Pony.ai (PONY.US): The Behind-the-Scenes of Robotaxi Turning Profitable

While the autonomous driving industry is still debating “when it will become profitable,” Pony.ai (PONY.US; 02026) has answered the question with a financial report: right now.

In 2025, this nine-year-old autonomous driving company, in the Robotaxi segment—dubbed the “first truly real application of physical AI”—has run a successful commercial closed loop. The full-year financial report shows that its autonomous driving mobility service revenue reached RMB 116 million, surging 128.6% year over year. Among them, passenger paid revenue grew by nearly 400%. More importantly, the company has already turned single-vehicle profitability in Guangzhou and Shenzhen, two major first-tier cities—meaning Robotaxi is no longer a money-burning experiment, but a business that can add up.

“Scale mass production is the foundation of profitability.” Pony.ai co-founder and CFO Wang Haojun told Zhitong Finance at a media exchange in Guangzhou on April 1.

Indeed, a breakdown of the performance shows that the deployment speed of Pony.ai’s seventh-generation autonomous driving taxis exceeded expectations. The fleet size expanded from 270 vehicles at the end of 2024 to more than 1,400 vehicles today, an increase of over four times. The company expects it to exceed 3,000 vehicles by the end of 2026. Among these, the performance in the Shenzhen market is the most convincing. On March 22, 2026, the market’s single-day average net revenue per vehicle hit a record high of RMB 394. The average number of orders per vehicle reached 25. Achieving profitability in the early stage of large-scale ramp-up validates that the commercial model of L4 autonomous driving taxis is economically feasible.

With single-vehicle profitability turning positive in regions such as Shenzhen, Wang Haojun pointed out that when Robotaxi reaches profitability on a unit economics (UE) basis, it is after removing fixed costs. A business economic model can then be directly derived—where expanding scale enables overall profitability. This means that as the fleet expands, the company will see greater gross margin room, ultimately driving continuous improvement in the group’s overall profitability level.

As the fleet expands toward 3,000 units, Pony.ai’s revenue structure is undergoing a qualitative change. In 2025, Robotaxi business revenue was RMB 116 million, up 129% year over year. The share of Robotaxi revenue in total revenue increased from about 10% to 18.5%, lifting the overall gross margin from 15.2% to 15.7%. Meanwhile, with the expansion of commercial applications, the Robotruck business also reached RMB 284 million in revenue. Driven by growth in demand for autonomous driving domain controllers, technology licensing and application service revenue was RMB 229 million, up 19.7%. The delivery volume of domain controllers increased by five times compared with 2024, demonstrating a steady growth trajectory.

And when it comes to how to balance the ongoing improvement in Robotaxi profitability with the pressure on gross margin from increasing investment in business expansion, Wang Haojun provided the explanation logic of “prioritizing expanding revenue.”

“First, as RoboTaxi is our high-gross-margin business, its continuously increasing share will inevitably lift the overall gross margin. Second, if we want to expand the RoboTaxi business into more regions—whether entering completely new cities or extending from existing cities like Guangzhou and Shenzhen into new areas—gross margin may, in the short term, face some pressure. However, the gross margin level of this business is still far higher than the RoboTruck and Licensing and applications segments. What is our current core priority? Is it to improve gross margin first, or to expand revenue first? The answer is without question the latter—revenue growth must be driven by the RoboTaxi business. Because RoboTaxi growth will increase its share in total revenue, naturally pulling up the group’s overall gross margin,” he added.

It is worth noting that although Pony.ai has expanded its domestic operating regions from the four major first-tier cities to new first-tier cities such as Hangzhou and Changsha, and overseas it has entered markets such as Croatia and Singapore through collaborations with Uber, and Rimac’s Verne, market data shows that its technology and operations costs are also declining. For example, in terms of the BOM (bill of materials) cost of the seventh-generation vehicle’s autonomous driving suite, it is down 70% compared with the sixth generation. Within that, the cost of on-board compute units is down 80%, and the cost of lidar is down 68%. The ratio of remote assistance personnel to vehicles has increased from 1:20 last year to 1:30, and even higher today.

In addition, Wang Haojun also revealed that subsequent cost optimization efforts will continue to be advanced—for example, vehicle costs, which account for half of total cost, are a key focus area the company is working to optimize. In addition, there is still room for further optimization in areas such as insurance, network services, and staffing for ground support personnel.