On September 2, Uber’s newsroom quietly published two stories within hours of each other.
One announced that roughly 3,300 people were losing their jobs.
The other announced that Londoners could now hail a Ford Mustang Mach-E with nobody meaningfully driving it. Read separately, they’re unrelated news items.
Read together, they’re the same story.
CEO Dara Khosrowshahi’s internal memo, posted publicly the same morning, framed the cuts as an org-chart exercise: flattening management layers, eliminating micro-teams of one or two people, folding three separate delivery divisions into a single structure, and ending remote work for all but about 1% of staff.
Khosrowshahi wrote that Uber had accumulated structures that made sense when businesses were smaller but no longer serve the company well at its current scale. Fair enough. Most companies eventually say some version of that.
What the memo didn’t spell out in dollars is where the savings actually go. Khosrowshahi gave a hint, writing that the changes would free up capital to build what he called the autonomous future.
A few hours later, Uber and UK startup Wayve announced that supervised autonomous rides had gone live across London, the first time the public could hail a self-driving car anywhere in Britain. The vehicle underneath that launch: a Ford Mustang Mach-E, retrofitted with Wayve’s AI Driver system.
Here’s the detail worth sitting with. Wayve didn’t build its first UK robotaxi on a Mustang because it’s a Mustang. It built it on a Mach-E because the Mach-E is an EV.
Electric vehicles already steer, brake, and accelerate through electronic controls rather than hydraulic or mechanical linkages, because that’s how regenerative braking and one-pedal driving work in the first place. Bolting a computer onto a system that’s already wire-controlled is a dramatically shorter engineering path than retrofitting a car that still relies on a hydraulic brake booster or a physical throttle cable.
That’s a big reason the entire autonomous vehicle industry has clustered around electric platforms. It’s less about optics and more about which cars are already halfway to being robots.
The second detail is the word supervised. Every Wayve vehicle running on Uber’s London network still carries a trained, TfL-licensed private hire driver, there specifically to intervene if the AI Driver gets something wrong. London’s first public robotaxi still has a professional driver in it.
That’s not a marketing footnote. Full driverless operation clears a much higher regulatory bar than supervised operation, and insurers price liability very differently depending on who, or what, is legally in control at the moment of a crash.
Amazon’s Zoox is still working through that exact question with regulators on its steering-wheel-free robotaxi, and Waymo has already learned the hard way what happens when the software misjudges something as basic as standing water.
Uber, for what it’s worth, already runs one of the largest driver-monitoring operations in the country, right down to pitching rideshare dashcam footage to outside surveillance companies. Swapping a monitored human for a supervised AI doesn’t eliminate that monitoring problem. It just relocates who, or what, is being watched inside the car.
So what did Uber actually trim 3,300 jobs to pay for? Not the vehicles, Uber doesn’t build them. Not the driving software, Wayve built that. Not even the safety driver in the seat, who isn’t an Uber employee.
What Uber cut was coordination: the managers, the go-between teams, and the layers of approval required to run a company that grew by orders of magnitude over the last five years. That’s the actual trade being made. Uber is betting its next phase needs fewer people managing people and more capital flowing into outside partnerships, Wayve in London, Waymo and others elsewhere, that scale with rides instead of headcount.
Detroit already ran this experiment and walked away from it. Ford shut down its in-house self-driving unit, Argo AI, back in 2022. GM spent billions building Cruise before winding down its robotaxi ambitions.
Both companies reached the same conclusion Uber’s org chart just acted on: building the driving software is a technology company’s job, not a car company’s, and definitely not a rideshare app’s regional operations team’s job either. What’s left for the automaker is the hardware, and increasingly that hardware just needs to be a good EV platform that somebody else’s sensors can be bolted onto. Ford, for now, seems content to just sell the car.
There’s an irony worth noticing here. The Mustang name was built on the idea of a driver in control, a car sold on how it felt to steer, shift, and occasionally step out of line. That badge is now pulling supervised robotaxi duty on some of the most complicated streets in Britain.
The car’s identity didn’t change so much as Uber’s definition of driver did. A rideshare platform never needed its drivers to be employees. Now it doesn’t even need them to be human. It just needs someone, or something, licensed to occupy the seat.
The 3,300 number will fade by next week’s news cycle. What shouldn’t: Uber didn’t shrink itself to save money on drivers. It shrank the layers of itself that coordinate humans, while carefully, deliberately, leaving the layer that still requires a human in the actual car untouched, for now. That gap, between an org chart that got leaner and a driver’s seat that didn’t, is where the next five years of this industry gets decided.