Robotaxis can’t be a solo act

That might seem to be stating the obvious. And yet, when Elon Musk announced in early April that Tesla was planning to unveil its own robotaxi, the first instinct of investors was to bail on a company that already had a platform of 150 million people using it at least once a month for rides and food delivery. Uber lost nearly one-quarter of its market value over the following four months.

The stock has since clawed back much of that ground, due in part to a smattering of deals Uber has signed with such robotaxi providers as Waymo and Cruise. Those deals help make the case that Uber’s massive platform of riders has value in a world in which expensive, driverless taxis need a steady base of passengers to make their economics work.

The deal with Waymo was particularly notable, as the company already runs its own robotaxi service in San Francisco, Los Angeles and Phoenix. Under that agreement, Uber will be the exclusive platform for booking rides in Waymo’s cars in Atlanta and Austin, Texas, when the service launches in those two markets next year. While Uber once aspired to build its own robotaxis, it has since decided to play to its strengths. “Generally, we want to build a marketplace, and we want to stay as pure-play a marketplace as we can," Uber Chief Executive Dara Khosrowshahi said at a Goldman Sachs investment conference in San Francisco last month.

Still, many investors seem to be betting on a Tesla win—at Uber’s expense. Tesla’s stock price has surged 43% over the past six months, ahead of the company’s robotaxi-unveiling event scheduled for Oct. 10. Uber’s shares have dropped 5% during the same time.

Some history might be at work. Uber did once consider Tesla a major threat, even spelling out a Tesla autonomous vehicle as a risk factor in its initial public offering papers in early 2019. And Tesla so far has been signaling that it wants to go its own way with robotaxis. On earnings calls and other forums, Musk has touted plans for a fleet of cars without pedals or steering wheels operating through the same network the company uses to direct drivers of its EVs to charging stations.

“This would just be the Tesla network," he said on Tesla’s earnings call in July, in response to an analyst’s question about how the company’s robotaxi service would work. “You just literally open the Tesla app and summon a car, and we send a car to pick you up and take you somewhere."

A lot like Uber, in other words. But replicating a ride-hailing network with a 14-year head start will be no easy feat, especially considering the scale Uber has achieved. The company’s mobility segment alone now handles more than $34 billion in gross bookings in the U.S. each year, according to consensus estimates by Visible Alpha.

While robotaxis spare the expense of paying a human driver, they are far from cheap. Analysts from Bernstein estimate that Waymo’s driverless cars cost between $150,000 and $200,000 a piece, factoring in the cost of the vehicle itself and the sensors and computing power necessary to run them. And those cars are nearly always operating and—thus—generating costs for their operators, while fleets of cars with human drivers can be scaled up and down based on demand and don’t generate costs when not carrying passengers.

The high fixed costs of a fully autonomous vehicle network “creates economic questions around upfront cost and utilization rates during non-peak hours," Bernstein’s analysts wrote in a report last month.

A wide network of riders is vital to the economics of such a business, both to amortize the large upfront costs and minimize so-called deadhead miles—referring to the distance and time between revenue-generating trips. In a June report, Brian Nowak of Morgan Stanley estimated that building a network of autonomous cars that would rival the size of Uber’s U.S. fleet would cost at least $25 billion—and could reach as high as $60 billion—depending on the technology used in the fleet.

Consider Waymo—the largest active robotaxi network, which surpassed the milestone of 100,000 paid weekly trips in August. The company is part of Google parent Alphabet’s Other Bets reporting segment, which reported $4.2 billion in operating losses for the 12-month period ended in June. Shweta Khajuria of Wolfe Research estimated that Waymo accounts for about half that loss. “Our analysis suggests it will require ~10 years for Waymo to reach payback on invested capital," she wrote in a report last week.

Is Tesla ready to stomach that? Musk is characteristically aiming high. “This will be one for the history books," he said of the robotaxi event on his X platform on Sept. 25.

But the voluble billionaire’s dreams don’t always line up with reality. And getting a workable robotaxi fleet operating on public roads is a bigger hurdle than just perfecting the technology of the cars themselves.

UBS analyst Joseph Spak said Tesla currently lacks the proper permits from the state of California to test its vehicles without a driver present, even as the company is set to unveil its cars at a Los Angeles event. “We believe wide-scale Tesla robotaxi deployment is unlikely within the coming years," Spak wrote in a report last month. Uber and its robotaxi partners might still have plenty of time on their side.

Write to Dan Gallagher at dan.gallagher@wsj.com