The Other Side of Pony AI's Commercialization Push: The Reality of "Short Trips" and Breaking Through via Co-built Fleets

After a long period of technical validation, leading companies in the autonomous driving industry are delivering their commercial mass-production report cards to the market.

Pony AI's financial report shows that revenue reached $90 million in 2025, a 20% year-on-year increase; during the same period, losses were $76.8 million, a year-on-year reduction of over 70%.

The most closely watched aspect of this report card is the progress of its autonomous driving mobility services (Robotaxi).

In 2025, Pony AI's Robotaxi generated $16.6 million in revenue, a 128.6% year-on-year surge.

The primary growth driver was passenger fare contributions following the expansion of the vehicle fleet. In the fourth quarter of 2025, passenger fare revenue from Pony AI's Robotaxi service increased by over 500% year-on-year.

Specifically, Pony AI achieved the operational milestone of turning its city-wide unit economics (UE) positive in Guangzhou and Shenzhen.

On March 22, 2026, the average daily net income per vehicle for seventh-generation autonomous taxis in Shenzhen hit a record high of 394 yuan, with an average of 25 orders per vehicle that day.

All Weather Technology calculated based on the aforementioned daily income and order volume that the average price per order in Shenzhen is approximately 15.76 yuan. Considering Shenzhen's base fare of 10 yuan and a mileage fee of 2.7 yuan/km, this means that current operations are still confined to a limited travel range.

In response, Pony AI CFO Wang Haojun admitted to All Weather Technology that operations in Shenzhen are indeed dominated by short-distance trips. This is mainly because the current deployment areas are concentrated in the Bao'an and Nanshan districts of Shenzhen. However, it is expected that as more urban areas in Shenzhen and Guangzhou open up this year, the order structure will shift from a short-trip model to a hybrid of short and long trips.

As mileage increases, Miles Per Intervention (MPI) data is garnering attention.

For instance, at the end of last year, a Model 3 equipped with FSD v14 traveled from Los Angeles on the U.S. West Coast to South Carolina on the East Coast in 2 days and 20 hours. The entire 2,732-mile journey was 100% dependent on FSD, covering complex scenarios such as highways, urban roads, nighttime driving, and multiple entries and exits from Superchargers. The entire trip was completed without any human intervention, sparking widespread discussion in the market.

Jim Fan, head of robotics at Nvidia, even exclaimed: "Tesla FSD v14 may have passed the 'Physical Turing Test'."

But in Wang Haojun's view, MPI is not applicable to the L4 stage.

"In fact, once we truly reach the stage of L4 scaled operations, people no longer mention MPI; the concept itself becomes irrelevant. Since there are no human drivers involved, the issue of intervention is moot. The main focus of L4 operations is large-scale deployment. The larger the scale, the lower the accident rate. Beyond that, the focus should be on the ratio of remote assistance," Wang Haojun pointed out.

Wang Haojun further noted that, in reality, observations show that companies like Waymo no longer emphasize the concept of MPI. Meanwhile, many L2+ companies are still mentioning MPI as they push toward L4. When examining the current situation, the keys to achieving L4 scaled operations are operational scale and the remote assistance ratio.

Looking ahead, Pony AI has set a goal to deploy over 3,000 Robotaxis in more than 20 cities globally by the end of 2026.

Such a massive increase in capacity relying solely on the heavy capital investment of self-operated fleets would clearly be an endless drain on cash flow.

In response, Pony AI's solution is a "two-pronged approach": city expansion and co-building fleets with third parties.

In terms of city expansion, Pony AI plans to not only continue densifying operations in domestic tier-one cities but also expand to new first-tier cities such as Hangzhou and Changsha.

Under the "co-built fleet model," Pony AI essentially transfers the heavy asset cost of purchasing vehicles to downstream partners. Third parties such as Ruqi Chuxing invest in purchasing vehicles and share operating revenue, while Pony AI provides the AI autonomous driving technology license to generate revenue.

Since cooperation under this model began in the third quarter of last year, the current scale of vehicles in operation is small. Wang Haojun expects that as more third-party Robotaxis are put into service in the second half of 2026, they will contribute more revenue.

However, the overall expansion speed still depends on the pace of policy liberalization in each city.

Currently, urban regional operations in China have not yet established a large-scale mutual recognition mechanism. This means that for every new city a Robotaxi company enters, it must progress step-by-step from road testing with safety drivers to final fully driverless commercial operations.

The pace of policy advancement overseas is similar.

Recently, Waymo co-CEO Tekedra Mawakana stated in an interview that in some cases, Waymo can complete the entire process from city mapping to paid rides in just a few months. But in other cases, progress is much slower, especially in cities or states lacking regulatory rules for Robotaxis.

Overall, the competition among domestic Robotaxi companies is still mainly focused on deploying as many vehicles as possible to gain a first-mover advantage.

As an industry leader, Waymo has already entered the stage of competing on order volume, planning to achieve over 1 million paid Robotaxi rides per week in the U.S. market by the end of 2026.

In the new segment of the Robotaxi race, technology is no longer the sole moat; whoever can first complete the commercial loop with scaled orders will truly remain at the table.