20 Aug 2026, Thu

Pony.ai Promised Europe More Robotaxis Than It’s Ever Built. Who Owns the Ones That Don’t Exist Yet?

Uber-branded autonomous robotaxi test vehicle with roof-mounted sensor array

Somewhere inside a routine filing with the U.S. Securities and Exchange Commission sits a sentence more revealing than anything in last week’s press release: vehicle funding and ownership can shift from partner to partner depending on the market. That’s corporate language for something simple. When one of these robotaxis eventually collides with a cyclist on a Zagreb street, or a pedestrian in whatever city gets named next, the company whose logo sits on the app may not be the company whose insurance actually pays out.

That’s the real story tucked inside the announcement Uber and Chinese autonomous driving developer Pony.ai made on August 14: an expanded partnership to deploy more than 2,000 robotaxis across Europe, growing out from an existing service in Zagreb into four still-unnamed cities, with the Middle East to follow. Neither company said which cities, or when the cars actually start picking up paying riders.

Here’s what the deal actually is, according to Pony.ai’s own filing. Pony.ai supplies the Level 4 self-driving software and the operational know-how it built running fleets across China. Uber supplies the app: booking, payment, customer service, the interface riders already trust. A third party, chosen market by market, runs the cars day to day. In Zagreb, that’s Verne, a company backed by Croatian hypercar maker Rimac. Uber’s global head of autonomous mobility, Sarfraz Maredia, described the arrangement as a shift “from individual launches to repeatable commercial scale.”

This isn’t really a story about robotaxis arriving in Europe. It’s a story about who owns the metal, and who inherits the liability when the metal makes a mistake.

Start with the fleet math, because it doesn’t add up the way a press release wants it to. According to Pony.ai’s own SEC filings, its global robotaxi fleet passed 1,400 vehicles as of this past March. By early June, it was running roughly 1,700, against a company-stated target of 3,500 vehicles by the end of the year. Now read the European pledge again: more than 2,000 robotaxis, in a market where Pony.ai currently operates commercially in exactly one city. Wait, really? Yes. Europe alone would absorb more robotaxis than Pony.ai has ever put on a road anywhere on Earth, and more than half of everything the company hopes to have running globally by New Year’s Eve.

That gap between promise and inventory is exactly why the deal is structured the way it is. Local partners like Verne carry the capital cost of the vehicles. Pony.ai isn’t necessarily buying fleets of cars any more than Uber is. It’s licensing software and operational playbooks, the same way a franchisor licenses a name and a fryer recipe rather than building every restaurant itself. Fast-food franchising conquered the world by separating the brand from the fryer. Robotaxis are now testing whether the same trick works when the fryer can run a red light.

That structure creates a second problem nobody in the press release wants to talk about: liability. A crash lawsuit involving a robotaxi already has to sort out, in the United States, whether fault sits with the sensor-and-software maker, the fleet operator maintaining the car, or the platform that dispatched it. Now stretch that three-way argument across national borders, five cities, and at least one local fleet partner that hasn’t even been named yet in four of them. Multiply that ambiguity by 2,000 vehicles and you have years of contract law working itself out in courtrooms nobody’s picked yet.

Here’s the part that should surprise anyone who assumes Europe is the cautious regulator and America is the free-for-all: it’s backward on paper. Since 2021, the National Highway Traffic Safety Administration has run a Standing General Order requiring makers and operators of automated driving systems to report to the agency any crash where the automation was engaged within 30 seconds of impact and the damage cleared a set threshold, a standardized, nationwide safety data feed most drivers have never heard of. Europe has no direct equivalent. Oversight is split among national regulators, which is precisely why European safety researchers have warned that commercial robotaxi rollouts are outrunning the bloc’s capacity to investigate them. The “move fast” country built the paperwork first.

So why now? Because Pony.ai has fewer places left to go. Washington has spent the past year working to lock Chinese-made and Chinese-connected vehicles out of the American market, robotaxis included, which leaves Europe as one of the only places a Chinese autonomous vehicle developer can prove, at real scale, that its technology works outside its home market. The timing is not subtle, either: Pony.ai reports second-quarter earnings this week, and the fleet-count line in that report is the number investors will actually trust. Not the one in a press release.

Uber, notably, has almost nothing riding on Pony.ai specifically succeeding. It has already committed more than $10 billion across more than 30 autonomous vehicle partnerships, spanning Waymo, Rivian, Nuro, Wayve, and Nvidia’s software stack. It has made this exact kind of multi-city, no-fixed-date pledge before, with Stellantis and Wayve, and quietly let the timeline slide. Pony.ai just tied a meaningful share of its own credibility to a partner that has roughly 29 other ways to fill the same seats.

The pattern extends further than this one deal. Waymo has spent months stockpiling hundreds of Chinese-built robotaxi hardware in Arizona, positioning inventory ahead of regulatory deadlines that haven’t closed yet. Every company in this industry, American or Chinese, is racing to lock in supply and market position before the rulebook finishes being written. That’s not a coincidence. It’s the strategy.

What should stick with readers isn’t the 2,000 number, and it isn’t the five cities nobody’s named yet. It’s this: every unresolved detail in this deal, the cities, the timeline, the name on the vehicle title, is a decision somebody is deliberately leaving until after the headlines move on. That’s not sloppiness. That’s the plan. Watch the fleet count when Pony.ai reports earnings this week, not the ribbon-cutting whenever it finally happens.

By John Lloyd

John Lloyd writes for The Auto Wire, where he covers the more entertaining corners of the car world—celebrity rides, motorsports drama, and whatever automotive thing happens to be blowing up online that week. He's drawn to where cars meet culture. One day that's breaking down why some celebrity dropped a fortune on a hypercar; the next it's explaining why a particular model is suddenly all over everyone's feed. He likes handing readers the context behind the headline, usually with a little attitude. The way John sees it, cars aren't just transportation—they're status symbols, money pits, lifelong obsessions, and occasionally pure chaos, and that's exactly the stuff worth writing about.

Join the conversation

No comments yet — be the first to share your take.

Your email address will not be published. Required fields are marked *