Will robotaxis ever be commercially viable? | FT Tech
San Francisco has become ground zero for US testing of driverless taxis, but how close are we to truly embracing them? The FT’s Patrick McGee pays a visit to two of the world’s top robotaxi operators and gets an insider’s take on how the tech and the business model are panning out
Produced by Alpha Grid
You can enable subtitles (captions) in the video player
Is the driverless society inevitable? If you visited my home city, San Francisco, you might think it's already arrived. Right now, on my phone, I'm going to hail a Cruise. And a fully driverless, empty vehicle is going to come pick me up. All right. So our vehicle is on the way. It's called Vermicelli, like the noodles.
In August, the California Public Utilities Commission gave the green light to allow driverless taxis, otherwise known as autonomous vehicles or AVs, to carry passengers without restrictions in San Francisco. Hello, Vermicelli. Oh. So if you haven't been in a driverless car or robotaxi, you might think, what's so different about this versus an Uber? And to some extent, the answer is, not a whole lot.
But there are a couple of advantages. One, the cost is going to get to a space where it's just much cheaper than having a paid driver. Second, the availability eventually should be just ubiquitous. I mean, they should be everywhere. Third, of course, is safety. Arguably, they feel really safe already. And this is the least safe they will ever be, in the sense that they're only going to get safer.
I'm on my way to visit two of the world's top robotaxi operators and get an insider's take on how the tech and the business model are panning out. You certainly get strange looks from passers-by. But after a few minutes, you just feel comfortable, and you're liable to start taking business calls, or searching on your phone, anything like.
Yes, I'm getting things done now. Watch me run, run, run, run, run.
Cruise is owned by automaker GM. In mid-2022 it became the first company to offer driverless, paid-for rides to locals, currently available between 10:00pm and 5:30am, taking a lead over its San Francisco rivals Zoox and Alphabet's Waymo. I meet Cruise CEO Kyle Vogt at Cruise's headquarters.
It's execution and focus. That's really all it is.
Until now, Cruise has adapted conventional cars, growing its fleet from 50 to 350 vehicles within the space of a few months. It's now gearing up to launch its six-seater Origin model, which was created with the concept of ride-sharing in mind. Unlike current models, it has no manual back-up controls.
This is the first purpose-built vehicle made by a major automaker that doesn't have a steering wheel.
Cruise's parent company has confidently predicted revenues of $50bn by 2030.
We know the cost of the technology is going to come down. We know the experience is going to improve. We know the availability is going to improve. And so long as people have a desire to get from point A to point B, robotaxis are going to emerge as the best option. I think the other approach, where you start with something cheap, where you restrict the hardware, it doesn't work yet. And you're hoping you can get there in software. It's viable. It's just riskier.
But for investors and corporate backers, a looming question is how soon the huge investments in AV tech will pay off. One barrier is the speed of adaptation to a new city or the lack of it.
Autonomous vehicles have to be paired with an environment. They have to map out the city extensively. They have to spend months of testing to understand all the dynamics of how traffic works in the city, how pedestrians work. All of those things differ so much that you're seeing that the time it takes to roll out, city by city, is pretty substantial.
Both Uber and Lyft talked big about putting fleets of driverless taxis on the roads, but, ultimately, they both failed to deliver. But outperforming people-driven taxis is only the first step.
Big tech tends to see the current ride-hail market as the gateway to an era defining transformation of the transport network as we currently know it.
When self-driving cars become more economical, more cost-effective than car ownership, the demand will go up. And that gives us a beautiful so-called network effect, which is, the more people use it the more of those cars are around, the higher the utilisation.
Transit authorities may eventually need to reimagine their networks and embrace driverless vehicles as part of a new holistic model of mass transport.
The goal of all these transportation as a service providers is to change the American culture and the American way of doing things, which is a dependence on personal car ownership. They're trying to monetise a service that people don't want to pay for. And they're doing it in an environment that is stacked up against them.
My next ride will be in a Waymo cab to its San Francisco operations centre. You can see my initials on top there, so I know that's my vehicle. Waymo took an early lead in multi-city operations, running vehicles in San Francisco and parts of Phoenix, Arizona. Driverless cars may be financially appealing, in one sense, doing away with the driver, but people will still be needed elsewhere to recharge, maintain, and oversee the fleet. Even so, the upfront cost of kitting out cars with a mind-boggling array of sensors has become less of a problem.
The fifth-generation driver that you're seeing, that sensor suite, itself, is dramatically cheaper than the fourth generation, while actually increasing the performance.
A key metric is the cost of driving per mile. For human-driven cars, the complete ecology of refuelling, maintenance, and other costs is well-established and cheap. Driverless cars will need to hit those numbers or do better.
When you first start, you have a certain number of cars and so on. As the ridership increases, just the cost begins to - on a per-mile basis or a per-trip basis - just drops, just by virtue of the scale.
But while companies like Waymo strive to realise the grand vision of a driverless society, the immediate challenge is running a sustainable operation.
I firmly believe that self-driving cars is a viable business model. I'm a car owner, and my car is parked 97 per cent of its time and only being used 3 per cent. I feel this is irresponsible. I'm locking up resources, space for something that isn't being used. It'll be quite a path to break even. And the reason is that, today, engineering costs still dominate the costs of self-driving car companies. And then the cars are still very bespoke. They're not yet mass-manufactured.
Financial success may not, in fact, rely on instant business performance, rather which wealthy parent company can stay the course. Zoox is the city's third robotaxi operator, a start-up that was bought by Amazon for $1.3bn in 2020.
It wasn't in the spotlight as much as Cruise and Waymo were. But the fact that it's owned by Amazon, a company that has an unlimited budget for these long-term potential projects, that could mean that Zoox actually is better positioned, in some ways, than Cruise is, from a capitalisation perspective.
Eventually, the robotaxi landscape could look very different, unbundling services and moving away from the current model where an operator provides everything, including an app, the driverless fleet, itself, and the infrastructure to run it.
Siemens, or ZETF, or Schneider, or Alstom - all these companies that make trains, I mean, they're profitable industrial companies. They don't run the service. They make the technology that governments buy to run reliable, effective, affordable transport services.
Whichever specific form the robotaxi model takes, these industry leaders are bullish in their outlook.
First of all, we have to understand that AVs, one day, will be far, far safer than humans. So I think we're already past the point of asking whether we should have AVs and are they safe enough. It should be, how quickly can we improve them from here? And how quickly can we expand them, to have that net safety benefit in more places more quickly?
It is the case with almost any innovative technology, the work doesn't get done. It's continuously improves. We want to make sure that we are truly serving the riders, listening to their feedback, improving the product, and then expand to other areas.
Sometimes I've been cynical about driverless vehicles because they've never met the deadlines, the aspirations that were set out mainly from the years 2015 to 2018. But the slow-and-steady approach is also the one that, as a consumer, rather than a journalist, I want to support, right? And the more miles that are tested, the slower the approach is, the more that the cars are going to be safer and probably have a more relevant future of being used daily.
And if robotaxis do prove to be viable...
Almost there. Don't forget your belongings.
...it wouldn't be the first time that big tech has banked on a radical shift in human behaviour to sell its technology.