The autonomous driving industry is undergoing a paradigm shift in 2025. After a decade of lab experiments and conceptual demonstrations, Robotaxi businesses are now being tested by real-world financial metrics.
Pony AI's co-founder and CFO Wang Haojun revealed in an interview that their seventh-generation Robotaxi fleet in Guangzhou has achieved positive unit economics (UE), marking a critical transition from R&D to commercial viability. This milestone demonstrates Pony AI's ability to establish standardized operational processes that can be replicated by partners during scaling.
Wang emphasized that while Robotaxi commercialization previously focused on the 0-1 validation stage, the industry has now entered the 1-1000 scaling phase. Pony AI has outlined an ambitious roadmap: expanding to 1,000 vehicles by end-2025, 3,000 by 2026, and 100,000 by 2030.
The breakthrough comes from comprehensive cost optimization. Compared to its sixth-generation vehicles, Pony's seventh-gen Robotaxi reduced autonomous driving kit BOM costs by 70%, leveraging domestically produced L2+ components and mass-produced solid-state LiDARs like Hesai's AT128. Algorithm improvements enhanced sensor noise processing capability by 30x, enabling safer performance with cheaper hardware.
Operational efficiencies were equally crucial. Pony's Robotaxi now achieves 23 daily rides generating ¥299 revenue in Guangzhou - enough to cover hardware depreciation and operating costs. Insurance premiums are 50% lower than human-driven taxis due to superior safety records. Remote assistance ratios reached 1:20 (staff:vehicle), targeting 1:30 by year-end.
The industry landscape is becoming increasingly competitive. Waymo continues its engineering-perfection approach with 450,000 weekly orders, while Tesla pushes vision-only solutions. Domestic players like Baidu's Apollo Go report 250,000 weekly rides, and XPeng plans L4 models without LiDAR by 2026.
Wang noted that with hardware costs falling below ¥250,000, the competition has shifted to operational efficiency. UBS research predicts operational expenses (maintenance, insurance, energy) will grow from 48% to 55% of total vehicle costs, making urban mobility expertise more valuable than pure algorithm capabilities.
Looking ahead, Wang identified the Middle East as a promising market due to strong policy support, though China and the U.S. remain primary battlegrounds. He stressed that L4 adoption requires regulatory-approved safety records that can't be shortcut by L2+ data, giving early movers like Pony an advantage.
The CFO also differentiated L4's reinforcement learning approach from automakers' imitation learning methods, arguing that only the former can achieve the 10x human safety standard required for full autonomy. As the industry moves toward 100,000-vehicle scale, Wang believes the winners will be those who master both technological depth and operational excellence in building mobility ecosystems.
Comments