· via The Verge
Waymo and Serve Robotics step back from Uber as autonomy firms go direct
Waymo will end its exclusive robotaxi deal with Uber in Austin and Atlanta by 2028, and Serve Robotics will let its Uber Eats partnership lapse, as robot operators move to own their customer relationships.

Two of the biggest players in physical-world autonomy are walking away from Uber. According to The Verge, Waymo plans to end its exclusive robotaxi partnership with the ride-hail giant, while Serve Robotics has decided not to renew its Uber Eats delivery deal when it expires in 2027. Together, the moves suggest that companies building autonomous robots increasingly want to own the customer relationship rather than operate through an intermediary.
The reversal came quickly. On a late August earnings call, Uber CEO Dara Khosrowshahi described the Waymo partnership as "very strong" and said he expected the company to keep operating in those markets next year. Several weeks later, Waymo co-CEO Tekedra Mawakana revealed in an interview that her company planned to end the exclusive arrangement, describing the mixed-fleet experience for riders as a "crapshoot." Starting in January 2028, riders in Austin and Atlanta will be able to hail Waymo driverless rides directly through the Waymo app.
The mixed-fleet problem
Consistency sat at the heart of the disagreement. Uber's model paired riders with either a robotaxi or a human driver, whichever was closer. But as The Verge reports, many Waymo riders preferred to know exactly what they were getting. TC Eley, an assistant professor at the Villanova School of Business, told the outlet that predictability matters enormously early in adoption: riders who specifically choose a robotaxi but get a human driver lose the chance to form their own impression of the technology, while riders who did not want an autonomous vehicle and receive one anyway may be put off for good.
Uber disputes the suggestion that it refused to adapt. According to Bloomberg, the company proposed a dedicated robotaxi button in its app for Austin and Atlanta, an offer Waymo declined.
Serve goes its own way
Serve Robotics shows the same dynamic playing out in delivery. The company runs more than 2,000 sidewalk delivery robots across dozens of US cities, which The Verge notes makes it the second-largest autonomous fleet in the country behind Waymo. Rather than renew with Uber Eats, Serve launched a direct developer API that lets any small business plug robot delivery into its own website and checkout.
The decision was largely economic, CEO Ali Kashani told The Verge. Delivery apps typically take a 15 to 30 percent cut of every order, largely to cover the cost of human couriers. By letting bakeries, florists, and pharmacies connect straight to its fleet, Serve wants to spare local shops those middleman fees. Kashani framed it as a fundamental difference in vision between the two companies, each doing what is right for its own business.
Uber spokesperson Noah Edwardsen placed the blame elsewhere, saying in a statement that Uber believes autonomy remains key to the future of both delivery and mobility and continues to scale with many partners. He added that Uber repeatedly raised concerns with Serve about operational performance and reliability beginning in early 2026, and that the situation never improved enough to meet its requirements.
Uber's platform bet
The defections are a setback for Uber's ambition to become the default app for autonomous rides and deliveries. The company still counts more than 30 autonomy partners worldwide, including Amazon's Zoox in Las Vegas, China's WeRide in Abu Dhabi, Dubai, Riyadh, and Madrid, and Avride, which has vehicles and sidewalk robots active in Dallas. Uber says it expects tens of thousands of unsupervised autonomous vehicles across more than 50 cities by 2028, and a Jeffries analysis it cites found its ride-hailing business growing faster, and taking more market share, in cities where robotaxis have operated for a while, such as San Francisco, Phoenix, and Los Angeles.
Harry Campbell, who writes The Driverless Digest newsletter, described Uber's role as a "Trojan horse" that introduces autonomous vehicles to riders who would never have sought them out. But he also noted that successful partners eventually want to split off, and estimated that Waymo realized within roughly six to twelve months that the experiment was not working well.
A history of bad blood
The break-up marks another twist in a tangled relationship. In 2017, Waymo sued Uber over trade secret theft and patent infringement. The trial ended abruptly a year later with a surprise settlement after a week of testimony. Uber later admitted it had misappropriated Waymo technology and agreed to license it, and former Google engineer Anthony Levandowski was sentenced to 18 months in prison for stealing Waymo trade secrets.
Why it matters
Autonomy has become a genuine commercial business, even if not yet a profitable one, and the companies doing the physical work no longer want an intermediary standing between them and their customers. If the strongest operators keep leaving once they reach scale, Uber's aggregation strategy faces a structural problem: a massive marketplace may be the best way to launch a robot service, but not necessarily the best way to run one.
- #uber
- #waymo
- #serve-robotics
- #autonomous-vehicles
- #robotaxis