The project is slated to get underway soon in the Croatian capital of Zagreb soon, with initial development work already happening, Pony.ai announced Thursday (March 26).
“The three companies plan to collaborate on the deployment of a commercial robotaxi service, combining Pony.ai’s autonomous driving system, Uber’s global mobility platform, and Verne’s service ecosystem and operational framework,” the announcement said.
“Under this model, Pony.ai will provide its autonomous driving solution; Verne will act as fleet owner and service operator; and Uber will integrate the service into its global ride-hailing network, complementing Verne’s own customer-facing platform,” the announcement added.
Also Thursday, Pony.ai announced its quarterly earnings, along with plans to deploy robotaxis in more than 20 cities around the world. The company’s robotaxi fleet has now surpassed 1,400 vehicles, and has extended its operations into Croatia as well as the Chinese cities of Hangzhou and Changsha, the announcement said.
The company said its robotaxi revenues rose by 160% year-over-year during the quarter, with fare-charging revenues jumping by more than 500%.
“The robotaxi is the first true application of physical AI, with its core differentiator lying in driving capability, an area where our competitive edge remains unparalleled across the industry,” said Dr. Tiancheng Lou, Pony.ai‘s chief technology officer.
Physical AI, as covered here recently, refers to systems that go beyond generating content into operating in real environments: robots, autonomous machines and the foundational models that instruct them how to behave.
“The category has been building momentum for years, but 2026 marks a credible inflection point,” PYMNTS wrote. “At CES in January, Nvidia CEO Jensen Huang declared that the ChatGPT moment for robotics had arrived, suggesting the combination of AI models and computing infrastructure could soon unlock large-scale commercial adoption.”
The report also touched on the theory that machines that can perceive, reason and function in the physical world may be on a more direct path “than software running in a data center” to reaching AGI (artificial general intelligence) or a form of artificial intelligence (AI) that can perform at/above the level of humans.
“If that thesis is right, the physical AI buildout now underway is not just a manufacturing story. It is the early infrastructure of something considerably larger,” PYMNTS added.