Written by | Xu Zhi, Edited by | Wu Xianzhi
Following the "9-Series Battle," large five-seat flagships have once again become a new battleground for domestic automakers. On September 17, XPeng also presented its answer: the G9L.
After the G9L launch event, Photon Planet and others held a dialogue with He Xiaopeng, Chairman of XPeng, and Wu Anfei, General Manager of the Product Matrix.
On the surface, the discussion was about the G9L, but behind it lies XPeng's steadily growing capabilities. Long-termism, technological luxury, and global layout are the most core keywords. He Xiaopeng even shared further insights on the robotics business, which has recently attracted high attention from the outside world.
For flagship large five-seat models, the usual practice of competitors is to directly modify the large three-row flagship version. However, the XPeng G9L and GX are completely two different concepts, and it is also a completely different model from the G9. To this end, XPeng has made brand-new treatments in terms of design, handling, safety, and other aspects. This perhaps also explains why XPeng is a relatively late player to "submit its paper" in the large five-seat product segment.
As one of the earliest players to gain a firm foothold in the European auto market, XPeng's globalization is further accelerating. Starting with the MONA L03 a few months ago, XPeng's new vehicles have gradually stepped into the rhythm of global synchronization. The G9L will make its global debut at the Paris Motor Show on October 12 and be launched simultaneously in 64 markets worldwide.
Supporting XPeng's global expansion, besides infrastructure such as production lines, sales outlets, and supercharging networks, technological capabilities represented by intelligent driving are becoming the most powerful engine.
Not long ago, the bet between He Xiaopeng and Liu Xianming officially concluded, with Liu Xianming winning as expected. XPeng's colleagues in Silicon Valley are about to welcome their own cafeteria. He Xiaopeng also stated in this dialogue that, even compared to the full-feature version of Tesla's FSD, the second-generation VLA has achieved superiority in many aspects. "In the next two years, the world's most advanced intelligent assisted driving will definitely include XPeng."
Compared to domestic embodied AI unicorns, XPeng's robotics pace has been relatively slow in the past, with very few public appearances. This stems from XPeng's proactive restraint. He Xiaopeng believes that rather than frequently releasing capability demonstrations that cannot be mass-produced, it is better to wait until the robot has sufficient certainty before promoting it.
Currently, XPeng's robotics is still in a "very painful" stage. Hardware, software, and design capabilities are all under intensive research and development, especially the supply chain. The existing supply chain is hard to meet the needs of XPeng's robots, so most of them have to be independently researched and built.
The root of the pain is that XPeng has always been committed to building a "completely different" robot. This is not only reflected in the ultimate anthropomorphism but also in its integration of production tools, productivity, and production relations, rather than "building a hammer first and then looking for nails" like most competitors.
The following is an excerpt from the dialogue. Some parts have been edited and streamlined without affecting the core viewpoints and expression logic:
No Shortcuts, No Quick Success
Q: With 9-Series configurations at a 6-Series price, what is the rationale behind such a large-scale configuration trickle-down for the G9L?
Wu Anfei: Over three years ago, we hoped to design two flagship vehicles. The fundamental consideration was customer demand. For example, real customers wanted the experience of a 9-Series flagship but only needed a five-seat vehicle. From the very beginning of the design, we wanted a large five-seat car of the 9-Series.
We wanted it to be more than simply shortening a six-seater; we aimed to redesign a completely new vehicle with a golden ratio. Therefore, the initial design of the G9L was not a six-to-five conversion, but a brand-new, native five-seat concept, which led to the 0.605 golden ratio design.
At the same time, we considered it a global vehicle from the start, featuring both left-hand drive and right-hand drive, as well as two-wheel drive and four-wheel drive. In terms of powertrain, it offers both pure electric and extended-range options, because charging infrastructure is not that well-developed in some countries.
He Xiaopeng: It's a bit different from the traditional six-to-five conversion. I think a five-seat car must first be good to drive, and driving with a golden ratio is a very enjoyable experience, whether it's the P7 or the G9L. Friends in this price segment mainly buy it for their own driving pleasure, occasionally carrying family and friends, so we actually implanted our own ideas into it.
Q: Why not just make it a five-seat version of the GX?
Wu Anfei: If we heavily modified the GX, the investment cost would be lower, but it would only change the number of seats, and handling might suffer some compromises. We didn't want to take shortcuts. We wanted to design a globally adaptable vehicle that natively balances space and sportiness, so we abandoned the mature, simple, and low-investment model.
Q: In the face of the current phenomenon of quick-to-market cars, the G9L took 3 years, crossed 26 countries, and completed 6.74 million kilometers of global validation. What is the strategic consideration behind this long cycle and massive mileage?
Wu Anfei: When we developed a car over three years ago, we kept raising the lower limits in terms of durability and safety. Developing a global vehicle requires considering tests in various scenarios such as extreme cold and extreme heat. We deployed many more vehicles compared to other automakers.
Including conducting corresponding tests in right-hand drive countries. Therefore, in the matrix logic of our product development, we hope to keep raising the lower limit of development reliability. Regarding the development cycle, we made the development validation more sufficient. Compared to an ordinary Chinese car, our biggest difference is that the validation is more comprehensive.
Breaking Out of Hyper-competition, Going Global
Q: What will be the characteristics and core competitiveness of XPeng's overseas expansion? How to convince them to accept Chinese luxury flagship models?
Wu Anfei: We rank first in the segment in multiple countries including Germany. In regions like Europe, where the auto industry was originally very developed, we have already taken the stage.
Then, a flagship car like the G9L has many inherent advantages. First is the intelligent driving capability. The G9L is now equipped with the second-generation VLA and is being tested in Europe. In this aspect, we are already ahead of the vast majority of enterprises.
Second is the 9-Series flagship configuration. Most European luxury brand cars either require extra payment for optional features or don't have them at all. In terms of configuration, we have no rivals globally.
Third is providing dual powertrains, with both pure electric and extended-range options. In countries with highly developed charging infrastructure like Europe, pure electric is completely fine; for places where charging infrastructure is not well-developed, the extended-range version also provides a choice. In the Middle East and other places, the GX has already received over 3,000 orders, mainly for the extended-range version.
Q: Is intelligent driving the advantage XPeng can build now?
He Xiaopeng: This time, the XOS 6.3.0 version of XPeng's second-generation VLA may not be a huge leap for the user level, but it is a massive leap at the technical level. I think the performance on main roads is consistent with FSD, but the performance on minor roads clearly exceeds FSD, as far as I can see. Many countries in Europe and Australia have the full-feature version of FSD. We simply took the current 6.3 version to run locally and found that we are much better than them.
XPeng overseas might just need a little bit of local data for post-training to perform very well; this is a basic generalization capability. The combination of our VLA and VLM capabilities will build a completely different core differentiation. Now L2 and L3 have been defined, and the whole of Europe and other countries will gradually open up in tiers after the end of March this year. XPeng is also one of the core committee members. Imagine that in the next year or two, the world's most advanced intelligent assisted driving will definitely include XPeng. I am very confident.
Q: Does the G9L have plans for overseas production lines? What specific adjustments have been made regarding overseas adaptation and compliance issues?
Wu Anfei: The G9L will go on sale in Europe starting October 12 at the Paris Motor Show. Besides Paris, we are deploying in over 60 countries globally. They will be launched and start delivery one after another from October to December. XPeng now has three manufacturing bases globally, with factories in Graz, Indonesia, and Malaysia. Currently, the first step for the G9L will be produced at the Graz factory in Europe to complete the mass production action. The overall pace of going overseas is completely synchronized with the domestic market.
As for how many regulations to adapt to and how many changes to make for going overseas, first of all, we are integrated in terms of manufacturing quality. The domestic supply chain, manufacturing processes, and inspection processes share the same roots and sources globally. Of course, different countries will also adapt to their regulations, including adapting energy forms, seats, and so on. There are also some specific needs in Europe, such as the tow bar seen today, which is for many European users who want to install bicycles or travel.
He Xiaopeng: XPeng has three core capabilities that were not reflected overseas before, and all of them will be manifested in 2027. First is intelligent driving. There was no regulatory support before, but we will launch VLA globally next year. Second is voice dialogue without data traffic. There are many voice languages overseas, and data traffic fees are also very high. Third is super-fast charging. From 2027 to 2028, XPeng's super energy storage charging stations will begin large-scale construction globally.
Q: What supports the mid-to-high-end positioning overseas, and what is the specific strategy?
He Xiaopeng: Relying on technological differentiation. If you ask in Europe whether they know XPeng, surely quite a few people know, and many consider it mid-to-high-end and high-end. The G9 is sold for 80,000 to 100,000 euros in Europe and is the number one in the pure electric mid-to-large SUV category in multiple countries.
This is also why the G9L and MONA are both European-spec versions and can go directly overseas to do a good job in technological differentiation. Friends in Europe, Southeast Asia, and Mexico are quite looking forward to a different car, not competing on price, not playing with low configurations, but relying on technological differentiation.
The "Painful yet Joyful" Robotics Business
Q: After robotics exploded in popularity last year, you mentioned not wanting it to have too much exposure. Is it because you want to make a stunning debut, or wait until a new cycle to release it?
He Xiaopeng: The R&D of robots is very difficult. Making a good robot is equivalent to about 20 times the difficulty of making a car. Therefore, our hardware, software, design, and supply chain are basically all independently researched and built.
In the past, we were always unwilling to release the robot because the difficulty was too great, and we were not sure when we could achieve a steady launch. Today, there are many videos of robots outside, but they are actually released after editing. I think it's meaningless. They ultimately cannot be mass-produced; it's just a demonstration of reaching this capability.
Just like there are many robot companies in the world working on the "brain," constantly saying their brain has reached a new level. Why don't autonomous driving companies globally talk about their scientific research progress? Because autonomous driving has already been mass-produced. If it's only scientific research without mass production, there is no value.
Using the auto business as an analogy, the robot today might still be at the level of not having produced the first car. This year is roughly equivalent to the XPeng G3 in 2018. The G3 was released at the end of December that year, and deliveries only started in April the following year. At the end of December 2026, our goal is to mass-produce the robot. The target for the second quarter of next year is to start being able to sell in China, followed by global sales.
Our robot, from scientific research to engineering, then to product, and then to mass production, has entered a very painful but also joyful rhythm period. So we are just starting to warm up now. By the fourth quarter or next year, everyone will see a leap in our robot's capabilities.
Q: What are the differences between XPeng's robots and those of peers?
He Xiaopeng: Our robot is different from most robot companies. They make a hammer, a production tool, and have to look for where the nails are, help hammer a few times, and then charge.
We believe the robot is a trinity of production tools, productivity, and production relations. There is no need to look for scenarios. It can start in sales stores, where it can also sell cars while selling itself. It is a completely different product, very hard to make but very interesting.
Q: Bringing the robot truly into its own production system, what does this step mean for the commercialization of XPeng's robots?
He Xiaopeng: As mentioned just now, robots are particularly difficult. Hardware capability, software capability, and design capability are all hard. Moreover, every detailed change in the robot may change many related areas and may affect the layout parameter capabilities.
Even in unexpected places, there are related impacts. For example, the camera installed on the robot's head, how should the angle of this head camera be set? A human's eyes can automatically focus, looking near and far, but for a robot, it's a problem.
We encountered a major problem with the robot some time ago: shaking hands. Humans don't need to bow their heads when shaking hands because they need to look into the other person's eyes. If the robot's eyes are not configured well, it needs to bow its head slightly, which will restrict all hand movements, foot movements, waist movements, and the whole body's motion control and the whole body's VLA system.
I think the vast majority of robot companies have seriously underestimated the complexity of hardware. Just like when we do autonomous assisted driving, without an OK car, chassis, computing power, and wiring harness, it's impossible to do it. XPeng has spent a long time thinking in the past: what should the hardware be like? Where is the production line? Where is the entire supply chain? Many things we are doing now do not exist in the industry.
At the recent WAIC and WRC robot exhibitions, many robots have mechanical hands and mechanical arms, but those elbow joints are not only thick, but generally have one or two bare wires exposed outside. What we want to make is thin, safe, durable, and powerful. Basically, all external supply chains cannot meet the requirements, and everything has to be redone. That's why it's said to be a very difficult step.
Mass production is the real integration and landing of the production line. Moving from small-scale integration to the path of mass production is a very critical big step. I also look forward to the next few years, hoping that everyone can see that robots can help us in business, industry, and family.