Conversation with Pony.ai CFO Wang Haojun: The biggest competitor in the autonomous driving industry is ourselves.

Since 2026, the global autonomous driving industry has entered a key turning point, gradually moving from technology R&D and validation toward large-scale commercial deployments. At the same time, behind the rapid growth of the industry, multiple challenges—including technology, costs, policies, and ecosystems—have yet to be resolved, testing the ability of the entire industrial chain to achieve coordinated breakthroughs.

As a representative company in autonomous driving, Pony AI has also released its 2025 performance results. On March 26, Pony AI published its 2025 Q4 and full-year financial reports. The financial report shows that in 2025, Pony AI’s total revenue reached 629 million yuan, up 20% year over year. Among them, the Robotaxi (autonomous taxi) business, which has drawn the most attention from the outside world, recorded revenue of 116 million yuan over the past year, up 128.6%. In Q4, Robotaxi revenue was 46.6 million yuan, accounting for 40% of the full year, up 160% year over year. Passenger fare revenue grew by over 500% year over year.

Pony AI has achieved an operational milestone in Guangzhou and Shenzhen, with the unit economics (UE) for cars across the city range turning positive. This means that in first-tier cities like Guangzhou and Shenzhen, Robotaxi revenue can already cover all operating costs, including vehicle depreciation, insurance, operations and maintenance, remote assistance, and more, while generating positive cash flow.

Recently, at a media briefing by Pony AI, co-founder and CFO Wang Haojun and several other media outlets including Caijing Network Technology held a discussion. He said, “For a growing company, the most important thing is always the Top line (revenue).”

In the recently disclosed financial report, Pony AI has continuously emphasized that UE has turned positive in its dual-city operations in Guangzhou and Shenzhen. Wang Haojun also said candidly that for Pony AI, the most important milestone last year was UE turning positive. “Pony AI doesn’t adopt a low-price strategy. It’s not that Pony AI is doing promotions and customer acquisition by discounting in order to get more orders. The end result of a low-price strategy is that there won’t be a stable stream of daily, repeatable user orders, and that also makes it impossible to ensure that UE can remain relatively stable with sustained growth. Likewise, it can’t guarantee better performance on revenue.”

From the size of the vehicle fleet, as of March 25, 2026, Pony AI’s fleet size has expanded from 270 vehicles at the end of 2024 to 1,446 vehicles. Pony AI’s target for this year is to exceed 3,000 vehicles by year-end, grow mobility service revenue to more than three times, and expand into 20+ cities. Can this goal be achieved? It also tests Pony AI’s capabilities across areas such as technology, fleet scale, and ecosystem.

The following are excerpts from part of Pony AI’s media briefing (with edits):

Question: In this financial report, you achieved profitability for the first time within a single quarter, but mainly as a result of investments in Moore Threads. How do you view this single-quarter profitability, and do you have expectations for a timetable to reach operating profitability?

Wang Haojun: : This is just a reflection of a number, and it isn’t that important. It isn’t an item directly related to our Fundamental (the most core business). Back when we invested in Moore Threads, the most important point in our view was that we needed to develop toward a larger scale, which requires integration across the entire ecosystem, including domestically produced chips, all of which need to be considered for layout. In fact, the investment back then in Moore Threads was more driven by strategy. It just happened to allow us to achieve single-quarter profitability. We don’t think this is something worth making a big deal out of, and it has no major connection with our current core business.

In terms of our company’s eventual operating profitability, it is closely tied to how many units Robotaxi is ultimately deployed at. Today, there is a consensus in the industry that Robotaxi will reach around 100k units by 2029 or 2030, and many companies should be able to achieve operating profitability milestones.

Question: After removing the stock gains from Moore Threads, Pony AI is still at a loss, and the year-over-year loss is still expanding. Why is the loss still growing when UE per vehicle has turned positive?

Wang Haojun: : UE per vehicle itself is a growth in revenue. The reason for the loss is that last year we made many investments that had to be done for commercialization—for example, the R&D costs for the seventh-generation vehicles. Without the R&D for the seventh-generation vehicles, we wouldn’t be able to achieve UE turning positive. In my view, this investment is necessary. We’re not saying that once UE turns positive, the company will quickly move onto a so-called path to profitability. We’ve always emphasized that if UE per vehicle turns positive, what the company needs to do is grow revenue—and that growth needs to be much faster than the growth in spending. That is what a good model looks like.

Question: Since last year, you’ve achieved UE turning positive in two cities, Guangzhou and Shenzhen. What are the challenges in replicating this model in other cities to achieve UE turning positive? After scaling up, how will costs and revenues change?

Wang Haojun: : In a new city, the cost of the vehicles is the same, and operating costs may be a bit lower—for example, my employees’ salary. But another issue will arise: in these cities, will the per-kilometer charging for cars be kept consistent with first-tier cities, or will it be lower? So, we still need a certain scale of network coverage in order to ensure that passenger fare revenue can be achieved, so that ultimately each city can, potentially, achieve UE turning positive like a first-tier city. Pony AI has achieved UE turning positive in Guangzhou and Shenzhen. More importantly, the significance is that it has set a benchmark. After setting a benchmark, when expanding, you won’t be so focused on whether each city must first reach UE turning positive before expanding. You should place more emphasis on how fast revenue is growing. Under the premise of revenue growth, more UE turning positive will naturally follow.

Question: In terms of revenue structure, Robotruck has the highest share, and the growth rate of Robotaxi is also quite fast. The revenue from technology licensing is relatively steady. What are the priorities and focus of investment for these three lines of business this year?

Wang Haojun: : Currently, Robotaxi is the most important. Robotruck is indeed highly linked with Robotaxi in terms of technology. We can share 80% of the technology between Robotruck and Robotaxi. Our approach is: when Robotaxi research advances to a new generation, we apply some corresponding technologies to Robotruck’s new generation. We expect Robotruck’s revenue in the second half of the year will continue to grow. Looking at technology licensing and application, last year we were a domain controller supplier for the low-speed logistics sector in the market. Later, we also expanded into embodied intelligence, street-sweeping vehicles, and other scenarios. In general this year, the orders are still very strong, at least enough to maintain the same revenue level as last year.

Question: Regarding Pony’s “dual-engine” strategy, you also discussed in the earnings call that overseas markets have more attractive profit margins. Could you elaborate in detail on the overseas operating cost structure and where the biggest differences from domestic operations are?

Wang Haojun: : For a growing company, the most important thing is always revenue. As I mentioned earlier, the profit opportunities in overseas markets are relatively more attractive. But in more detail, it’s not that every market is like that. In any overseas market, if its labor costs are very high and passenger fares are very high, then you can consider its marginal benefits to be high.

Question: Following up on the overseas topic, I’m also quite curious about what you just mentioned—that unmanned commercial operations will be introduced soon in the Middle East. What considerations go into pricing locally? Will your pricing be aligned with local taxi fares, or will it be lower or higher?

Wang Haojun: : This is related to local regulatory thinking. Taking the Middle East as an example, the entry costs are actually not high, which means I can pursue a low-price strategy. But the question is: why would I pursue a low-price strategy? Or does that market really need a low-price strategy? Or will they also recognize the value of autonomous driving services and be willing to pay a price similar to that of taxis. Different regions have different regulatory focuses. A sudden move to a low-price strategy could affect local taxi businesses. Regulators there might also judge whether they should provide guidance pricing to the company, similar to taxi fares. If there is such guidance, we will certainly comply with local regulators’ views.

Question: Guangzhou Toyota and Toyota China, together with Pony AI, are jointly building the seventh-generation vehicle, which is set to be deployed this year. I just saw it’s 1,000 units. If this year the fleet continues to grow to 3,000 units, will the cost reductions brought by economies of scale be reflected in a concrete way?

Wang Haojun: : Economies of scale will reduce operating costs. If it’s 1,000 vehicles, that volume is basically nothing to the OEM. But we’ve also said that even if vehicle cost reductions aren’t achieved through scale, but through continuous optimization of ADK (autonomous driving kits) costs, then this year Pony AI can reduce ADK costs by 20% compared with last year. This is achieved through our own technology, not simply through economies of scale.

Question: As the fleet size expands and the company expands into more cities, will Pony AI’s Robotaxi business gross margin experience relatively large fluctuations? Especially with the co-built fleet model, in the long run, will it dilute the company’s profit space?

Wang Haojun: : Actually it won’t, because co-building itself brings two lines of revenue. One is vehicle sales revenue. After all, the capital expenditures for this vehicle are not on Pony AI’s side, so we will have vehicle sales revenue. The second is technology licensing fees. Licensing itself is a high-gross-margin business. If you look just at your own Robotaxi, its gross margin is higher than Robotruck’s technology licensing fees. Today, Robotaxi is fluctuating in total revenue due to the impact of some project-based factors, which is also why we’re working hard to increase the portion of recurring revenue. If recurring revenue can be increased, in the future we can ensure businesses with higher gross margins.

Question: Right now, automakers, ride-hailing platforms, and technology companies are all entering the Robotaxi track. In your view, what kind of competitive landscape will it form in the future? Who will be Pony’s biggest competitor, and what will Pony’s moat be?

Wang Haojun: : The biggest competitor in the autonomous driving industry is definitely itself. What we care about most is still pushing the whole industry forward. Although we believe a turning point has arrived, people still have more questions in mind—whether the turning point truly has come. How exactly should revenue be generated? For this industry, we need to push it forward. So what I’m more concerned about is how we can move faster. It’s a good thing that many players are entering autonomous driving—whether it’s the industry or the ecosystem—because once they come in, some costs can continue to go down. It’s just that in the end, how many of these new players will actually succeed? Today, the safety requirements for L4 are much higher than for L2+. That means only those who can build such a technology stack can ensure very high levels of safety while also covering overall operating efficiency. The complexity is far higher than L2+. My prediction is that the number of companies that can make it to L4 will be far fewer than those that can make it to L2+. If that’s the case, there will be only a few players ultimately, but their number will still be much smaller than the L2+ companies.