Uber needs to find way to get self-driving unit back on track
Without a fleet of autonomous vehicles Uber risks losing its grip on the ride-hailing market
San Francisco — With deep pockets, some of the best brains in robotics and its hallmark hustle, Uber Technologies was once a serious contender in the race to build self-driving cars and revolutionise global transportation.
Today, the effort is lagging far behind rivals, and investors are getting antsy about the rationale for continuing a long-term science project that is not helping Uber become profitable or safeguarding its future.
In recent months, the company’s two largest shareholders, SoftBank and the venture firm Benchmark, have privately encouraged CEO Dara Khosrowshahi to find more investors for the division, which is expected to exhaust its funds by the end of 2021, and re-evaluate its strategy, according to people familiar with the situation.
Khosrowshahi is exploring various options. One would entail Uber accelerating the project by open-sourcing it to coders around the world, said the people, who requested anonymity to speak about a private matter. Or the company could continue building a proprietary car with money from existing partners Toyota and Denso, and possibly new ones, they said.
It is an existential decision, because without a fleet of autonomous vehicles, Uber risks losing its grip on the ride-hailing market to any number of companies, from Alphabet’s well-established Waymo unit to upstarts such as Amazon.com’s Zoox, which recently got the green light to test autonomous vehicles on Californian roads without a safety driver.
PitchBook mobility analyst Asad Hussain describes a nightmare scenario where Waymo develops its own network, Amazon gives free rides to Prime subscribers, and Uber becomes irrelevant. “Self-driving is the next smartphone,” he said. “You don’t want to get left out.”
When Khosrowshahi assumed Uber’s helm from ousted founder Travis Kalanick in 2017, he inherited a mess of cultural, strategic and financial problems. Khosrowshahi set to fixing the most pressing ones first, repairing Uber’s reputation and values and offloading losing bets in overseas food delivery and ride networks; the self-driving group went on the back burner.
Work did not stop, but the division was badly demoralised by a pair of events: the conviction and imprisonment of its former chief, Anthony Levandowski, for stealing intellectual property from Waymo, and the 2018 death of a pedestrian struck by an Uber vehicle in Phoenix, Arizona, while the human driver was watching The Voice on her cellphone.
Meanwhile, Uber’s rivals were making progress. Waymo cars have logged hundreds of thousands of kilometres in more than two dozen cities; Lyft recently surpassed 100,000 rides in Las Vegas using vehicles supplied by Motional. General Motors’s Cruise has completed 50,000 grocery robodeliveries during the pandemic alone. By contrast, Uber is operating fewer than 10 cars a week on public roads in Pittsburgh and Washington.
Last year Khosrowshahi bought himself some time by turning the autonomous driving division, called the Advanced Technologies Group (ATG), into an $8bn subsidiary with its own set of investors, including SoftBank, Denso and Toyota, to help fund an operation that is consuming about $500m a year. The move got the unit off Uber’s balance sheet and helped silence critics. But it did not deal with the more fundamental question of strategy and whether Uber should be in the business of building its own autonomous driving technology.
Investors and executives including former Uber chief technology officer Thuan Pham have said that open-sourcing ATG’s self-driving technology would speed up development and establish Uber as a leader, similar to how Google’s open-source Android mobile operating system created an alternative to Apple. Going this route would require Uber to release the source code and grant rights to everyone interested in using, changing, studying and distributing it however they choose. Proponents say this would let Uber remain on the cutting edge while progressing faster because the work would be done collaboratively by coders around the globe.
“There’s no question in my mind that [open source] is the perfect solution for a highly complex problem,” said Benchmark’s Bill Gurley, who added that the move would boost Uber’s share price by as much as 20%. “I would sacrifice whatever value there might be to lead an open-source movement to get the company to a better place faster. I’m highly confident the market would respond to that.”
Khosrowshahi earlier in 2020 directed ATG chief Eric Meyhofer to explore the idea, according to people with knowledge of the situation, who said the conversation quickly became “contentious.” Meyhofer confirmed the conversation and, though he disagrees with that characterisation, he does believe the idea of open-sourcing ATG’s technology is “too complicated.”
“Open sourcing opens a can of worms,” he said. “We don’t think it makes sense for our strategy.”
Amid the infighting, Uber’s partnerships with other self-driving companies have withered. The original idea was to have a backup in case Uber’s proprietary effort was late or did not come to fruition. That is why Uber struck a deal with Daimler back in 2017 to put Daimler’s own self-driving cars on Uber’s ride network.
But all new work with the German-US carmaker has ceased, according to people with knowledge of the matter. Uber tentatively agreed to work with Motional, with tests planned for later in 2020, but several ATG employees advocating for the partnership have left, putting the deal in purgatory, the people said. Daimler did not respond to requests for comment.
Meyhofer said ATG “remains interested in and committed to pursuing partnership opportunities that serve our mission of bringing safe, reliable self-driving technologies to the Uber platform”. He declined to comment on details of the existing Motional arrangement, but said no “deal” has been signed and nothing is set to be tested in 2020. Motional said it would be “premature” to comment on the status of conversations with Uber and other ride-share companies.
Work on Uber’s own autonomous technology is not going well either. ATG engineers are in the midst of replacing the software platform the company has been licensing from Carnegie Melon University since 2015. Originally, the project was going to take six months, but that deadline has been pushed out to more than 18 months, according to people familiar with the work. Earlier in 2020, cars running on the new platform occasionally lost computing power, going just a few hundred yards before stopping, the people said. Then, just as some of those initial kinks were being worked out, Covid-19 hit.
ATG suspended all testing apart from simulation work for a “brief period”, until it could ensure employees would be able to safely resume work, according to its financial report for the period ending June 30.
Today, five years after Uber launched its self-driving effort, the company operates vehicles in the nation’s capital but only with the autonomous system turned off. In Pittsburgh, meanwhile, the autonomous system is engaged only “as needed,” with a human driving the rest of the time. Khosrowshahi has little in the way of milestones to show potential investors.
Would you like to comment on this article or view other readers' comments?
Register (it’s quick and free) or sign in now.
Please read our Comment Policy before commenting.