Why robotaxi rides still lose money after years of hype

HostIt feels like we have been hearing about the end of driving for a decade now. You see those cars with the spinning buckets on top rolling through cities like Phoenix or San Francisco, and it looks like the future has finally shown up. But if you look at the books for the companies making these cars, the math just doesn't seem to add up yet. Even without a driver to pay, these rides are still burning through cash every time someone hops in the back seat. I wanted to sit down and really dig into why the dream of cheap, driverless travel is still such a money pit. Is it just a matter of making more cars, or is there something deeper in how these things work that keeps them so expensive?

GuestIt's a bit of a shock when you see the numbers. We were told for years that the driver is the biggest cost in a ride. If you take the person out of the seat, the ride should be almost free, right? But it turns out that when you fire the driver, you end up hiring a whole team of other people and buying a mountain of gear that costs more than the car itself. A normal car you might use for a ride-share app costs maybe thirty thousand dollars. One of these top-tier robot taxis can easily cost two hundred thousand dollars. Most of that's for the eyes of the car. They use these light-based sensors called lidar that scan the world thousands of times a second. They're amazing, but they're still very pricey to build and even pricier to fix when a pebble chips one on the road.

HostBut we see this with all tech, don't we? The first ones are expensive, and then they get cheap. My phone has a sensor in it that used to cost a fortune. Why hasn't that happened here yet?

GuestWell, the sensors on a car have to be much tougher than the ones in your pocket. They have to sit in the sun for ten hours a day, handle freezing rain, and never, ever fail while the car is moving at forty miles an hour. That kind of gear stays expensive because you can't cut corners on safety. But the real secret cost isn't the hardware. It's the people who are still driving the car from a room miles away. Most people think these cars are totally on their own, but they really aren't. They have a bit of a ghost in the machine. When the car gets confused by a stray traffic cone or a weirdly parked truck, it stops and calls home. A human in a call center looks through the cameras and tells the car what to do.

HostWait, if there's still a person in a room watching the car, doesn't that just mean we have built a very long, very expensive steering wheel? If you still need people to watch them, you haven't really solved the labor problem.

GuestThat's the big catch. Right now, the ratio of people to cars is still pretty high. In the early days, it was almost one person watching every one or two cars. Companies are trying to get that to one person for every twenty or fifty cars, but we aren't there yet. And those people in the office are often making more per hour than a gig worker would. Then you have to think about the physical car. When you drive your own car for work, you pay for the gas, the tires, and the car washes. When a company owns a fleet of robots, they have to pay for huge hubs to store them, specialized mechanics to fix the sensors, and teams to go out and rescue a car when it gets stuck or has a flat tire.

HostI guess I never thought about the simple stuff, like who cleans up the mess. If a rider spills a drink in a normal ride-share, the driver cleans it up right then.

GuestExactly. With a robot, that car might just keep driving with a sticky seat for five more trips until someone complains. Then the company has to take that car off the road, drive it to a center, and have a pro clean it. That's time the car isn't making money. And since the car is so expensive, it has to be moving almost twenty-four hours a day to pay for itself. But humans don't want rides twenty-four hours a day. We want them in the morning and the evening. So for half the day, these two-hundred-thousand-dollar machines are just sitting in a lot, or driving around empty to find a spot to wait. That empty driving is a huge waste of money.

HostSo it sounds like they're fighting a war on three fronts. The gear is too pricey, the office staff is a hidden cost, and the cars spend too much time doing nothing. Is there any part of this that's actually getting cheaper?

GuestThe software is getting better at handling the easy stuff, which helps. But as they move into new cities, the costs reset. A car that knows how to drive in sunny Phoenix has no idea what to do with a snowstorm in Chicago or a confusing five-way stop in Boston. Every new city needs new maps, new testing, and new office staff. There's also the insurance problem. If a human driver gets in a wreck, the insurance payout is pretty standard. If a robotaxi hits a person, it's a massive news story and a giant legal headache. The safety tax on these companies is huge. They have to spend billions of dollars on testing just to prove they're slightly safer than a teenager with a license.

HostIt feels like the bar for success is just so much higher for a machine. We forgive humans for making mistakes, but we don't forgive code.

GuestWe really don't. And that means these companies have to keep spending on research and development forever. They can't just stop and say the car is finished. They're constantly chasing that last one percent of weird road situations, like a man in a wheelchair chasing a turkey, which has actually happened. Until the car can handle those weird moments without calling for help, the person in the office stays on the payroll, and the ride stays in the red.

HostEven if they get the tech perfect, they still have to compete with a guy in a ten-year-old sedan who's willing to work for twenty bucks an hour.

GuestThat's the hardest part of the whole business, because the cheap car and the human driver will always be the baseline they have to beat.

HostIt turns out that replacing a human behind the wheel is a lot harder than just teaching a computer how to see the road.

GuestThe real test is whether these companies can finally stop paying for a chauffeur in the cloud and let the car truly stand on its own.

HostThose spinning buckets on the roof might look like they have it all figured out, but there's still a very expensive team of humans just out of sight keeping the whole thing moving.