When people talk about driverless taxis, it is often Waymo that gets the shoutout. Few mention Pony AI.
CEO James Peng is hellbent on flipping this narrative.
“In two years, we’ll be bigger than Waymo on fleet size,” he tells IZON’s Executive Editor Ben Hubbard.
“At the start of 2025 we had 270 robotaxis, today we’re at 2,000 and by the end of this year we’ll have 3,500, so we’re on a very strong growth trajectory.”
Pony AI Inc is already running Level 4 commercial robotaxi services in four Tier-1 Chinese cities: Beijing, Guangzhou, Shenzhen and Shanghai.
“Those are fully driverless with paid customers, and we have 24/7 services in some cities,” says Peng.
Outside China, the company has quietly been extending its footprint across a widening map of early autonomy markets, including Singapore, Doha, Dubai, Seoul, Luxembourg and Zagreb.
At the core of its strategy is what Peng calls the “Virtual Driver”: a software-defined autonomous driving system designed to sit inside multiple vehicle platforms and power different mobility use cases, from robotaxis to logistics and low-speed urban services.
In that framing, Pony is attempting to become the essential brain layer in any autonomous rollout.
Pony’s edge
Autonomous driving is often discussed as a hardware race: who has the best sensor suite, the safest car, the most refined perception stack, but Peng sees it differently.
“Robotaxis are such a complex system,” he says. “Technology is just one part of it. We’ve spent the last 10 years getting the technology right, with more than 85 million kilometres of autonomous driving on public roads and a very strong safety performance.”
But the next constraint, he argues, is not capability but unit economics.
On cost, Pony has been aggressively working to compress the price of deployment. The company expects its 2027 robotaxi model in China to fall below $33,000 per vehicle.
That shift matters because it changes the shape of scalability. The lower the vehicle cost, the less capital-intensive it becomes to expand dense urban fleets, and the faster utilisation can compound.
Yet even with improving cost curves and accumulated road experience, Peng argues the industry is still waiting on a final unlock.
“I think we are waiting for one thing to get ready: regulation,” he says. “Once regulation is there, the scalability path will be exponential.”
Europe: fragmented markets, modular strategy
If China represents scaled deployment, Europe is being treated as a structured proving ground.
The company’s European strategy is deliberately non-vertical. Instead of attempting a full-stack rollout, Pony is positioning itself as a modular technology provider embedded within local ecosystems.
“The European market, first and foremost, is a very interesting market to us,” Peng says. “It has a population of 500 million people and is economically very strong. But we are definitely not going to come to Europe alone.”
That approach reflects a clear division of labour across the mobility stack. At a high level, Peng likens it to a taxi business with four components: user acquisition, vehicle supply, the driver system, and back-end operations including charging, cleaning and maintenance.
“We are providing the Virtual Driver,” he says. “For everything else, we will have partners. It is going to be an ecosystem, not a single partner.”
In practice, that means working with ride-hailing platforms such as Uber and Bolt, alongside automakers and local fleet operators depending on the city.
Rather than standardising a single deployment model, Pony is adapting to Europe’s fragmented regulatory and operational landscape on a city-by-city basis.
Early rollouts prioritise dense, tightly defined urban zones where utilisation can be maximised and operational data can be gathered quickly. In Shenzhen, Peng says Pony is already achieving around 25 rides per vehicle per day, with an average wait time of five minutes, a benchmark it is now attempting to replicate in Europe.
Croatia and Luxembourg
The first signals of that strategy are already visible in Croatia and Luxembourg.
In Zagreb, Pony works with Uber on the demand side, while Verne manages local operations. Around 15 robotaxis are currently operating across central routes and airport corridors, still with safety operators onboard.
“Our internal data suggests it is safe to remove the safety driver, and we will be testing that very soon,” Peng says.
In Luxembourg, Pony’s partnership with Bolt and Stellantis is being used less as a commercial deployment and more as a regulatory and systems validation environment, a place to test safety performance, operational readiness and policy alignment.
“With Luxembourg, you can make an appointment with a minister very quickly,” Peng notes. “For us, that is great because it means we can grow fast while working with policymakers to demonstrate to the public that the technology is safe, reliable and convenient.”
The real bottlenecks
Across Europe, Peng sees two dominant constraints shaping the pace of deployment: regulation and partnerships.
Autonomous vehicle frameworks remain fragmented across borders, with differing approval processes, testing permissions and data requirements from country to country. That fragmentation forces companies like Pony into a partnership-heavy model by default.
“We must have strong partnerships with automakers, ride-hailing platforms and service providers to scale,” he says.
Recent moves toward cross-border coordination between European states are, in his view, an early sign of convergence but still far from a unified system.
At the same time, he notes that more cities are now actively queueing for autonomous vehicle trials, suggesting demand is beginning to outpace available deployments.
What next
Peng confirms that another European city is already in preparation, though details remain undisclosed.
“There will be something this year,” he says. “There is a lot happening.”
Globally, Pony is targeting operations in 20 cities by the end of the year.
The goal is as much narrative as it is scale: to move the conversation away from Waymo as the default reference point for autonomy, and toward a more contested global hierarchy.
