Xiaoma Zhixing's Commercialization Surge, Part B: The Reality of "Short Orders" and the Breakthrough with Co-Building Fleets

After a long period of technical validation, the leading companies in the autonomous driving industry are now turning over their results from commercial mass production to the market.

The financial report released by Pony AI shows that in 2025, revenue reached $90 million, up 20% year over year; in the same period, the loss was $76.8 million, narrowing by more than 70% year over year.

What draws the most attention in these results is the progress of autonomous driving ride-hailing services (Robotaxi).

In 2025, Pony AI’s Robotaxi generated $16.6 million in revenue, up 128.6% year over year.

The main driver behind this growth is passenger fare revenue on the passenger side after the vehicle fleet size expanded. In the fourth quarter of 2025, Pony AI’s Robotaxi passenger fare revenue increased by more than 500% year over year.

More specifically, Pony AI has achieved an operational milestone in Guangzhou and Shenzhen—turning city-wide per-vehicle unit economics (UE) from negative to positive.

On March 22, 2026, in Shenzhen, the seventh-generation autonomous taxi achieved a new record high for average net daily revenue per vehicle of 394 yuan. On that day, average order volume per vehicle reached 25 orders.

Based on the above daily revenue and order volume, using 全天候科技’s calculations, the average fare (per customer per trip) in Shenzhen is approximately 15.76 yuan. Combined with Shenzhen’s local start fare of 10 yuan and the per-kilometer charge of 2.7 yuan per kilometer, this means that operations are still limited to a relatively short driving range.

In response, Pony AI’s CFO, Wang Haojun, acknowledged to 全天候科技 that currently, operations in Shenzhen are indeed mainly made up of short-trip orders. This is largely constrained by the current deployment areas concentrated in Bao’an and Nanshan in Shenzhen. However, it is expected that as more districts in Shenzhen and Guangzhou open up this year, the order mix will shift from primarily short-trip orders to a hybrid model combining short and long trips.

As the number of kilometers increases, average takeover mileage (MPI) data is coming under increased scrutiny.

For example, at the end of last year, a Model 3 equipped with FSD v14 departed from the U.S. West Coast in Los Angeles and crossed the entire continent, reaching the East Coast in South Carolina within 2 days and 20 hours. The full 2,732 miles relied 100% on FSD, covering complex scenarios including highways, urban roads, nighttime driving, and multiple entries and exits from supercharging stations. Throughout the entire journey, there was no manual takeover of any kind, sparking extensive discussion in the market.

Even Jim Fan, head of NVIDIA’s robotics business, exclaimed, “Tesla FSD v14 may have already passed the ‘Physical Turing Test.’”

But in Wang Haojun’s view, MPI does not apply to the L4 stage.

“Actually, once you truly reach the stage of L4 scaled operations, people don’t even talk about MPI anymore—because this thing itself doesn’t hold up. Since there are no human drivers involved, you don’t need to discuss takeover issues. The primary focus of L4 operations is still on scaled deployment. The larger the deployment scale, the lower the accident rate. In addition to that, we also need to look at the proportion of remote assistance,” Wang Haojun pointed out.

Wang Haojun further noted that in fact, if you observe the current landscape, you’ll find that companies like Waymo are no longer emphasizing the concept of MPI. Meanwhile, many L2+ companies, as they push toward L4, still mention MPI. When we truly examine the current situation, we find that achieving scaled L4 operations hinges on operational scale and the proportion of remote assistance.

Looking ahead, Pony AI has set a goal of deploying more than 3,000 autonomous taxis across over 20 cities worldwide by the end of 2026.

A fleet capacity surge at this scale, relying solely on capital-intensive in-house vehicle fleet investment, is clearly a bottomless pit that devours cash flow.

In response, Pony AI’s solution is to “handle both: expanding into cities and co-building fleets with third parties.”

In terms of city expansion, Pony AI not only plans to continue increasing density in China’s first-tier cities, but also plans to expand into new first-tier cities such as Hangzhou and Changsha.

Under the “co-built fleet” model, Pony AI effectively shifts the capital-intensive vehicle purchase costs to downstream partners. Third parties such as Ruichi (如祺出行) fund the purchase of vehicles and share the operating returns. Pony AI then takes a back seat, earning revenue through licensing its AI autonomous driving technology.

Since collaborations under this model began in the third quarter of last year, the number of vehicles currently in operation is relatively small. Wang Haojun expects that in the second half of 2026, as Robotaxis co-developed with third parties gradually come into service, they may generate more revenue.

However, the overall expansion pace still depends on how quickly policies open up in each city.

At present, China’s city-region operations have not yet opened up large-scale mutual recognition mechanisms. This means that for each new city a Robotaxi company enters, it must go through a step-by-step process—from road testing by safety drivers to gradually advancing toward final commercialization with fully driverless operations.

The pace of overseas policy rollout is also similar.

Recently, when Tekedra Mawakana, co-CEO of Waymo, was interviewed, she also said 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 that lack Robotaxi regulatory rules.

Overall, the competitive phase for Robotaxi companies in China is still mainly focused on deploying as many vehicles as possible to gain a first-mover advantage.

As an industry leader, Waymo has already moved into the phase of competing on order volume. It plans to achieve more than 1 million paid Robotaxi rides per week in the U.S. market by the end of 2026.

In this new stage of Robotaxi, technology is no longer the only moat. Whoever can first run a scalable commercial closed loop through order volume will be the one who can truly stay at the table.

Risk Warning and Disclaimer Clauses

        There are risks in the market; invest prudently. This article does not constitute personal investment advice, and it has not taken into account the special investment objectives, financial conditions, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article are consistent with their specific circumstances. Invest at your own risk; responsibility lies with you.