Uber Bets on Human-AI Hybrid Network to Compete in the Autonomous Taxi Market
Uber is leveraging its network of human drivers as a core asset in the competition for autonomous taxis: scheduling both human drivers and self-driving vehicles within the same app, using human labor to fill the gaps in autonomous driving capacity, in order to compete against closed fleets like Waymo and Tesla that only use self-driving cars.
Uber CEO Dara Khosrowshahi refers to this strategy as the "hybrid future," which seamlessly integrates human drivers and self-driving vehicles into the same market. Uber claims that in Austin and Atlanta, self-driving vehicles on the Uber platform have 30% more daily orders per vehicle compared to pure self-driving platforms, and passenger wait times are reduced by 25%. Khosrowshahi revealed during the Q2 earnings call that the average daily orders per vehicle on the platform typically range from over 20 to just above 30.
Uber's autonomous driving business is currently operational in seven cities, with a goal to expand to 15 cities by the end of 2026. The proportion of autonomous driving orders to total orders is still less than 0.5%. Uber has committed over $10 billion to invest in autonomous driving software companies, build ground infrastructure, acquire real estate, and secure vehicle procurement, with a total commitment of approximately 120,000 vehicles from related partners.
The partnership network covers multiple technological routes: Nuro and Lucid supply vehicles in the San Francisco Bay Area, planning to procure up to 20,000 vehicles; Zoox and Motional operate in Las Vegas; Wayve is responsible for London and Tokyo; Baidu has also entered London; Pony and Verne operate in Zagreb; Rivian has an order for 10,000 vehicles with an additional option for 40,000, deploying in San Francisco and Miami starting in 2028; NVIDIA plans to enter Los Angeles and San Francisco by 2027. Additionally, Uber has placed an order for 25,000 vehicles with Waabi.
On the regulatory front, Uber has been approved to operate 1,000 autonomous taxis in Clark County, Nevada, over the next 12 months, executed by Motional and Zoox, with Zoox holding a separate license for 100 vehicles. Uber's core argument to local governments is to first use human drivers to meet peak demand, gradually integrating autonomous vehicles into urban traffic, rather than deploying a purely autonomous fleet all at once.
Competitors are "de-Uberizing." Waymo has ended its three-year partnership with Uber in Phoenix as of May 2026, with collaborations in Atlanta and Austin expected to terminate by 2028. Waymo currently operates in Phoenix, Atlanta, and Austin, and has partnered with Lyft to enter Nashville; approximately 70 remote assistants support around 3,000 vehicles. In the high-end market, Uber has acquired the private car booking platform Blacklane, while Lyft has acquired TBR Global Chauffeuring, with both betting on the high-end segmentation of "human driver services."
In terms of market mechanisms, the buyers are demand aggregation platforms like Uber, which exchange order density for vehicle supply from autonomous driving companies; the sellers are autonomous driving software companies and automakers, which need high utilization rates to spread the high costs per vehicle. This round is not driven by a single event but is a structural capital allocation: Uber's free cash flow in Q2 was $2.8 billion, and its rolling 12-month free cash flow has surpassed $10 billion for the first time, with this cash flowing into equity and vehicle orders from autonomous driving companies. Beneficiaries include supply-side companies like Nuro, Lucid, Zoox, and Wayve, which lack demand entry points, as well as flexible part-time drivers (61% of drivers are online for less than 20 hours a week). The pressured parties are Waymo and Tesla, which are pursuing vertical integration: they must build their own demand networks while bearing the costs of idle vehicles during non-peak hours.
Source: Public Information
ABAB AI Insight
This is not Uber's first bet on autonomous driving. In 2015, it poached about 40 robotics researchers from Carnegie Mellon to form the Advanced Technologies Group (ATG). Former CEO Travis Kalanick defined autonomous driving as a matter of the company's survival. In February 2018, Uber settled a lawsuit with Waymo over the theft of lidar trade secrets, paying approximately $245 million in Uber equity. In March of the same year, an ATG test vehicle struck and killed pedestrian Elaine Herzberg in Tempe, Arizona, marking the world's first fatal accident involving a self-driving car and a pedestrian. In December 2020, Uber sold ATG to Aurora and invested $400 million in Aurora. This ended its self-development route, shifting Uber to a "platform + external supply" model.
In terms of capital strategy, Uber shifted its focus from "developing technology" to "buying supply shares." In 2025, Uber invested $300 million in Lucid, tying Lucid's vehicle orders with Nuro; it later participated in a $1 billion financing round for Waabi and invested $100 million to build electric autonomous vehicle charging infrastructure. In January 2026, Uber established AV Labs to collect road data using its own fleet, which will then be supplied to partners. Each investment comes with vehicle procurement or exclusive access terms, essentially locking in future capacity through equity and orders to prevent autonomous driving companies from bypassing Uber to directly serve passengers.
The industry landscape resembles the past competition between Android and iPhone. Waymo and Tesla pursue an Apple-like vertical integration model, building their own software, vehicles, applications, and scheduling; Uber adopts an Android-style open aggregation model, integrating over a dozen suppliers including Nuro, Zoox, Wayve, Baidu, and Pony. Lyft sold its self-development department Level 5 to Toyota's Woven Planet for $550 million in 2021, following a similar path to Uber's shift away from self-development, and both companies are now using hybrid networks and high-end human services to hedge risks. The industry is still in the early stages of expanding from "establishing footholds" to "urban-scale" operations, with autonomous driving orders accounting for less than 0.5%, and control has yet to be solidified.
Structural judgment: This represents a transfer of pricing power, with the focus on who controls pricing power on the demand and supply sides. Autonomous taxis are heavy assets with high per-vehicle costs and rigid depreciation, and profitability entirely depends on utilization rates; however, travel demand exhibits clear peak and off-peak periods. A purely autonomous fleet would face significant idle times during off-peak hours if it is deployed based on peak demand, while it would be underutilized during peak hours if deployed based on average demand. Human drivers provide flexible supply that can be activated or deactivated at any time, perfectly filling these fluctuations. Therefore, Uber can schedule autonomous vehicles only during high-utilization periods and allocate overflow demand during peak times to human drivers, thus pressuring autonomous driving companies to lower their revenue share. Waymo's exit from the Phoenix partnership is a response to this trend of pricing power shifting towards aggregation platforms.
ABAB News · Cognitive Laws
Fixed assets fear idleness, flexible supply enjoys premiums.
Technology determines whether one can run, demand determines who sets the price.
If you can't build the engine, then monopolize the accelerator.