As AI reshapes industries, trade tensions deepen and societies adjust to new realities, the new economy inspires both hope and unease.
Join this conversation for an open conversation on seeing beyond today's short-term pressures and imagining how the new economy can work for everyone.
In a wide-ranging conversation at Davos, Uber CEO Dara Khosrowshahi argued that AI is already core to Uber’s operations—pricing, routing, matching, and recommendations—yet the most disruptive impact will come from autonomous vehicles (AVs). “If you think about the world 20 years from now, your Uber is going to be driven largely not by [a] human being, but by a robot driver.” He framed AV adoption as a societal design choice, not just a technical one: “How safe is safe enough?” While AVs can be “safer than a human driver,” he suggested society may ultimately raise standards far above human performance, potentially reshaping who is allowed to drive.
Khosrowshahi tempered near-term investor expectations, citing economics and scaling constraints: AVs cost “well over $100,000,” require new fleet operations, and will be limited for “the next 3 to 5 years,” reaching significance over “10 to 20 years,” especially outside high-fare markets. Financing, he said, will be “revenue based” because AVs have “zero residual value.”
On enterprise AI, he warned against superficial adoption. Real winners will “throw everything away” and rebuild processes, not add an “AI veneer.” Uber is applying this to customer service, shifting from rigid policies toward outcome-driven AI reasoning.
Welcome. I'm Stephanie Flanders, head of economics and government at Bloomberg, and this is an honest conversation on the hopes and anxieties of the new economy. Like all the conversations this week seemed to be honest, I don't know what they were before. But anyway, Dara Khosrowshahi, you are the chief executive of Uber. We were having this conversation before a lot of this week when he's not talking about Donald Trump is going to be thinking about AI. And the big question particularly kind of embedded in those valuations in Wall Street and elsewhere is how and whether and when companies will really be able to realize the benefits of AI. And I guess, as much a question for investors as for ordinary people. Uber is a company that we completely understand. Unlike many companies, we totally understand how you would how you would get the benefits of AI. But now just wondering how many jobs it's going to lose and how quickly you're going to do it.
Yeah, absolutely. So for us, AI has been a huge part of how we run our company. It's to some extent under the covers, but the price at which you're charged for an Uber, the route that the car takes, the driver that you're matched up with, the restaurants that we offer you on your Uber Eats feed, all of those are powered by AI algorithms, simpler AI algorithms than, let's say, some of the more complex llms that are now becoming increasingly a part of our reality. So AI has always been a part of our genetics. It's how we build the company. It's how we scaled. It's how we're able to have essentially one system that is able to operate effectively in Lagos and is also able to affect up to operate effectively in New York as well. These AI algorithms can adjust to their circumstances based on, a generalized set of rules, and then based on the data that they see in real life, they're able to act accordingly. Now, for us, the revolutionary, application for AI is as it relates to autonomous driving. And if you think about the world 20 years from now, your Uber is going to be driven largely not by human being, but by a robot driver. That will be a piece of software on top of a car, which, the cars are going to get much more sophisticated. It'll be, you know, more and more cars are going to be computers on wheels. And so we are our first kind of order of business is working with the ecosystem. There are many AV providers out there that are looking to build these software drivers, so to speak. And the way that we're positioning ourselves is working with the entire AV ecosystem. Just as, if there's a Mercedes or a Toyota or a Chevy, we all want them to be on a platform. If there's a Waymo or a Tesla or an Ave ride or a we ride, these are all autonomous, driver providers. We also want them on the Uber platform. As long as it's a safe, qualified, licensed driver, we want them on the Uber platform. And we will amalgamate all of this content, and essentially offer up to, the populations of the cities in which we serve in a delightful way and easy, delightful way. I think at the same time, there are a lot of questions that we're asking ourselves, and societies and local cities are asking about this AV reality that is starting to become a part of reality on the streets. One is how safe is safe enough? There's no doubt in my mind that the robot driver can be safer than a human driver. It doesn't get tired, doesn't get distracted, doesn't text, can work all day and all night with no problem whatsoever. And it's constantly getting better. The average AV driver will have driven, you know, hundreds of thousands of more miles than the human driver and is constantly learning. Every single driver is learning from the other driver as well. So this is a driver that will be superior to a human driver. And if you're being perfectly logical, you would say, well, if a robot driver is safer than a human driver, then that's kind of the bar that should be acceptable. In the US, for example, there are 30,000, fatalities, human fatalities from cars one way or the other. And we can being road, we can improve road safety significantly with these robot drivers. Now, the potential of this technology is that it can be five x ten x eventually 100 x safer than drivers as well. So I think one of the questions that society has to ask itself is, is it enough to be better than human? Or do we take the opportunity actually to set the bar much higher? And that bar may be five x ten x better than a human? We can.
Then you have to stop humans driving.
You know, at some point. Right? It's it's, I do think again, that's another choice for society as well. It could be 20 years from now that, driving will be like horseback riding. People who are fans of a certain type of transport that used to be the predominant type of transport. Do it for fun. They have to be specially licensed to do so. But there's no doubt that ten years from now, there will be questions as to whether humans are safe enough as well. But the question we've got to answer right now is, how safe is safe enough for robot drivers? Is it is the human the test, or should we kind of raise the bar even higher? I think other questions that we have to ask ourselves are, you know, kind of the societal impacts as well. Obviously there will be job displacement. The job displacement will be gradual. But in driving in many other roles, whether they're information technology roles or otherwise, these robots are going to first augment humans and in some cases take the place of humans as well. How fast do we want to modulate that? Do we want to modulate it at all? I think is a question that we're asking ourselves. And then the availability of this technology to society as well. When we launched at Uber, one of our core precepts was that we want to build a service that's available to everybody in the city that is not just available to people in the center of the cities where more wealthy people live, but also is available the outskirts of the cities and the transportation deserts in the neighbors neighborhoods that are not, let's say, served by, by public transit as well. You see, some of the newer services essentially map out the center of cities, but not necessarily the outskirts of cities. You know, Waymo isn't available in Oakland yet. Should it be or not? Should we make sure that this technology is evenly available to all parts of these cities versus the wealthy centers of the of the cities as well? And then last of all, what are the interactions of these services with emergency vehicles, etc.? Again, you've had times when, when the power's out, these vehicles cause huge snarls of travel, etc., of traffic, ignoring school buses. So I'm incredibly optimistic about AV as, as radically improving safety on the streets, ultimately bringing down the cost of transportation and making it available to far more people. But we also have to, be careful in that as we introduce this product to society, we also slow down and ask the questions as to the effect of its society, not just safety, but fairness and transportation availability for everyone. And and we want to have that dialogue now, not 5 to 10 years from now.
I've heard you talk about this before, and you have this. You say 20 years time, it's going to be totally transformed. It's not going to happen tomorrow. And that sort of gets over a lot of the hopes and anxieties. It sort of pushes them. But of course, for an investor, I'm sure the investors that you talk to, the difference between tomorrow and 20 years from now is pretty big. You're in a very good position to judge how long it's going to take for autonomous vehicles to really be a meaningful relative to the size of a business like Uber. What's your best guess?
Well, I think that, the biggest factor in terms of autonomous vehicles, obviously there are a lot of there are a lot of players that are trying to get to human level safety, and there will be multiple players who will cross that line. But it is the availability of affordable vehicles that have the sensor kits and the compute necessary to pass and then exceed that level of safety. I think for the next 3 to 5 years, a number of vehicles is going to be somewhat limited. And so while it will be 5 to 10 to 20% of of trips in higher fare markets where the economics of those vehicles are going to work early on, it will take some time. It will take 10 to 20 years before Avs become a significant part of the business, especially outside of high fare markets like the US and Europe.
And that's and that's not just on safety grounds, though, because I mean, economically, they're currently much more expensive than.
Much more. expensive people.
Who, you know, clean their own cars.
It's about cost and production capability. The cost of these vehicles now is well over $100,000. The human driver does a lot that we take for granted that clean the car, repair the cars, position the cars, etc.. So you will have to build out an ecosystem. There'll be a financing ecosystem that will be financing these fleets. There'll be an ecosystem of essentially fleet operators in every single city, both large companies and smaller companies. And we're building that ecosystem to make sure that the fleets are clean, repaired, etc.. And then, of course, there'll be a driver ecosystem and then demand ecosystem that we bring to the pie as well. That will take some time to, to develop. And again, it while there is a significant amount of capital going after the space, it'll be it'll be at least five years before it becomes a significant portion of the business.
And in approaching those five years, you talked about the, the financing for, for AV fleets. And I think it may be financial rights and others. Yes. What kind of milestones do you think they'll be needing to see before they go into that business? I mean, it's obviously it's going to be over time, but.
Well, they'll have to the, the complication with with these assets is that there's essentially no residual value for the assets. Car financing is obviously is a huge market there, but there's a very accepted residual value for a, a used Toyota. These cars, these AV robots have zero residual value. So the the financing model will have to be essentially revenue based. And one of the advantages that Uber has is that we know today how much essentially a human driver can bring in terms of revenue. We take about a 20% cut, etc., of that of that revenue. So 80% goes to the human driver. We can sign up for a revenue model that is has a high degree of certainty. And as a result, today we're working with very large financial institutions to essentially prove out, the financing model for these fleets. We will take balance sheet risk to begin with. But as we prove out this revenue model, then we see significant capital players who are very, very interested in coming into this market. It's a non-correlated asset, so to speak. It will be a very, very large market. And essentially, just like you see these retail real estate investment trusts that own hotels, that own shopping malls, etc., trading in the public markets. We think you'll see kind of these fleet entities as well, public and private insurance companies owning fleets in different cities. And we will be a very big part of that.
You think. But in terms of proving the case for that will be entirely revenue based. And you'll have the yardstick of Uber currently.
Exactly.
Okay. When you're working, I think you've mentioned in the past that as you have trouble sometimes on the hardware front, and I just wonder, are you finding it difficult? What are the challenges of trying to get car manufacturers to work with you in some of these areas?
Yeah, I think the car manufacturers are finally seeing this market as a very, very large potential market. So the issue is in having trouble on the hardware. It's bringing the cost of hardware down. And I think we're probably two generations away. Right now you can build a hallway, but every build is.
Bespoke generations.
Two car generations. So it's the next generation of cars. You won't have bespoke, bespoke builds. The cost of sensors is coming down very significantly. The cost of computers coming down. Nvidia is building a reference architecture that we think is very, very good for, for the for the industry generally, but it will take two generations of hardware to get to pricing below well below $100,000, which truly makes us a mass market product and something that we're very, very much looking forward to.
Are there particular manufacturers who are closer to that than others?
Every car company is talking about it. I wouldn't say anyone is in the lead. This market is a market that's ready to be taken by, OEMs that take a bet on it.
When you talk, when you look at, as I'm sure you do, some of the assumptions built into investors models, the ones who are most bullish about this, which seems to be all the people who are buying, how stretching. I mean, I wouldn't expect you to say, oh, it's all overvalued, but how much, in your view, are they conservative, or are they highly optimistic in terms of how long it's going to take to make revenue from some of these things?
You know, the AV, the the market for AV investment, it's mostly private companies. And Waymo is the leader there. But there are a number of Chinese companies out there. And I don't think actually that market is particularly overheated. I think that the potential is trillions of dollars in incremental spend, transportation spend. I think, again, every single vehicle sold, new vehicles sold ten years from now is going to be L3 or L4 ready. So I think the market is enormous. And, I don't think the market is overheated by any means. Now, the overall AI market in terms of foundation models, hardware providers, semi semi providers, etc., that is a market that, as you know, has heated up significantly. Consumers or investors are paying for very, very aggressive assumptions. Forward growth. And you know, I'm not an investment business. I'm not in the valuation business necessarily. But everything that we're seeing on the other part of AI, the the potential for AI to transform how businesses are run, the potential for AI to make every human into a superhuman in terms of whether you're a researcher or a marketer or a developer. It is absolutely game changing. And we are very much leaning into AI, also changing kind of how Uber operates as a company. Every single one of our developers using AI, using cursor cloud code, etc., and the productivity gains that we see in terms of the throughput of, for example, our developers, we have never, ever seen these kinds of advances, and these kinds of productivity gains. So I do think that what you're going to see is, I think where investors can do very well is that finding companies that are truly looking to transform themselves using AI versus companies that are saying the right words and kind of play, acting their way into a pretend transformation. And I think the companies that truly lean into AI and change the way that they work, are going to be the winners. And I think kind of the, the investor's job is to recognize the play acting from the real stuff. And I will tell you that, that that truly changing how you work with AI. We found it to be much harder than it sounds. It's pretty easy to put kind of an AI layer and let's say improve productivity by 10%, 20%. You know, have your sales team have an AI agent summarize a client and what a client pitch for you. That's the easy stuff. Anybody can do that. That's not going to differentiate you, but actually changing your processes from the ground up, actually throwing everything away and not putting kind of this AI veneer, but actually using the power of AI from the ground up, that is actually much more challenging. You, you have to survive through a bunch of car crashes internally to do so. But those are the kind of transformation, transformational events that will separate companies that truly take this to heart versus companies that are kind of playing make pretend.
We're going to. There's so many things to talk about. We're going to run out of time. But just it's sort of tantalizing what you've just said about the pretenders versus the actual Transformers. And I suspect Davos is full of both. You wouldn't necessarily want to distinguish.
The pretenders here.
Yeah. But knowing what you know about how hard it is when you look at across sectors, where would you be most suspicious that they can, as easily as they purport, really transform their business? I mean, what are the kind of sectors that you would be most skeptical of, like immediate transformation?
I don't think it's about sectors. It's about culture. Every single sector has a potential to transform themselves. I'll give you an example. We are using AI to transform empower our customer service interactions. Right. We want our customer service to be better. We want it to be cheaper. All the good stuff. And there's if someone calls Uber with an issue, there's a number of things that happen below the covers. One is, you know, you call us, it's Stephanie, a great customer or a fraudster. Lots of fosters out there. What's the issue that you describe our understanding what your issue is. How does that issue compare to the real world that, you know, you say your food is 20 minutes late. Well we know when did you order, when did we promise it, etc. we can actually confirm that or understand that what you may say is not true. Based on that confirmation, what are the policies that we have? And then what do we do about how do we communicate back to you in terms of the resolution, do you get your money back, etc.? All of these are processes that have been put in place, built out over 15 years, etc.. There's you can have essentially what we found is that building out AI that essentially follows the policies and follows the procedures that we put in place for many, many years is actually, working. But it's working. It's okay. It will get some people out of the loop. It will save us some money while actually completely rebuilding things from scratch and actually telling the AI agent, what is it that you're after? What are you trying to accomplish? I want a loyal customer like Stephanie to be treated well and to feel great after this transaction. And I don't want to spend all the money in the world, let's say, and allowing the AI actually to reason through that and throwing away all of the old policies is turning out to be the most promising way forward. Now, we fell upon this approach because a couple of our developers who were frustrated at kind of our old approach said, screw this, let's throw everything out and let's try to build this all from the ground up. That's the kind of work that it takes to actually make a breakthrough. That's not about what industry you're in. It just goes to actually fundamentally ground up rethinking how you work versus building yet another layer on top of this kind of layer that companies have been built on. You know, companies have been built that there are rules, you know, what's a company, essentially it's a rule. It's a bunch of policies that say you're working for this entity and you all have a common goal to some extent. You've got to break down those rules and start over with AI in order to get the full potential of AI inside of your company.
Just briefly on, you did mention in passing that you thought Waymo was perfectly it was made sense the valuation. And does that mean that you're actually going to Uber might take part in the funding round that they've got?
No comment on? You thought it was such a good, such a good price. Listen, they've got plenty of money that they don't need our money, but we are investing in the entire ecosystem. We're very, very bullish.
You wouldn't rule it out, though. What's that? Sounds like you don't rule it out. I mean, there's such good value.
Don't rule it out. But again they've got plenty of money.
You've obviously you know, you're one of the now, sort of poster child for not not necessarily AI, but certainly the new economy that people are so anxious about. One of the things that's getting more and more scrutiny is dynamic pricing. I mean, as an economist, I think dynamic pricing is at least a beautiful thing in in theory, yes, but a lot of people find it in practice annoying, unfair. It gets to people's sense of of fairness as consumers. So are you worried that there's going to be a kind of more widespread questioning of that? I mean, it's obviously at the center of your business.
Yeah, I think that there are there's a version of dynamic pricing that takes advantage of the customer and basically tries to maximize maximize company value. How much is a customer willing to pay? And let me target that customer individually. We think that's a we think it's right to push back against that, that kind of dynamic pricing because it's essentially it's a win lose.
Was that do you think that's more the sort of Instacart type?
I don't know whether that's what Instacart is doing. But but some but we see some companies doing that. Well tickets the. Yeah exactly. The purpose of dynamic pricing for Uber is to retain the reliability of the service. There are times in which the demand for service exceeds the supply for that service. And in order to get more supply into the roads, you know, if there's a concert or there's a sports event that is winding up in order to get more drivers onto the road at 10 p.m., which is not a great time to be driving around, we need to pay drivers more. And so in order to pay drivers more, we have to up the price of an Uber ride to then let drivers know 10 p.m. at night at the, you know, Oakland Coliseum. You're going to make a lot of money. And so that's the goal of dynamic pricing is to retain reliability for the service. And that is essentially serving the driver because they get to make a lot of money. And it's serving the the rider as well because they get higher reliability should they choose to use that service. That is, I would say, a constructive methodology of dynamic pricing versus let's say it's not destructive, but it's certainly a company saying I'm going to maximize my earnings at the expense of a customer's willingness to pay. Those are two very, very different circumstances.
I guess, is what you want dynamic pricing, where there's also dynamic supply. If the supply is fixed, then you're just getting the maximum amount of money.
Exactly.
Yeah. Yeah. You are I mean, there's obviously been we, you know, over the years, we read about Uber's at war with Barcelona or they're at war with the mayor of London. Just random examples. But the new mayor of New York, seems to be pretty concerned about, for example, the new way of, having tips, being able to give tips on the Uber platform. Have you talked to him about it? He was pretty angry.
I haven't spoken to him specifically about it.
You have spoken. Have you spoken to him since he became mayor?
No, no, I spoke to him before he became he became mayor. Listen, I think that if it's, If he's serious about affordability, in terms of affordability, rents, affordability, food, etc., then, we would love to talk to him, because a lot of when you look at New York City and, for example, the cost of Ubers in New York City versus, general cost of living in New York City, the cost of Ubers in New York City have increased far faster than the cost of living.
Due to regulation.
Yeah, the average driver in New York City Uber driver is making 50, $55 an hour. And that's because of the rates that the taxi limo commission is setting, tons of taxes, etc., fees that are inside of the system. And so if if he's serious about affordability, we'd love to talk to him about that, because we think Uber's unfortunately in the city are becoming more unaffordable, and it's actually because of overregulation than the other.
Why don't you call him up? He's quite busy, so.
He's busy right now. But yeah, we'd love to have a conversation with him.
I'm going to get on to questions in a minute, but I one of the things that's sort of evident, we've obviously had a big war of words between Europe and the US over Greenland, which I assume you don't have a strong view on.
There's no Uber in Greenland yet.
We'll see. But there is a sort of concern that I guess there's two things. I mean, just companies that are active in as many jurisdictions as you are and are also fundamentally felt very obviously an American company. Do you worry about sort of the amount of retaliation back and forth, but also specifically that it's a bit more kind of high risk being an American company in some of these places than it used to be?
I worry about it as an American citizen. As, let's say, globalist, and that that may not be in in vogue, but it's not going to affect our business that much. You know, ours. Uber is fundamentally a local business. Even the money flows with Uber. You know, if you take an Uber here, 80% plus of the money is staying here, not just in the country, but actually in the town, you know, from consumers who can afford to pay two drivers, etc., who are making a living in their local in their local business. So we tend to be relatively unaffected by global conflicts, etc. because fundamentally, while we're global platform, we're very much a local business. So it's not something that affects us.
You know, you get affected by local conflict enormously.
By local conflict, sure, but not by some of these global political issues.
Okay. We've got five minutes and I will take 2 or 3. But if you could just be brief. Yeah. So wait for the mic.
Hi. Whether we're driven by a robot or.
Could you just.
Say yes. So I'm Sue chan. I'm from the Telegraph in London. Whether we're driven by a robot or human in the future, do you think the future of driving is still electric? I know Uber wanted to make London the first net zero city by last year. Are you pushing back that target?
We are pushing. So our target of being all electric by 2030, that's just not going to happen based on everything happening in society, etc.. We continue to increase the percentage of EVs on Uber. I think one of the benefits of AV, of autonomous, is that the the vehicles, the the vehicles that can carry kind of, the sensor kit compute, they all happen to be electric. So the autonomous revolution will also be an electric revolution. And we think that's a great thing.
Yeah.
Hi. Thank you for an amazing session. I am from Pakistan, and Uber happens to be a household name along with other competitors for all the ride hailing apps that we use. So my question is, you know, you're you're talking about innovating and you're actively working across it. But when we talk about expansion to the global South, that is not very technically friendly. How do you plan to navigate that expansion into our part of the world or, you know, parts of the world that are like ours while being technically equipped because we're not technically equipped till now? So, like, what's the plan ahead?
Well, I think I think that, as long as there are cell phones and broadband and decent cell service, we'll get into these markets. So I think the global South Africa, for example, in the Middle East, etc., these are all markets that we are targeting and we're getting into, it's it's it represents the the next 20 years of growth for us. We're in 15,000 plus cities and our businesses, our systems are designed to operate both in low cost markets and in high cost markets as well. So you should you should expect to see us in many of the markets in which you're talking about.
You're testing the we've talked a lot about the Avs, and I know that you're testing, with supervisors at the moment. Correct. In a so presumably you have to do the driverless testing. Are we going to have driven in a Uber robotaxi by the time we meet again in a year's time?
I'm certainly hoping to get some robotaxis here if we can. But I think it'll be a heck of a lot of fun.
But in terms of but actually in terms of operating in cities, will a customer of Uber be in a robotaxi within a year?
It depends on where you live. If you live in Atlanta and Austin, you'll be getting robo taxis. You can get robo taxis on Uber today. Abu Dhabi, Dubai, London, hopefully coming up. So there will be many, many cities in which, robo taxis will be available as part of the Uber platform. Absolutely.
London this year.
London isn't here yet, but we're talking to London and we would love London to be a part of robo taxis either later this year or the next year. London is definitely leaning in, and it's terrific to see the UK in general leaning in on AI and, the transformation there. The talent base there is excellent. And we would love to play a part.
Thank you very much.
Thank you.
Appreciate it.