Uber vs. Tesla, Robotaxi Timelines, and the End of Human Driving | Uber CEO Dara Khosrowshahi | #243

Khosrowshahi details that Uber provides a demand aggregator layer, fleet maintenance, cleaning, repair, and eventual financial ownership structures. Early on, Uber will buy cars directly; over time, it expects an asset-light model where financial institutions own fleets and Uber manages them, akin to hotel REITs. Standard APIs allow partners to plug into Uber’s demand and data, learning optimal pick-up/drop-off points. This frees AV companies from operational overhead. The company currently works with over 20 autonomous players (Waymo, WeRide, Pony.ai, Wabi, Nvidia, etc.) and will be in 15 cities by year-end.

Uber launched Uber AI Solutions to employ drivers in data labeling and model oversight. Khosrowshahi frames Uber as a ‘platform for work’ that has expanded from driving to delivering, shopping, and now AI tasks. The plan is to augment and eventually replace some driving labour, but in a managed way. He also details the option for drivers to become fleet owners, buying autonomous cars and earning passive income, thus shifting from labour to capital.

Khosrowshahi joined Uber during a severe crisis; he believes that the people who joined then were the kind who run toward a fire. He states that impact doesn’t come with a free lunch and that the company is upfront about the grind. He contrasts this with a culture of excessive prep—he wishes employees would just talk to him directly rather than practise with his avatar—but he accepts that the avatar helps some teams. The message is that working hard on difficult problems is fun and gives a higher order of satisfaction.

After 15–16 years of development, autonomous technology is now on streets, led by Waymo. Khosrowshahi argues there won’t be a binary outcome—either all autonomous or none—but a hybrid where fleets consist of both autonomous and human-driven vehicles. Uber’s unique position as a demand aggregator with massive rider bases allows it to offer autonomous partners a route to market. By year-end, Uber will be live with autonomous rides in 15 cities; by 2029, it projects facilitating more autonomous and robottaxi rides than anyone. It is working with over 20 partners, from Waymo to WeRide, Pony.ai, Wabi, Avride, Nvidia, and others. The platform provides standard APIs, data for training models (pick-up/drop-off optimization), fleet management, cleaning, repair—so partners focus solely on building a safe, affordable autonomous driver. Khosrowshahi notes that humans are actually more complicated than robots, so Uber’s existing infrastructure handles a lot of the complexity. Even Tesla is welcome when its camera-only FSD becomes safe for driverless operation; currently they have tens of thousands of Teslas on the platform, some using FSD with a human backup.

Khosrowshahi says that within 10 years, every single new car sold will come with autonomous driving software and a sensor stack (lidar, cameras) as costs plummet. This is based on current declining costs and industry trends. However, the average lifespan of a car in the US is over 10 years, meaning the installed base will turn over slowly. In developing markets, where Uber operates in 70+ countries, adoption will be even slower. He envisions autonomous costs dropping to the point where owning a car doesn’t make economic sense—the cost per trip will go down, safety will increase, and the user experience will improve with personalised settings. This will expand the total addressable market for mobility and delivery. The transition is not an overnight shift but a decades-long overlay that Uber intends to ride.

Khosrowshahi says the business is growing over 20% annually and driver churn is roughly 20% per year, so autonomous will primarily slow the recruitment of new drivers rather than cut existing ones. By 2030, he predicts significantly more drivers on the platform, including in the US. To handle eventual labour displacement, Uber is expanding the types of flexible work available: driving, delivering, shopping, and now data labelling through Uber AI Solutions, where drivers tag images and help evaluate AI models. The vision is to make drivers into fleet managers who own autonomous vehicles, turning labour into capital income. The company is consciously managing the social contract so that the technology wave doesn’t leave drivers behind.

The eVTOL service allows a user to push a button and summon an Uber to a vertiport, fly on a Joby aircraft, and take another Uber to the final destination—saving hours of traffic time. The technology is ‘absolutely real’, with Joby partnering with Toyota for manufacturing and safety. The first market is Abu Dhabi by end of 2025; the US to follow in 2026. Khosrowshahi identifies the Middle East (UAE, KSA) as the most forward-leaning region on innovation, with young, tech-savvy governments actively inviting such services. He also hints at Europe opening up, with UK partner Wave building a mass-market autonomous software stack for OEMs. The move completes Uber’s vision of ‘wiring up anything that moves’.

Khosrowshahi explains that delivery is harder than mobility because food doesn’t walk to the customer—pickup/dropoff precision is critical. For suburban areas without high-rises, drone delivery (with partners like SkyTrax, Mana) will be a ‘surprise and delight’: food prep may take 10 minutes, drone flight 2 minutes, total 15 minutes. In dense urban areas, sidewalk robots (Coco, etc.) are suitable for short trips within a mile or two because they travel slowly. Ultimately, specialized road robots using bike lanes will appear. Uber’s goal is to wire up anything that moves, making it a multimodal network on sidewalks, roads, and in the sky, fulfilling the demand for anything delivered within an hour.