 2020, another $750 million two months later, then $2.5 billion in June 2021. In October 2024, Waymo closed a $5.6 billion round, its largest ever.

These numbers reflect both the promise and the challenge of autonomous vehicles. The potential market is enormous—Americans spend trillions of dollars on transportation each year, and a company that could capture even a fraction of that would be immensely valuable. But getting there requires sustained investment for years before profits appear.

Waymo operates at a loss. The company doesn't disclose detailed financials, but 450,000 rides per week, even at premium prices, doesn't cover the cost of running 2,500 vehicles plus thousands of engineers plus massive computing infrastructure plus the ongoing research needed to improve the technology. The bet is that costs will fall and volumes will rise until the math works.

The Competition Stumbles

By late 2024, the New York Times described Waymo as being "far ahead of the competition." The assessment reflected not just Waymo's progress but its rivals' setbacks.

Cruise, the self-driving subsidiary of General Motors, had been Waymo's closest competitor in deploying robotaxis. But in October 2023, one of Cruise's vehicles struck a pedestrian in San Francisco. The pedestrian had already been hit by another car and thrown into Cruise's path, but what happened next made headlines: the robotaxi dragged the injured woman twenty feet before stopping. Cruise suspended all operations and faced investigations from California regulators.

The incident crystallized fears about autonomous vehicle safety. Even if the Cruise car wasn't primarily at fault for the initial collision, its behavior afterward—continuing to move with a person trapped underneath—suggested gaps in how the system handled unusual situations.

Other competitors have faced their own challenges. Tesla has promised "full self-driving" capability for years but continues to require human supervision and has faced scrutiny over crashes. Apple spent billions developing autonomous vehicle technology before reportedly abandoning the project in 2024. Uber sold its self-driving division in 2020 after a fatal crash in Arizona.

Waymo's safety record isn't perfect—no technology operating in the chaotic real world could be—but it has avoided the catastrophic failures that derailed competitors. The company publishes safety reports and has submitted data to regulators showing its vehicles are involved in fewer serious incidents than human drivers. Whether these comparisons are fair remains debated, but they've helped Waymo maintain regulatory permission to operate.

The Technology Arms Race

Self-driving cars depend on specialized computing hardware that barely existed when the project began. In 2009, the cars used whatever processors were available—mostly graphics processing units, or GPUs, originally designed for video games but useful for the parallel computations that machine learning requires.

Google invested heavily in developing custom chips called Tensor Processing Units, or TPUs, specifically optimized for AI workloads. These chips can perform the mathematical operations underlying machine learning far more efficiently than general-purpose processors. Though TPUs were developed primarily for Google's cloud computing business, the same technology supports Waymo's vehicles.

The onboard computers have to process enormous amounts of data in real time. Lidar systems generate millions of data points per second. Cameras capture high-resolution video from multiple angles. Radar tracks dozens of objects simultaneously. All this information must be fused together, interpreted, and acted upon faster than a human could blink.

In 2017, Waymo unveiled new sensors and chips that were cheaper to manufacture, cameras with better visibility, and even windshield wipers specifically designed to keep the lidar sensors clean. These mundane-sounding improvements matter enormously. A self-driving car that works perfectly 99.9 percent of the time but fails catastrophically the other 0.1 percent isn't good enough—not when that 0.1 percent could kill someone.

The sixth-generation Waymo vehicles, which began testing in July 2024, use electric vehicle platforms from Zeekr, a Chinese automaker. The partnership, first announced in 2021, represents Waymo's latest effort to reduce costs while expanding capability. Each generation of hardware has been cheaper, more reliable, and more capable than the last.

What Comes Next

Waymo's immediate plans are straightforward: more cities, more vehicles, more rides. The goal of reaching 20 cities by 2027 would make robotaxis a significant form of transportation rather than a curiosity limited to a handful of markets.

But the longer-term implications are harder to predict. If self-driving taxis become ubiquitous and affordable, what happens to car ownership? Why maintain a vehicle that sits idle 95 percent of the time when you can summon a ride whenever you need one? What happens to parking lots and garages when fewer people own cars? What happens to the millions of Americans who drive for a living—truck drivers, taxi drivers, delivery drivers—when machines can do their jobs?

These questions have been asked since self-driving cars first seemed possible, and the answers remain unclear. The technology's slow rollout has given society time to adapt, but "time to adapt" isn't the same as "actually adapting." Few cities have seriously planned for a future where autonomous vehicles are common.

Waymo has also expanded into delivery services, partnering with Uber Eats and DoorDash to bring food to customers in Phoenix. These services use the same vehicles and technology as the passenger service, just with meals instead of people in the back seat. It's a logical extension—if the cars can safely navigate to pick up humans, they can navigate to pick up tacos—but it hints at how deeply autonomous vehicles might eventually penetrate daily life.

The Quiet Revolution

Perhaps the most remarkable thing about Waymo's current operations is how unremarkable they've become. In Phoenix, seeing a car without a driver is no longer unusual. People hail robotaxis to go to dinner, to run errands, to get home from the airport. The vehicles stop at red lights, yield to pedestrians, and parallel park without drama. They are, in the most mundane sense possible, just cars.

That mundanity represents the culmination of nearly two decades of work and tens of billions of dollars in investment. The flashy demonstrations and bold promises of the early years have given way to the unglamorous reality of operating a transportation service: dealing with customer complaints, maintaining vehicles, navigating local politics, and grinding through the endless work of making a complex system reliable.

Sebastian Thrun, who started Google's self-driving car project, once predicted that autonomous vehicles would be common by 2020. He was wrong about the timeline but perhaps right about the destination. The cars aren't everywhere yet, but they're no longer nowhere. They're on the streets of Phoenix and San Francisco and Los Angeles, picking up passengers, dropping them off, and doing it again—thousands of times a day, without human hands on the wheel.

The revolution didn't arrive with a bang. It arrived with a quiet hum of electric motors and the soft click of automatically locking doors.