Europe's First Robotaxi Runs on Chinese AI

Europe's first commercial robotaxi service is launching in Zagreb, Croatia — and the self-driving brain inside it was built in Guangzhou, China. Uber, Chinese autonomous driving company Pony.ai, and Croatian startup Verne announced the partnership on March 26, with on-road testing already underway using Chinese-manufactured vehicles running Chinese AI.

The deal marks the first time a Chinese autonomous driving system will power commercial rides on European streets. It also quietly answers a question the continent has been avoiding: in the global race to put AI behind the wheel, where does Europe actually stand?

The Zagreb Deal

The structure is simple. Pony.ai provides the autonomous driving technology — its Gen-7 system, installed on an Arcfox Alpha T5 built by Chinese automaker BAIC. Verne owns and operates the fleet. Uber plugs it into its ride-hailing network so anyone with the app can hail one.

Uber is also making an undisclosed strategic investment in Verne, signalling this isn't a one-city experiment. The three companies say they plan to scale to "thousands of vehicles across Europe" in the coming years, though no specific launch date or expansion cities have been named.

What makes this unusual isn't the robotaxi itself — it's the supply chain. The AI is Chinese. The vehicle is Chinese. The fleet operator is Croatian. And the distribution layer is American. Europe, the continent where the service actually operates, contributed none of the core technology.

The Man Behind Verne

Verne's origin story is almost too ironic to script. It was created by Mate Rimac, the Croatian engineer who built the Nevera — a $2.2 million electric hypercar that accelerates faster than a Formula 1 car. Rimac started the robotaxi project in 2019 under the name "Project 3 Mobility" inside Rimac Group, his growing empire that also includes Bugatti Rimac.

Rimac's logic was counterintuitive for a supercar maker: he believed autonomous vehicles would eventually make human-driven cars obsolete for daily transport. Rather than compete in a dying market, he wanted to build the replacement.

Verne launched independently in July 2024 with €100 million in funding. It's building purpose-built two-seater electric robotaxis at a new factory in Lučko, Croatia, expected to begin production later this year. Sixty verification prototypes have already been built and tested. But the company's own vehicles aren't ready for commercial service yet — which is why Zagreb's first robotaxis will be the Chinese-made Arcfox Alpha T5s instead.

China's Quiet Lead

The Zagreb announcement landed on the same day Pony.ai released its Q4 2025 earnings, and the numbers tell a story that most European and American readers haven't been following.

Pony.ai's robotaxi revenue grew 160% year-over-year in Q4, reaching $6.7 million. Fare-charging revenue — meaning money from actual passengers paying for rides — surged over 500%. The company reported consecutive quarters of positive unit economics, meaning each robotaxi ride is now generating more revenue than it costs to operate.

The fleet is scaling fast. Pony.ai currently operates fully driverless commercial services across Beijing, Shanghai, Guangzhou, and Shenzhen. By the end of 2026, it plans to more than double its fleet to over 3,000 vehicles across more than 20 cities globally. Zagreb will be its first European market.

Here's the number that matters most: according to analysis by the Special Competitive Studies Project, Chinese autonomous vehicle operators have collectively logged roughly 149 million autonomous miles, compared to around 106 million for US operators. China didn't just catch up in the self-driving race. It pulled ahead — and most people in the West have no idea.

Why Europe Can't Build Its Own

Europe has some of the world's strongest AI research institutions and a long tradition of automotive engineering. It has BMW, Mercedes, Volkswagen, Renault, Volvo. What it doesn't have is a single company capable of deploying a commercial robotaxi service with its own technology.

Waymo, the clear leader in the US, raised $16 billion in February 2026 and is expanding to new American cities including New Orleans, Minneapolis, and Tampa. In China, Pony.ai and Baidu's Apollo Go are already running thousands of driverless rides daily. Europe has neither equivalent.

The gap isn't really about engineering talent. It's about three things Europe lacks in combination: massive real-world driving datasets (China's cities generate them at scale), tolerance for aggressive urban testing (European regulators move slowly), and the kind of patient capital that lets companies burn cash for a decade before turning profitable. Pony.ai's total 2025 revenue was $90 million against a non-GAAP net loss of $49 million per quarter. That's the kind of investment horizon European venture capital rarely supports for hardware-heavy autonomous systems.

Croatia, of all places, may have found a workaround. Rather than trying to build every layer of the stack domestically, Rimac's strategy treats autonomous driving like a commodity input — buy the brain from whoever makes the best one, then compete on fleet operations, vehicle design, and urban deployment. It's the Android model applied to robotaxis.

The Geopolitical Wrinkle

This deal will make regulators uncomfortable. Chinese AI systems processing European streets in real time raises questions about data sovereignty, national security, and technology dependency that Europe has been debating in theory but never had to face in practice.

The EU has spent years crafting the AI Act, the world's most comprehensive AI regulation framework. But the AI Act was designed primarily to govern AI systems developed by companies operating within or selling to Europe. A Chinese autonomous driving system deployed by a Croatian company through an American app doesn't fit neatly into any existing regulatory box.

Pony.ai's Gen-7 system processes camera, LiDAR, and radar data continuously as it navigates city streets. That data includes pedestrians' faces, building locations, traffic patterns, and infrastructure details. Where it's stored, who has access, and what happens to it are questions that will need answers before this service expands beyond Zagreb.

For now, the three partners haven't addressed these questions publicly. The focus is on getting the service running.

What Happens Next

The real test isn't Zagreb. It's whether this model — Chinese AI, local operations, American distribution — becomes the template for autonomous mobility in markets that can't build their own.

Nvidia declared at its GTC conference last week that the "ChatGPT moment for physical AI" has arrived. Its new Alpamayo 1.5 model brings reasoning capabilities to autonomous vehicles, and its GR00T N1.7 humanoid robot model is now "commercially viable for real-world deployment." The tools to build autonomous systems are proliferating. The question is who deploys them first in each market.

In the US, that's Waymo. In China, it's Pony.ai and Apollo Go. In Europe, as of this week, the answer is: a Croatian hypercar maker, a Chinese AI company, and an American ride-hailing app — working together because none of them could do it alone.

The self-driving future everyone's been waiting for is arriving. It's just arriving with a different accent than anyone expected.