In 2025, there are two main ways you can expect to encounter self-driving technology. One is in your own new, or even relatively new, car: While Tesla has the most comprehensive and widely used driving-assistance features, lots of automakers — Ford, GM, Mercedes, and BMW to name a few — now offer features that will in some situations drive for you, with supervision and occasional intervention.
The other way is in the form of a fully driverless taxi appearing at your door — in other words, in a car you don’t own. After years of testing in a few markets around the country, autonomous vehicles from companies like Alphabet’s Waymo are going to start showing up in Uber and Lyft:
The ride-hailing leaders are preparing to bring driverless taxis to your door with new app features that allow customers to use their phones to open trunks and honk horns. They are building infrastructure to maintain the high-tech taxis and training human support staff to handle riders without drivers.
Both companies will have driverless cars — from Alphabet’s Waymo and others — on their apps this year. In the coming months, riders in Austin, Texas, and Atlanta will be able to hail a Waymo through the Uber app. Lyft plans to offer May Mobility’s driverless taxis in Atlanta.
The Wall Street Journal has some reporting on how rideshare companies are preparing for this rollout, and their choices provide a sketch of one plausible path for autonomous vehicles. Uber and Lyft will effectively be maintaining fleets of self-driving cars, which means building, or outsourcing, new infrastructure to support vehicles with particular needs: electric charging, specialized technicians, and storage. This won’t be a huge or particularly rapid rollout for a few reasons. One is that the types of vehicles Lyft and Uber will include in these programs are still in testing and only legal in a few markets. Another is that, while they cut out the cost of human labor, they’re not yet cheap to build, own, or operate. Waymo and others are racing to develop lower-cost vehicles, but the hardware on the road today consists of costly conversion vehicles laden with expensive, specialized hardware. For now, autonomous taxi rides aren’t necessarily cheaper than manned ones; in San Francisco, for example, Waymo rides are often the more expensive option.
But periods of self-driving acceleration have often been followed by stall-outs, and Uber and Lyft still face several challenges. They can reasonably expect capabilities to improve, although they don’t know how fast. They can assume that hardware prices will come down, although they don’t know for sure by how much and when. They can assume, broadly, that the regulatory environment will be more favorable in the future, but in ways that are hard to predict or control specifically. The basic bet here is that the next big thing for self-driving cars — and the first business case built around truly autonomous vehicles — involves acquiring and maintaining big fleets of specialized vehicles and renting out access to riders.
This is a more complicated proposition for ride-sharing companies than it might seem. Human contractors don’t just drive for Uber and Lyft; they also often own their own cars, pay for gas and maintenance, and assume a degree of liability for what happens to those cars and the people inside them. Rideshare companies get a lot more from people than the simple work of driving, in other words, and replacing them means either shouldering new costs and responsibilities or finding ways to outsource them. It’s easy to imagine a future rideshare company where self-driving cars support a simple, profitable business — indeed, Uber and Lyft spent billions of dollars trying to manifest that future for themselves, before pulling back and outsourcing self-driving tech to outside partners. But for Uber, which spent 15 years hemorrhaging money to become the dominant rideshare platform and just recently reported its first profit, autonomous vehicles represent both a long-term opportunity and a short-term challenge — a disruption to its business model as it exists today. If it goes all in on this, it would put it back in lose-money-to-make-money mode but presumably with an even bigger prize ahead of it.
That’s one bet on where this is going: that expensive, specialized, managed, but actually-self-driving cars that already exist are the way forward. The other sounds more like this, from Elon Musk:
… Think of every car that we sell or produce that has full autonomy capability as actually something that in the future may be worth as much as five times what it is today. Because average — vehicle is doing like maybe ten hours of driving a week. If sort of — if this says 1.5 hours a day on average, that’s ten hours a week-ish. If you’ve got on autonomous — if the vehicle is able to operate autonomously and use either dedicated autonomous or partially autonomous like Airbnb, like maybe sometimes you allow your car to be used by others. Sometimes you want to use it exclusively just like Airbnb — doing Airbnb with a room in your house. The value is just tremendous.
Now, Musk has been saying stuff like this for a long time. He has framed such comments both as a prediction about how self-driving might go mainstream — as a feature in private cars that gets better until it’s good enough to take over — and as a way to sell more cars now, by suggesting that Tesla owners might be able to turn their vehicles into moneymaking taxis when they’re not using them (Teslas with autonomous features are now “appreciating assets,” Musk claimed all the way back in 2019). But this has not panned out for a wide range of reasons. And given the stated timeline for Tesla’s recently unveiled Cybercab concept — a 2027 target from an admittedly “optimistic” Musk — brand-new Tesla owners shouldn’t expect to be able to put their cars to work, either. Musk’s affinity for this theory is inseparable from his role as the CEO of an automaker who makes money by selling cars. It’s also intertwined with Tesla’s fundamental approach to self-driving technology, which has seen it pivot away from reliance on expensive but effective sensor stacks like Waymo’s toward cheaper but less reliable camera-centric systems, a capability gap Tesla claims it’ll be able to close with advances in camera technology and, mostly, AI.
This, too, is based on assumptions that come with caveats. Driver-assistance-style features will get better, but it’s not clear how quickly; hardware will become more capable and cheaper, at some rate; more car owners will live in places where self-driving cars are legal to use, somewhere by sometime; Tesla’s ability to collect data from hundreds of thousands of cars on the road will give it an advantage, for some problems to some extent. Also like Uber’s plan, this sort of speculation should be understood in the context of major setbacks and contractions in the broader autonomous-vehicle industry, which saw the abandonment of Apple’s car project and the closure of GM’s self-driving taxi project in the last year. Lots of companies still believe autonomous vehicles are coming, and are spending money on versions of the concept. But they’ve also found progress toward real full-self-driving to be slower, less linear, and more expensive than they’re willing to bear, at least for now. Self-driving cars are both more clearly inevitable than they’ve ever been and turning out to be more complicated to deploy than they might have seemed when they were less obviously going to work at all. Not an unfamiliar technological predicament!
Still, both of these visions remain broadly plausible, albeit on different unpredictable timelines: autonomous cars as fleet vehicles versus autonomy as a feature in cars that owners mostly keep for themselves. It’s cars as a service versus cars as a strange, new, but semi-familiar, asset class. The two visions aren’t entirely incompatible — Lyft is already talking about building features for users to deploy their own small “fleets” of cars at some point — but with millions of semi-autonomous cars hitting the road, and early generations of fully autonomous vehicles actually driving people around at scale, they’re on course to eventually collide.
