Home / Transcripts / WeRide Inc. (WRD) · August 12, 2026

WeRide Inc. (WRD) Earnings Call Transcript

August 12, 2026

NASDAQ US Consumer Discretionary Automobile Components earnings 68 min

Earnings Call Speaker Segments

Xu Han executive
#1

Hi everybody, and thank you for being here today. Our second quarter this year was excellent. First off, I want to call out 3 important elements behind how well we did in Q2 and point to the encouraging momentum we have right now: faster growth abroad, scaling that relies less on owning assets, and an obvious route toward cash generation that sustains itself. Across each of those 3 fronts, we moved forward significantly this quarter, putting the business in position for its next stage of expansion. Now for the numbers. Total revenue grew strongly, almost 2x versus the same period a year ago and over 2x compared with the prior quarter, at 103% growth. Revenue from overseas climbed 164% year-over-year and roughly 170% quarter-over-quarter. At the same time, gross margin hit an all-time peak of 38%, an improvement of about 10 percentage points versus Q2 of last year. Thanks to ongoing gains in operating efficiency, EBITDA also got better, with the loss shrinking 8% against the second quarter of 2025. On the operational side, our L4 fleet stood at roughly 3,400 units, a rise of 22% from the earnings release in April. Inside that figure, our robotaxi fleet added 500 units, reaching over 1,800 vehicles, which works out to 40% growth across that same span. Beyond that, our 1-stage end-to-end L2+/L3 solution has moved past winning design awards into full-scale mass production. This business's revenue jumped nearly 2,600% year-over-year and rose 219% quarter-over-quarter during Q2. In the quarter, we shipped about 30,000 units of our L2+/L3 solutions, marking the start of scaled, high-growth commercialization for that L2+/L3 business. Looking at the broader industry, we think our position is unique, since we have mature technology stacks plus large-scale commercial deployment for both product lines: the multi-sensor fusion L4 autonomous driving solution and the proprietary one-stage end-to-end L2+/L3 solution built for mass production. Moving to strategy. During this quarter, we split our business into 3 areas: L4, L2+/L3 ADAS and AI infrastructure, which gives greater clarity for our long-term strategic planning. In Q2, we took further steps in our strategy as a physical AI company. That rests on proprietary infrastructure and foundation models, with autonomous driving as the most commercially advanced use of physical AI. Our physical AI infrastructure lets us distill what large foundation models can do into efficient, lightweight onboard models. What matters most is our capacity to transfer the underlying knowledge and the representations that large-scale models learn — not just their output — into smaller models, all while keeping the intelligence, generalization and decision-making abilities needed for real-world autonomous driving. That lets us deliver ever more powerful AI capabilities onto cost-efficient onboard systems, which supports both scaling our L4 business and the ongoing evolution of our WRD 3.0, meaning our L2+/L3 solutions. Even more significant, this produces a strong technology flywheel. Our real-world L4 operations create high-value data that keeps improving safety, robustness and generalization, while our expanding L2+/L3 adds more road data, speeding up L4 model development and broadening coverage of long-tail scenarios. There are 2 separate businesses. These are not 2 separate businesses. We have two mutually reinforcing layers within our physical AI platform, and with each vehicle put into service and each mile traveled, our technology becomes more intelligent, safer, and more capable. Now I'd like to take you through the particular advances we made during this quarter on both the technology development side and the commercial deployment side. Let's begin with our L4 business, specifically robotaxi. The main themes for us are asset-light overseas acceleration along with a strengthening regulatory mode. Our operations abroad rest on an asset-light model; we don't own the operating vehicle assets. The vehicle acts as a hardware entry point for entering new markets, and we partner with local players to deploy and run the business. Thanks to this approach, we can scale our footprint with substantially reduced capital needs and enhanced operational flexibility. Assuming optimal utilization under normalized fess operations, we project that steady state annualized technology service revenue for robotaxi might surpass USD 50,000. As the fleet grows, we anticipate growing gains from data network effects, cross-market learning, and algorithm generalization, which ought to further enhance vehicle level economics and generate meaningful operational leverage. I'd also like to make clear how we define unit economics. Our overseas UE isn't grounded in a fleet ownership plus ride-hailing platform model. Their revenue comes from fare charging at a gross level. Our economics, by contrast, rest on a licensed virtual driver technology model. We deliver regulated, locally verified and recognized autonomous driving capability, and we collect a recurring fee from autonomous driving technology services along with knowledge-based fees. Since our positioning and our business model are fundamentally different, the very same term UE carries quite distinct operational and financial implications compared with those found in a traditional ride-hailing platform. With backing from partners such as Grab, Uber, Green Mobility, SBB and so forth, we are replicating — we are replicating this proven asset-light operational model throughout Mid East, Europe and Southeast Asia. This signals that our overseas expansion has now reached a phase of large-scale commercialization. To give you more specifics on the second quarter, we announced new robotaxi commercial partnerships in Madrid, Spain and Zurich, Switzerland and Copenhagen, Denmark. In the Middle East, we keep expanding food driver age robotaxi operations in Abu Dhabi and Dubai this quarter, now covering over 70% of the core urban areas. In Riyadh, our robotaxi operation zone has grown to the airport terminals and surrounding central business districts. By the end of 2026, we will be fully entering a phase of meaningful growth supported by secure licenses, expanding fleet, recurring orders, solid revenue and profitability in overseas markets. This is going to show the successful execution of our overseas acceleration and asset-light expansion strategies. When it comes to the domestic market, we keep strengthening our operational capabilities and expanding our driverless operation zones in Guangzhou to key areas of Tianhe District, including Zhujiang New Town and the Haizhou District, including the Canton Fair Complex. This amounts to a nearly threefold expansion in our ODD relative to the end of 2026. Even more importantly, our operating efficiency kept improving meaningfully over the quarter. On average, each vehicle completed 21 rides per day, marking a 24% increase from the prior quarter, and the highest number of daily trips finished by a single vehicle rose to 28. As a ride-hailing operator running its own robotaxis, we saw registered users in China grow by almost 35% compared with the previous quarter. Growth across scale, utilization and our user base at the same time produced a 140% jump in ride-hailing revenue on a sequential basis, which shows even more clearly how scalable our operations are and how their profitability keeps improving. Going forward, with China's rules for autonomous driving steadily becoming more developed, we anticipate extending our robotaxi business into additional cities soon. Meanwhile, we will keep making Guangzhou our domestic reference point, aiming to bring driverless robotaxis into the public transportation system on a large scale. Taken together, these efforts will lay a solid groundwork for the coming stage of commercialization at home. Separately, I'd like to step back and talk about our regulatory approach. The operational licenses we've obtained abroad are not merely routine administrative sign-offs. They reflect a blend of many years of technology adaptation, ecosystem integration and strict safety validation. That forms a key competitive barrier placing us 2 to 3 years ahead of the market. This approach rests on 3 central pillars. The first is technology localization and regulatory engagement.

Xu Han executive
#2

Our technology gets adapted to the specific ways traffic is lost, the state of the roads, and how people drive in each place, and at every step we work hand in hand with regulators. Number two is about integrating with ecosystems and infrastructure. We achieved deep technical integration alongside local operators and service providers, and that's supported by strong execution on the ground. Number three is strict validation in the field and safety performance. Localized real-world data keeps building up across our fleet, and our safety framework plus our track record provide the basis for earning regulators' trust and securing commercial licenses. A key point: because regulatory frameworks are becoming more and more aligned from one market to another, whatever we've learned and proven in a single jurisdiction can assist in accelerating approvals elsewhere. That brings down the expense of staying compliant on margins, stops us from duplicating effort, and creates operating leverage that keeps growing. As a result, we can expand our asset-light self-driving operation around the world faster and with more efficient use of capital. The same mobility platform also noted during its Q2 investor presentation that the industry is increasingly aligned on what the core obstacles at L4 are. What determines technical competitiveness isn't how many total driving miles you accumulate — it's what it costs to obtain rare scenarios, how varied those scenarios are, and the practical quality of the data. On top of that, general-purpose large foundation models aren't able to meet what autonomous driving requires. AV systems depend on proprietary vehicle hardware and tight regulatory supervision, so you need models tailored to match hardware specifications and compliance obligations. Thanks to a strong proprietary tool chain that handles data end to end, plus a varied data set built up through years of cross-border L4 operations and excellent algorithms we developed ourselves, we continue to raise the bar on vehicle engineering safety. Regulators worldwide trust us because our safety results are proven in the field, which speeds up our expansion into new regions and creates a unique core edge for our global footprint. Now for our L2+/L3 business, three things show how we are doing: technology that leads the industry, fast commercialization at scale, and a route to generating cash that sustains itself. Building on our top-tier advantage in L2+/L3 algorithms, we have won China's Intelligent Driving Competition 6 times in a row, which strengthens our position as a leader in intelligent driving technology. On the commercial side, we are clearly speeding up, with several OEM mass production programs moving forward steadily. We just started an L3 autonomous driving proof-of-concept initiative together with Mercedes-Benz. That marks a key endorsement of our technology from a major global OEM and shows once more how strongly the industry recognizes what we can do technically. By June 30, we had delivered more than 30,000 vehicles cumulatively across various models fitted with our single-stage, end-to-end L2+/L3 offering throughout the reporting period. Going forward, we anticipate that over 100,000 vehicles will run on our solution by the close of the year, and cumulative deliveries should climb past 0.5 million units in the following year, which sets us up for substantial expansion in our L2++ ADAS business. Last, here's a quick update on our AI infrastructure. During this quarter, we formally introduced the WIT model, a large-scale physical fact foundation model, and with it we closed out an end-to-end technical stack along with full capability coverage across understanding data, producing data, and the whole chain. It works in close tandem with our Genesis world model. From road test data, the WIT model extracts and validates physical facts, building a cognitive groundwork that enables machines to perceive and make sense of the real world. Leveraging those confirmed physical facts, Genesis reconstructs, via simulation, a broad array of extreme operating conditions as well as long-tail scenarios. Together, the two models operate a closed loop that spans generating data, simulating, and iterating algorithms, thereby driving ongoing upgrades throughout autonomous driving technology. Right now, it backs our main business areas: L4 self-driving and high-volume L2+/L3 production. So to close out, in Q2 we reached important milestones and delivered solid performance across the whole business, which takes me back to three central themes I mentioned when I opened today's call: faster overseas growth, an asset-light approach to expansion, and cash generation that funds itself. Looking ahead, we know exactly where we're headed. When it comes to L4, in markets abroad, we'll leverage our edge in regulation and permits to replicate our proven robotaxi playbook across Europe and beyond, accelerating the move from successful validation to deployment at scale. In China, our deep market approach will continue to expand. We'll set up flagship operations that are completely integrated with urban transportation networks, and after that, we'll grow in a disciplined, repeatable manner across major cities throughout the country. When it comes to L2+/L3, our number one goal is simple: secure additional OEM partnerships, capture greater share with important OEMs, and roll out faster at scale. The ultimate aim for gross installed vehicles — apologies, growth is what matters most. Rather than pursuing expansion purely for its own sake, we'll direct capital toward areas where we have strong conviction and real potential for meaningful scale. At the same time, we'll remain committed to generating long-term cash flow as well as profitability, instead of scaling up just to scale up. Before we wrap up, I'd like to acknowledge my team beyond just the figures. Hard challenges are something this group takes on willingly; they look reality in the eye and stay unwaveringly committed to getting things done. Thanks to that ethos of honest thinking and superb operational discipline, we're able to reliably transform vision into products, then products into services, and services into substantial commercial value, even when conditions are unpredictable. Our physical AI foundation model, the WIT — word intelligence toward truth — serves as a reminder that genuine intelligence has to rest on facts that can be observed and confirmed about the physical world. That principle applies to our business too. In the end, what counts isn't the claims we make — it's what we keep delivering in the real world. Stories can shift, tools can advance, and ups and downs in the market will always come and go—yet the facts stay put. What holds up is actual results out in the field, and nothing speaks with more force than a history that has been tested and proven. With that, I'll now pass things to Jennifer, our CFO, to take you through the specifics of how we performed financially in the second quarter of 2026.

Xuan Li executive
#3

Thanks, Tony. Greetings, everyone. I'll begin with what matters most from this quarter. For WeRide, Q2 represented a significant turning point, featuring robust top-line expansion, further gross margin gains, and rising operating leverage throughout our operations. Crucially, that expansion came from three areas where we see potential to underpin a more durable financial model down the road: faster growth in our business abroad, scaling up our asset-light L4 approach, and swift commercialization of our L2+/L3 operations. We are seeing more and more evidence of this strategy paying off in our financials, especially in revenue growth, gross margin, and capital efficiency. During Q2, revenue came to RMB 232 million, which is 82% higher than a year earlier and 103% above the prior quarter. Crucially, that growth is spreading across both of our businesses and geographies. Revenue from L4 hit RMB 125 million, rising 47% compared with the same period last year, mainly on the strength of robotaxi. Our priority continues to be expanding the core L4 business, robotaxi in particular. At the same time, several of our other L4 operations follow a natural seasonal pattern, with commercial negotiations usually wrapping up in the first half and revenue being recognized as deployment occurs during the latter half of the year. As those projects enter deployment, we anticipate a bigger contribution from our wider L4 portfolio during the back half of this year. Our L2+/L3 business kept speeding up considerably, with revenue climbing roughly 26x versus a year ago and 219x versus the previous quarter. That shows a major shift from project development toward mass production and scaled vehicle deployment. As additional OEM programs advance toward mass production, we expect L2+/L3 to turn into a progressively more significant engine of our total gross revenue growth. Markets outside China make up almost 40% of group revenue and are also becoming a more and more important source of growth. Revenue from overseas climbed 164% year-over-year and roughly 170% quarter-over-quarter. Even more significantly, the economics of our international operations are becoming more and more appealing. Because our model is asset-light, we can expand via local partners and the operating ecosystem without needing a matching rise in investments on our own balance sheet; that provides us with three key benefits: quicker geographic expansion, better margins as operations scale, and lower incremental capital needs. We see this approach as a crucial basis for our long-term journey toward growth that sustains itself. Now to profitability at the gross profit line. Gross profit rose 143% year-over-year to RMB 87 million, while gross margin widened by 9.4 percentage points to 37.5%, compared with 28.1% in the same quarter a year ago. That gain came from continued scaling of our high-margin, asset-light overseas L4 business, fast growth in L2++, and a broader move toward higher-value AI service-oriented revenue. As this mix keeps improving, we think gross profit has the potential to outpace revenue growth over time. Total operating expense came to RMB 533 million in Q2, up slightly by 9.2% year-over-year, a touch slower than revenue growth. This is an early sign that operating leverage is starting to take effect. R&D expense rose 36% year-over-year to RMB 434 million, largely reflecting our added investment in AI infrastructure and foundation models. Simultaneously, the fact that total operating expense grew far more slowly shows we're starting to get better efficiency as the business gets bigger. G&A dropped sharply to RMB 69 million, largely from reduced share-based compensation and professional fees, whereas selling expense rose only to RMB 29 million amid our commercial expansion. On the whole, operating leverage is becoming visible, with operating expense climbing far under revenue. Tying that together, our net loss shrank 1% year-over-year to RMB 401 million in Q2, and EBITDA loss shrank 8.1% year-over-year to RMB 335 million. This outcome captures the early payoff of our asset-light model and rising operating leverage, especially as we divest our overseas business and strengthen our route to sustained cash generation. Last, at June 30, we hold roughly RMB 5.4 billion in cash and other liquid financial resource. That gives us a solid financial base for further expansion. Let me finish with 4 key takeaways. One, revenue growth is speeding up and turning more diversified. Two, our revenue mix is moving toward the higher-margin physical AI service-oriented business. Three, operating leverage is starting to show up, with expense rising much more slowly than revenue. And four, our robust balance sheet and asset-light model give us a capital-efficient base for further expansion. We think Q2 shows an early financial payoff of the model, and we've been building global expansion, asset-light deployment, higher physical AI service revenue and rising operating leverage. Going forward, we're focused on turning our commercial pipeline into scale deployment, keeping improving margin and steadily converting revenue growth into stronger cash generation. We see these as the core building blocks for durable capital-efficient growth, a leading position in global autonomous driving and a lasting route to profitability. With that, operator, we're ready for questions.

Ming Chung analyst
#4

Congrats on those impressive results. My first question: we see that WeRide right now has a business portfolio that's diversified, covering L4, L2++, L3, and AI infrastructure too. What I'd like to know is how does management decide where resources go among the various businesses and regions? And what targets do you have for the second half of this year?

Xu Han executive
#5

Okay. Thank you for asking. To begin, I'd like to give a somewhat detailed explanation of our financial information model. As you've seen recently, virtually every industry, I believe, has gotten a capability boost similar to what happened 5 years back, thanks to large language models and AI advances. I don't believe any company worldwide is capable of doing L4 and L2++, L3, and AI infrastructure all at once. Yet lately we've observed that, with our AI model improvements, particularly our Genesis and WIT, we realized we can definitely create plenty of technology-to-data synergies. Incidentally, I want to say a bit about why the WIT name was chosen. This model, WIT, was announced at the 2026 AI World Conference. It's a tribute to a well-known German philosopher, whom I regard as the greatest, Ludwig Wittgenstein. His point — his celebrated statement that the limit of my language is the limit of my world. That's precisely what we've just come to understand. Every video can be segmented into what we term minimal fact units. Essentially, we're able to process each video and break each video down into an extremely atomic fact. Then, together with our Genesis model, we ultimately discovered we can produce any kind of data we want, much like in the matrix someday, when Leo suddenly realized he saw the world's true nature. That's the sort of feeling we have. Consequently, we cut down dramatically on the resources required. And we can — through a single investment, accomplish 3 things, namely L4, L2+/L3 ADAS, and AI infra; so we've built our infra extremely robustly and generate data from L4 and generate data from ADAS as well, letting these 2 complement one another, establishing what we call a double data flywheel. In fact, this flywheel explains why we're the sole company globally to have reached driverless vehicle operation at large scale. Meanwhile, we have numerous car OEMs embracing our ADAS system. Regarding our second quarter goal, there are several aspects. One is that we aim toward the accumulated 0.5 million installation goal; and we intend to really drive forward having many car OEMs embrace our ADAS system; and hopefully, during the fourth quarter, Tesla FSD might come into China, and we'd like a direct head-to-head comparison with FSD, since I'm a heavy FSD user in California, and I've driven our own ADAS system, our own cars built on ADAS system WRD 3.0. In my view, those 2 items can be compared, so I'd like additional car OEMs to take up the ADAS system. At the same time, we'll keep growing our robotaxi fleet. And thirdly, regarding our AI infrastructure, our aim is for additional businesses in the robotic sector and the humanized robot sector to take up our infrastructure. That's the objective. This brings my answer to a close.

Tim Hsiao analyst
#6

Tim here, from Morgan Stanley. To begin with, congratulations on the strong top line growth and the global expansion you achieved this quarter. I have a quick question regarding the overseas business. On this call, I believe management has emphasized both accelerating overseas expansion and large-scale commercialization. So I'm curious whether you could provide more detail. How should we think about your overseas strategy? And at the same time, is it replicable in different regions? That's what I wanted to ask.

Xuan Li executive
#7

Thanks, Tim. As we just said, the overseas market serves as a key growth driver engine for us, and it has already delivered roughly almost 40% of group revenue; we do think there's substantial room to run. The software case is quite compelling; every market we're entering faces severe labor shortage and climbing labor costs, which generates strong natural demand for autonomous mobility. Let me give you a sense of the pace. During Q2, in the past quarter, we rolled out 3 new robotaxi deployments in Spain and in Switzerland and in Denmark, and in the Middle East. Since last quarter, our robotaxi fleet has roughly doubled to around 400 vehicles. If you remember from the Q1 earnings, back then we had like around 200; and even considering the recent geopolitical tension in the region, we doubled the fleet during the quarter. So the momentum is real and it's accelerating. Also regarding the rapid capability, this is where we believe our model is generally differentiated. When we operate overseas, we don't own any of the vehicles in operation. What we do instead is handle the localized tax adaptation and regulatory compliance. Then we license our digital AI driver to local partners. That way we capture recurring technological service revenue especially as our operation scales; this allows us to expand very quickly without heavy capital, and we can also maintain a good margin and good cash flow. And more importantly, looking at the unit economics, the annualized recurring revenue per vehicle exceeds USD 40,000 to USD 60,000. It also carries meaningful upside as density builds up in each city; for us, our primary goal is simply to increase the fleet size in cities where we already operate. We're very disciplined about where we go next. We really pick the market with proven commercial potential and a reasonable regulatory barrier to entry. And the playbook is well established. Everywhere we go, we secure the permits, we adapt the technology locally. We set up a benchmark project in the region and flagship project in the region and then we replicate at a lower marginal cost. Our belief is that this can be duplicated at scale on a consistent basis. Right now, 4 operations are active in 12 countries, and every one of them is progressing as planned. We have confidence in both the model and the pipeline we've built, and we're eager to serve travelers in additional markets going forward.

Paul Gong analyst
#8

Two things I'd like to ask. First, on ADAS — could you share where things stand now and what's coming next for that side of the business? And since the group working on it isn't very large, where does your edge come from in such a fiercely crowded race? Then, second, if I may: how do you view OEMs, and the ADAS players too, being able to develop robotaxi services by drawing on data gathered through ADAS? What's your read on where this is heading? Could it pose a danger to your robotaxi business?

Xu Han executive
#9

Thanks a lot for those two very significant questions. I'll take them in turn, alright? The opening one concerns the most recent developments and roadmap for our L2+/L3 ADAS offering. To keep it brief, I'll simply refer to ADAS as our business, covering both L2++ and L3 systems; and that question also touches on the edge we hold at present. So to begin with, I should note that rivalry in ADAS, both within China and around the world, is extremely fierce; and at this point everyone is attempting to measure their own ADAS against FSD 14 — that is, version 14. Based on what our internal assessment shows, I'd say right now, in urban scenarios within China specifically, we stand level with FSD. I don't know if you happened to see it. Only last week, I ran a live stream showing our ADAS experience in Zhujiang New Town — that is, the Pearl Harbor New Town, the central CBD district of Guangzhou. I was driving the N60, which runs on our WRD 3.0 ADAS system. Meanwhile, I was taking questions on that live stream from everyone in my live stream room; and I kept that live stream going for 2 hours, driving through Guangzhou rush hour with zero intervention, alright? The video is available online. You're welcome to look at it. From what I understand to the best of my knowledge, this marks the first occasion that a founder of an ADAS or autonomous driving company has done a live stream in Russia or in a first-tier city with no intervention while answering questions. That demonstrates the breakthrough our technology has made. As for installation units, well, our revenue climbs by 2,600%, that's 26x year-over-year, and 220% quarter-over-quarter. Within a very brief span, we've already reached something like 30,000 units installed in a single quarter. Yet our target for this year is to hit 100,000 units installed; and next year, cumulatively, we aim to reach 0.5 million, and that's by far the quickest pace I've ever seen in ADAS growth when it comes to installation units. And that's really the reason we can pull this off. Thank you for the high regard — we're managing it with quite a small team. The core secret sauce lies in our 2 key large language models, the Genesis model and the WITT model. And I genuinely think, as all of you have already observed with AI, folks sometimes say a single person can create a company of one. And thanks to our excellent AI infrastructure plus foundation model, we can genuinely create the finest ADAS solution in China and across the globe with a crew of 200 or 300 individuals. That's what AI can do, right? So that's what I want to say regarding where we stand today. Now the second question — thank you. It's a fantastic question; namely, we can — we've witnessed quite a few OEMs, whether car OEMs or some firm offering ADAS solutions, declaring in a very boastful manner that they are getting into the robotaxi business, and they'll achieve wonderful things, with plenty of PR. So my reply to that sort of declaration or that aim is, in fact, first of all, we embrace that sort of rivalry; and since that rivalry will strengthen us, we remain the leader and the pioneer in the robotaxi field, and we're expanding worldwide. And we recognize that we also possess a very robust background in ADAS. We recognize the huge distinction. We recognize what redundancy demands for reliability. There's a 1,000x gap; so I simply want to — as I stated at the start of this year, I put forward a qualification threshold for robotaxi. Namely, a company must run for no less than half a year with a fleet of 100 driver robotaxi with no significant accidents, with no severe accidents, before it can call itself a robotaxi company. If not, it's merely a declaration. Folks can declare things daily. And I don't have to answer that sort of declaration. But I simply want to say there's a huge distinction between an ADAS system and an L4 robotaxi; and folks will say, fine, with this progress in large language models, we can fix the issues that previously required something like 10 years of your efforts, we can fix them in probably half a year; now I'll make a further claim, which is that hallucination in digital AI may be entertaining, yet it is fatal in physical AI; so in physical AI, you can — our society, our public roles cannot tolerate a fatal accident if we can prevent it. Yet in the digital realm, you can simply — you can simply reboot the computer or simply do it once more. But in the actual world, a life lost is lost. You can never recover it. So with that kind of philosophy, with that kind of priority on safety, we know that we possess a very, very strong competitive edge. We have a very deep moat for our L4 business. Right? Those are my replies to your 2 excellent questions. Thank you.

Tianyu Lu analyst
#10

Just a single item from me. In what way does leadership perceive the broader direction of upcoming operating costs, especially the money going into R&D?

Xuan Li executive
#11

Tianyu, happy to address that. As we've just walked through the financials, the operating leverage is clearly starting to show. Revenue was up 87% during this quarter, while R&D rose only about 36%. Much of our R&D is shared — for instance, the technology platform underpinning L4, L2++, L3 and every other product line. That means it doesn't truly grow in proportion to revenue or to how many vehicles get deployed. Consequently, as we expand across various businesses, the R&D expense tied to each vehicle should keep falling. We've likewise gotten beyond the heaviest investment period for building out AI infrastructure, and those ongoing deployment expenses drop considerably once operations are up and running; so going forward, we intend to stay highly disciplined about R&D outlays and keep funding the technology approach — the foundation model, say, on Genesis, as Tony described. With our asset-light L4 operation expanding and L2+/L3 moving into large-scale commercialization, we anticipate operating leverage picking up speed. That keeps us squarely on course for positive cash flow within one quarter by 2028, and our goal is a full-year breakeven in 2029.

Mai Liu analyst
#12

Yes, the company just rolled out the WIP model. So, Tony, would you mind going into a bit more detail on that?

Xu Han executive
#13

All right, I'd be glad to say more regarding the WIP model, all right? As you can observe from the WIP model, essentially, the crucial element of the WIP model, I would say, is that it attempts to — it gives WeRide a distinctive analytical instrument for video, and when paired together with the Genesis model, we are able to — so to begin with, when handed a video, it contains numerous facts, and doing causal inference across separate videos simply isn't possible; it's rather like handing you a machine, and the WIP model assists you in detecting that machine and uncovering how its various components relate to one another; meanwhile Genesis grants you an additional ability to learn what the components do and assemble them so as to construct a different machine; so essentially, it's similar to giving you a LEGO toy of, perhaps, let's say, a LEGO toy of a Jeep, and then I can identify it and transform it into a cargo ship, all right? So by combining these two, we think we've discovered a highly distinctive route to examine hundreds of videos, billions of videos, and uncover the causal links among all the facts, after which we employ them to produce the pertinent data, or extremely long-tail data, for our autonomous driving. In doing so, we in fact enhance our training capacity and reduce our training expense while constructing the desired distilled onboard model in the urban challenge, such as when we captured 6 consecutive championships. Our onboard domain controller runs comparatively slower, and its top is far lower than that of our rivals. They might possess a 2,000 TOPS domain controller. We have only a 200 TOPS controller, yet we keep winning. Why? Because we can distill a model tailored to our particular requirements, and we can obtain long tail. So that's the strength of our tool foundation model. I'll pause there. Should you have additional questions, feel free to ask as well.

Ming-Hsun Lee analyst
#14

I've really got just a single item to ask. We've observed that the leadership keeps stressing your competitive strengths. Right now, opinions in the market regarding this industry are divided. Might management lay out what your competitive moat actually is?

Xu Han executive
#15

I'd like to get a clearer picture of where those divisions lie — could you give me a bit more detail? What sort of differing perspectives do you mean?

Ming-Hsun Lee analyst
#16

What I'm getting at is this — right now, in my view, the rivalry in this space keeps heating up. That's presumably why the capital market sees the competitive landscape differently. So I'd just like the management team to spell out more about what your competitive moat actually is.

Xu Han executive
#17

All right. Understood. All right. Thanks for clearing that up. I believe I just walked through the foundation model we built, correct, WIP and Genesis. And folks might attempt to pull off the very same thing, yet our objectives are identical. Still, just as everyone wishes to manufacture electric cars, whoever creates the finest motor, the finest battery, whoever nails the finest design aerodynamics will land huge orders. And when it comes to ADAS, there's a grand challenge, as I keep repeating, that we took 6 championships in a row. That's a challenge where we go up against every top-notch system company. Plus, drawing on our L4 capability, we run a really large-scale L4 driver lease operations fleet, and it runs globally too, in Europe and in China. We're able to collect every sort of data and sharpen our model. And I'd say we're the sole company on this planet with what's called a dual flywheel mechanism — pulling data off L4 robotaxi to help sharpen the L2/L3 ADAS system, and likewise taking L2, L3 ADAS data to help sharpen L4, since with WeRide design, they rest on the same foundation model built on the same — quite similar sensor configuration; we can employ the 2 things in tandem so they leverage one another and lift one another. So, given that, I'd say our competitive edge is one of a kind, and our results, our financial numbers, our competition numbers, our evaluation numbers already demonstrate that we truly sit in a leading position. Just a simple question: which other company has its founder, has its ADAS system doing a live stream video in the Tier 1 city for 2 hours with zero intervention, parking lot to parking lot. As I see it, I'm the only guy who tried it first. And next month, I'll give it another try. So, I'd say all the results make it pretty clear that WeRide has very — is a leader in both L2++ and L4, and our competitive edge is extremely strong.

Walter Patrick analyst
#18

Regarding the asset-light approach and the platform partnerships, in this setup, your partners are the ones who hold the vehicles, while the platforms essentially control the riders. So we're basically in the middle, licensing the driver, and that's what determines how lasting those relationships are — is that the crux of it? Then concerning Uber, a partner of yours in certain markets, clearly their history there, you had them cycling far more in Phoenix. They've just ended their serve robotics tie-up. So, I suppose could you just walk through what exclusivity or minimum volume commitments exist in the Uber markets; and under the asset-light model, where you don't possess the cars or the customers, what leverage do you ultimately have if the platform, Uber or someone else, chooses to send demand toward different partners?

Xu Han executive
#19

Jeniffer, would you like to handle this question, or would you prefer that I handle it?

Xuan Li executive
#20

Absolutely, I can handle that one. Regarding water, our approach uses an asset-light model, which lets us grow more quickly around the world. Wherever we enter, we aim to create an ecosystem alongside local partners and, naturally, platform partners too. That means we collaborate with a number of platform partners, and also with many local partners. For us, the market opportunity is enormous; we're truly just at the start; and we believe that establishing a sound ecosystem will lay a solid groundwork for everyone involved. When you examine actual deployment numbers worldwide, we're of the view that our robotaxi fleet is the biggest outside China and the U.S. among all AV competitors, and by a wide margin. Our strategic moat isn't merely putting assets in place and stopping there; essentially, wherever we operate, we sell the assets to local partners or fleet operators; and since we enjoy a BOM advantage originating from China, we intend to retain that edge. Even more crucially, WeRide's outstanding safety record is the reason we've been able to secure autonomous driving licenses in so many nations. As of now, we hold official autonomous driving licenses in 8 countries. Our operations span 12 countries. To my knowledge, no other company has an operation scale like that. For us, the duty — and this is also the central duty — is, sorry, to obtain the regulatory permit. That duty is correct; and it will be essential over the longer term. Anyone can join a platform, yet not everyone is able to obtain a regulatory permit. And with regulators, all decisions rest on merit. So an AV player that consistently maintains a strong safety record both locally and globally will certainly be able to obtain additional permits as the business grows. However, for those lacking such a strong record — indeed, certain competitors are receiving — they'll face questions — or get suspended by the regulator when safety accidents keep occurring; so we hold great confidence in our technological capability, our ability to deploy locally, our regulators, our communication, and our safety capability, and we hold great confidence in WeRide's safety record. With all of that in place, that's how we've established a strategic moat globally. I hope that addresses it.

Analyst analyst
#21

No, that really does help a lot, particularly regarding the regulators and which players hold leverage in the marketplace. I've got just one more question about the OEM side, and it's about Mercedes specifically. During this call, you described the L3 proof of concept as a sort of confirmation for the top OEMs; yet Mercedes is likewise an NVIDIA flagship partner running this autonomous program on the complete NVIDIA stack. So they aren't merely adopting NVIDIA's open model — clearly they're deploying the entire stack. Could you help us grasp what Mercedes — how this OEM, as an illustration, is assessing WeRide, something NVIDIA wouldn't already handle? Is this purely a China market distinction, where the NVIDIA stack wouldn't necessarily be embraced? Does it come down to cost? Is it once more about regulation? Is it that OEMs will want multiple suppliers for L3 or L4 or L2++, whichever it may be? And then I suppose the larger issue is, as time passes, assuming NVIDIA has plenty of capital and they mature, what will ultimately make your software prevail at a company such as Mercedes? Clearly, NVIDIA, with its open model and everything else it does, whether that's China, regulation or simply expense?

Xu Han executive
#22

Okay. Allow me to respond to that. To begin with, according to the contracts, we aren't permitted to disclose too many specifics. Still, I truly value this question. Internally I often joke, saying, look, if you examine the internal character of the company, we were engaged in many activities much like we speak. Previously, I kept hearing, for instance, Sam Edelman replying to questions from an India engineer, saying, okay, if you lack that kind of money, I can't recall the precise number; perhaps it's something like unless you possess USD 20 billion, never consider attempting a large language model. But consider what DeepSeek accomplished, okay? So I've always held that genius, that outstanding engineers, that brilliant minds can overcome capital, okay? That's the reason I love Silicon Valley. I have faith in humanity's creativity. Consequently, I believe WeRide is exactly this kind of company, one possessing tremendous creativity along with plenty of genius, and we shall construct the finest ADAS system, the finest robotaxi, both already demonstrated by the market. Consider the resources we invest in our company and consider the outcomes we've achieved thus far. I believe the response to this question is quite clear-cut, and I genuinely appreciate this excellent question. Thank you.

Xu Han executive
#23

Thank you very much. Bye.

Read the full transcript via the API

You're viewing the first half of this call. Get the complete WeRide Inc. transcript - plus 132,000+ transcripts from 7,000+ companies, speaker segments and full-text search - through the EarningsAPI REST API or hosted MCP server.

Get an API key View API docs →

For developers and AI pipelines

Programmatic access to WeRide Inc. earnings transcripts and 132,000+ others is available through the EarningsAPI REST API and the hosted MCP server. Quarterly plans from $162 - full transcripts, speaker segments, full-text search, and the /api/v1/transcripts/recent polling endpoint for ETL pipelines.