Every layoff comes wrapped in the same corporate vocabulary: flattening, simplifying, refocusing. Uber’s September 2 memo, cutting roughly 3,300 corporate jobs, checked every box. CEO Dara Khosrowshahi framed it as removing bureaucracy to fund what he called “build the autonomous future,” and on paper the numbers back him up: Uber cut its micro-teams of one or two direct reports by nearly half and reduced the number of employees sitting seven or more layers below the CEO by a fifth.
That’s the story Uber told employees. Here’s the one it didn’t: the autonomous vehicle company most responsible for making that future look real to millions of riders, Waymo, has spent this year quietly disentangling itself from Uber’s app.
Start with why the layoffs don’t add up as a cost-cutting story. Uber isn’t shrinking because business is bad. It generated $52.02 billion in revenue in 2025, up 18% from the year before, while gross bookings grew 19% to $193.45 billion. In the most recent quarter Uber has reported, revenue rose 12% year over year to $14.2 billion, trips climbed 18%, and operating income jumped 30%. Companies don’t usually cut a tenth of their corporate workforce in the middle of numbers like that unless they’re redirecting the savings toward something specific.
Uber said exactly what that something is: autonomous vehicles. But look at who is missing from the victory lap. Reporting throughout the summer confirmed that Waymo and Uber ended their robotaxi partnership in Phoenix in June, closing out nearly three years of running Waymo vehicles through Uber’s app in that city. Waymo has since been weighing a broader exit from Uber’s platform entirely, with plans to run its own standalone service, outside any Uber app, in Austin and Atlanta by January 2028. Coverage of the relationship elsewhere on Uber’s app describes a contract that runs only through May 2028.
That’s not a renewal. That’s a wind-down with a date on it.
Uber doesn’t talk about Waymo much these days. It talks instead about the roughly 30 autonomous vehicle partnerships it has assembled worldwide, including a supervised, driverless Mustang Mach-E now running through Wayve’s technology in London, a robotaxi service with WeRide in Madrid, and deals touching Baidu, Pony.ai, Volkswagen’s MOIA, Avride, and a Nuro-built, Lucid-powered robotaxi program still in development. Thirty logos on a partnerships slide can look like diversification. They can also look like what a company does when its best partner is heading for the exit, and there’s no replacing that kind of leverage with volume alone.
Uber has been in exactly this position before, and the last time it ended badly enough to shape everything that came after. In March 2018, one of Uber’s own self-driving test vehicles, developed by its in-house Advanced Technologies Group, struck and killed 49-year-old Elaine Herzberg as she walked her bicycle across a road in Tempe, Arizona. It remains the first recorded pedestrian death involving a self-driving vehicle. Uber pulled its self-driving tests off Arizona roads within days, and by December 2020 it sold the unit outright, taking a minority stake in Aurora Innovation rather than keep building the technology itself.
That sale set the template Uber still follows. Uber doesn’t build self-driving cars anymore. It rents shelf space to the companies that do, and collects a cut every time someone hails one through its app.
Here’s the part that doesn’t show up in an org chart. Uber’s business, since 2010, has been solving a matching problem: connecting scattered human drivers with scattered riders, and handling the recruiting, background checks, insurance coordination, and payment processing that made that matching work at scale. A robotaxi doesn’t need any of that. It doesn’t get recruited. It doesn’t get background-checked. Its insurance is a commercial fleet policy written directly with the operator, not a patchwork of state rideshare endorsements layered onto a driver’s personal coverage. And once an operator like Waymo has enough of its own riders in a city, permitted directly through agencies like California’s Public Utilities Commission rather than riding on Uber’s transportation network company license, Uber’s percentage of every fare stops looking like a service fee and starts looking like rent for storefront space the tenant no longer needs.
That’s the leverage problem sitting inside every one of these friendly-sounding partnerships. The bigger and more self-sufficient an autonomous vehicle company becomes, the less it needs Uber’s app, Uber’s brand recognition, or Uber’s decade of demand generation. Waymo, backed by Alphabet’s balance sheet, is the partner with the least reason of all of them to keep paying that rent.
Seen that way, Uber’s flattened org chart stops looking like a cost-cutting story and starts looking like a hedge. A company anchored to one deeply integrated AV partner can afford patient, specialized teams built around that relationship. A company juggling thirty separate AV relationships, controlling none of the underlying technology, needs something closer to a deal desk: fewer coordination layers, faster sign-offs, people who can move capital between bets as partners like Waymo drift away and newer ones scale up. Uber didn’t flatten its structure to build cars. It flattened its structure to place bets faster, because it can no longer count on any single bet lasting.
Detroit already lived through a version of this math. Ford shut down Argo AI in 2022. GM spent billions on Cruise before winding down its robotaxi ambitions entirely. Amazon’s Zoox is still negotiating that regulatory territory for its steering-wheel-free vehicle, the same territory Uber would rather leave to somebody else’s legal department. Waymo, for its part, has been busy building its own next-generation robotaxi, the Ojai, at a facility in Arizona, a sign of a company preparing to stand on its own rather than lean on a partner’s app.
None of this means Uber’s robotaxi bet is doomed. Spreading risk across thirty partners is a reasonable hedge in a market where nobody, including Waymo, has fully solved the economics of driverless rides. But the layoffs aren’t evidence that Uber is winning the robotaxi race. They’re evidence that Uber no longer controls it. Uber spent six years and one dead pedestrian learning it couldn’t build the robot itself. The lesson it drew wasn’t to stop trying. It was to stop owning, and start renting shelf space instead, from whichever robot shows up.
The 3,300 jobs will be a footnote by next quarter’s earnings call. What won’t be: whether Uber’s most important partner still needs Uber’s app once the rent comes due in 2028.