Xpeng (XPEV) Q2 2026 earnings call transcript

I will now turn the call over to your host, Mr. Alex Z, Head of Capital Markets of the company. Please go ahead, Alex.

Thank you. Hello everyone and welcome to EXPO's second quarter 2026 earnings conference call. Our financial and operating results were issued by our news file services early today and available online. You can also view the earnings price release by visiting the IR section of our website at IR. xiafeng. com. Participants on today's call from our management team will include our co-founder, chairman and CEO, Mr. He Xiaofeng, life chairman and president, Dr. Brian Wu, Vice President of Finance and Accounting, Mr. James Hu, and myself. Management will begin with prepared remarks and the call will conclude with a Q&A session. A webcast and replay of this conference call will be available on the IR section of our website.

Before we continue, please note that today's discussion will contain forward-looking statements made under the Safe Harbor provisions of the U. S. Private Securities Litigation Reform Act of 1995. Fore-looking statements involve adherent risk and uncertainties. As such, the company's results may be materially different from the views expressed today. Further information regarding these and other risks and uncertainties is included in the relevant public filings of the company as filed with the U. S. Securities and Exchange Commission. The company does not assume any obligation to update any formal notice statements except as required under applicable law. Please also note that EXPENG's earnings press release and this conference call include the disclosure of unordinated GAAP financial measures as well as unordinated non-GAAP financial measures.

EXPENG's earnings press release contains a reconciliation of the unordinated non-GAAP measures to the unordinated GAAP measures. I will now turn the call over to our co-founder, chairman, and CEO Mr. 何晓彭. Please go ahead.

Thank you. I believe that the GDVS will not only be able to use the GDVS program, but also to provide the GDVS program. I will now turn the call over to management for the XPENG INC. Thank you. The financing will provide ample capital to support the mass production and continued development of our advanced general purpose humanoid robots, IRAM.

We will accelerate our progress towards the chat GPT moment for physical AI, whilst bringing additional strategic resources to strengthen the robotics ecosystem and expand real-world applications. As a global leader in physical AI, XPENG will not only lead the large-scale deployment and commercialization of autonomous driving worldwide, but also build the world's most valuable humanoid robot company. Today, we're very happy to see that we have taken another important step towards that goal. Thank you. I have also taken up on the role of the CEO of our robotics business. Over the past 12 years, XPENG has remained committed to full-stack in-house R&D across both software and hardware, building a solid technological and commercial foundation for the physical AI era.

We are able to bring together and integrate the strengths and resources of the entire group. These include the supply chain, automotive-grade manufacturing capabilities, and global footprint developed through our automobile business, as well as the Turing AI chips, AI infrastructure, and world foundation models developed through our Ados business. By applying these capabilities to our robotic business, I believe that we can accelerate the mass production and commercialization of XPENG's humanoid robots.

I think it's possible to do this. In all of the companies in China, only in all of the companies are doing their own work. Because of this, the performance of the AIRN and the current performance of the world, I believe there is a difference in the quality. In many ways, we have the ability to do with the ability and the quality of the quality of the world.

We have been conducting our research and development in the area of robotics for more than eight years, And I have always believed that the technological challenges and level of innovation required for advanced general purpose humanoid robots are far greater than those for smart EVs by at least 20 times. To accomplish something that is this difficult, we need to have a broad and deep R&D and system integration capabilities across design and styling, hardware and chips, software and AI, data and control systems, and quality and manufacturing. Only then can we succeed. XPENG is the only robot company in China with such comprehensive in-house R&D capabilities across the entire technological stack. This is why XPENG IRON fundamentally stands apart from other humanoid robots currently on the market with differentiated capabilities across multiple areas.

Thank you. The Currently, our full in-house technology stack covers IRON's body, brain, cerebellum, data, and infrastructure. On the hardware's front, XPENG INC features the industry's most human-like form and design. XPENG pioneered the industry's first fully enclosed flexible lattice structure for IRON, combining aesthetic appeal with enhanced safety. With 76 degrees of freedom across the body and 21 degrees of freedom in each hand, both are at industry-leading levels.

XPENG has independently designed and developed an AI-native hardware platform and all core components specifically for embodied intelligence, including chips, controllers, motion modules, and dexterous hands. Leveraging our established smart EVRMD and manufacturing capabilities, we can achieve automotive great quality and build the capability to manufacture and deliver at scale.

We can also make sure that the public is in the public and high-stress. We will now turn the call over to management for a long time. In terms of intelligence, XPENG INC is powered by three Turing AI chips, delivering effective computing power of up to 2250 tops. With the industry's highest level of edge computing power, XPENG's physical AI foundation model runs directly on IRON, enabling it to autonomously perform complex work tasks, without petty operation whilst ensuring low latency inference and data security.

IRON's highly human-like hardware platform provides a natural advantage in scaling data. It allows us to maximize the reuse of behavioral data generated in people's everyday lives and rapidly adapt to a broad range of environments designed for humans. As XPENG IRAN moves ahead to mass production in real-world deployment, we will gain access to vast amounts of real-world and human demonstration data, accelerating the training and iteration of our AI models. In turn, continued improvements in model capabilities will allow IRON to enter more scenarios and generate more high-quality data, creating a flywheel across data, models, and applications that will accelerate IRON's evolution in the real world.

We will now turn the call over to management for the best-to-beam. And from the open room, the open room, the market, and the open room. And in the 2017, the open room will turn the call over to a million dollars. I think the high-speed I hope that in the market, in the market, in the market, and abroad, will bring us an important growth and increase in our technology.

XPENG INC combines an exceptionally human-like design, the most intelligent AI brain, and the highest standards of safety and quality. Only by doing so can IRAM become a trusted companion to people and truly become part of everyday work and life. We have recently achieved several major milestones in the development of the mass production version. Starting from September, we will unveil and demonstrate a series of distinctive capabilities. We plan to enter scaled production by year-end with initial commercial deployments in XPENG stores and campuses.

In 2027, XPENG ARAM will officially launch and begin large-scale deliveries in China and overseas to external customers in the retail and service sectors. Next year, monthly production capacities can rapidly ramp up to several thousand units in response to market demand. I believe the technological barriers to advanced general-purpose humanoid robots are exceptionally high, while the supply of high-quality humanoid robots remain limited. As a result, the lifetime revenue and gross profit contribution of each IRM, including hardware sales and recurring revenue from upgrades to its AI model capabilities, will be substantially higher than the current average setting price and gross profit per vehicle of our automotive business. I expect the commercialization of humanoid robots to scale rapidly in China and overseas following mass production, generating meaningful gross profit growth, supporting our investment in physical AI R&D, and further widening our technological lead.

Thank you. Good morning, and welcome to the PORTA. Thank you. Now, I would like to come back to our automotive business. In the second quarter, our vehicle deliveries reached 103,295 units, up 65% quarter over quarter, and we achieved year-over-year growth ahead of the broader industry. Despite industry-wide cost pressures, our operations remained resilient. Supported by our progress in the premium segment in international markets, companies' growth margin remained above 20% in the second quarter. Our tech-defined luxury flagship model GX stood out among a wave of large six-seat SUVs launched this year. Domestic deliveries exceeded 7,000 units in July, making it one of the top three models in China's NEV SUV segment priced above RMB 300,000.

Mona L03, the first SUV in the Mona series, became a breakout hit immediately after its launch, with orders setting a new record for any XPENG model. In the third quarter new uncancellable orders increased by 50 quarter over quarter to a record high Extreme weather and supply chain disruptions affected our pace of ramping up in the delivery Here, I would like to especially express my sincere appreciation to our customers for their patience. We have started two shift productions for the Mona L03 and are working closely with our supply chain partners, to accelerate the capacity ramp. I expect that the L03 deliveries will increase substantially over the coming months and continue to trend upward.

So, our 5th of the car will be in the 9th of the year, and will be in the 4th of the year. The success of the GX and the Mona L03 gives us more confidence in our upcoming models. We are translating our competitive strengths in the best-in-class intelligence and standout design into higher sales targets and stronger brand momentum.

Our flagship five-seat SUV, the G9L, will officially launch and begin delivery in September. The Mona L05 will also launch in China in the fourth quarter. With the launch of four brand new SUV models, we will cover all major SUV segments. We believe XPENG's deliveries to increase significantly in the fourth quarter, with monthly deliveries targeting more than 60,000 units. Thank you. I believe it will be in the future of the market. Our international business is the second growth engine for XPENG's automotive business, and also an important driver of improving profitability. Overseas quarterly deliveries exceeded 20,000 units for the first time in the second quarter, up 81% year-over-year. In the first half of the year, our international business accounted for more than 25% of total revenues.

Furthermore, our overseas operations boasted exceptional quality, with an average setting price of our exports exceeding €40,000, EUR, placing our per-value revenue and gross profit at the forefront of Chinese automakers expanding globally. Since its global launch in Munich in July, the Mona L03 has attracted significant attention and earned high praise from overseas consumers for its intelligent technology, distinctive styling, and spacious interior. I believe that Mona L03 will become XPENG's first major model to achieve leading sales across multiple international markets.

Overseas delivery of the Mona L03 are expected to begin in the fourth quarter, driving firm-wide quarterly overseas deliveries to exceed 40,000 units. In 2027, we will also introduce multiple star models, including expanded range EV models in overseas markets, further expanding our geographic coverage and market share.

We will now turn the call over to management for the market. We will see the results of the Q&A. And at the end of August, we will roll out a major upgrade to VLA 2.0. Once again, validating the scaling law in ADOS and delivering substantial improvements in both user experience and safety. With the new 6.3.0 major version, the number of parameters in the VLA 2.0 on-device model will increase by 3.5 times, putting its parameter count in order of magnitude above that of small models commonly used in the industry. The new version delivers a 300% improvement in perception sensitivity, introduces ultra-long horizon reasoning and predictive capabilities, and operates at an industry-leading frame rate. This enables the AI driver to see accurately, think ahead, and respond quickly.

The new version will also integrate ADAS and smart cockpit capabilities powered by VLA and VLM, bringing selected L4-level capabilities developed for XPENG RoboTaxi to our passenger vehicles. One example is voice-activated pullover parking. Users simply give a voice command, and VLA 2.0 will autonomously find a suitable roadside parking space and park the vehicle without exiting ADAS mode.

Thank you very much. I believe that the increase in the overall performance and the model of the model of the VLA after this year will be more powerful than the overall performance. We will also be able to exceed the overall performance of the VLA 2.0 and the latest others.

Recently, together with my colleagues, we tested the XPENG INC. It exposes VLA 2.0 and the latest others from a leading global peer in China, Europe, and North America, respectively. In my view, VLA 2.0 is already on par with the world's leading others on major roads. In narrow roads when negotiating, as well as in campuses and parking facilities, the user experience delivered by VLA 2.0 is even better. It can navigate directly to a parking space with both efficiency and safety. I believe that as we upgrade computing power and model capabilities, VLA 2.0 will develop even more powerful capabilities over the next several version upgrades, delivering an L4-level ADOS experience in mass-produced vehicles, surpassing peers, and establishing a generational lead.

At the same time, we all have to do a full-time job in the world. So, we have to do a full-time job in the world.

So, I am very excited to be here. Because the model is in almost no longer in the world. It is reached a high level in the world and in the world. I think this is the goal of the world and the world of all areas of the world. In the fall of the sea, we will be happy to have the first-time I will now turn the call over to management for the next year. Next year, we roll out VLA 2.0, bringing a safer, more comfortable, and more convenient driving experience to users worldwide.

After deploying in overseas markets, VLA 2.0 will compete directly with the world's leading others and become XPENG's defining competitive advantage for our global products. At the same time, we will actively explore new software-based business models, creating a positive cycle in which commercialization and technology developments reinforce each other.

We will now turn the call over to management for prepared remarks. We will now turn the call over to management for prepared remarks. As of now, our pre-installed, mass-produced, robotaxi powered by VLA 2.0 has completed more than 2,000 internal test orders in Guangzhou and validated the full end-to-end process for trial passenger operations, laying the groundwork for commercial operations. Recently, we have completed the development of our cloud remote takeover platform, and our goal is to begin passenger operations without a safety operator in the car next year. In 2027, XPENG will continue strengthening the technology and cost competitiveness of our robotaxi while partnering with leading domestic and international mobility platforms to expand our robotaxi business across key cities in China and around the world.

This will create greater commercial value through vehicle sales, technological services, and revenue sharing from operations. We think that the AI technology is not only about the technology, but it is also about the development of the AI technology and business. We want to improve the AI technology and business development. We have a team of the AI technology and business development. We are currently in the international and international cooperation. I hope that the tool in AI, the VAA, the model, and the logotip, and the technology that we can use can move forward to the world.

In the meantime, we believe that the large-scale application of physical AI requires more than technological breakthroughs. An open and collaborative technology and business ecosystem that creates a value for multiple participants is equally important. To accelerate the commercialization of physical AI, we recently established a group-level business development team within the group and are actively engaging with the partners in China and overseas to bring our industry-leading Turing AI chips, VLA 2.0, Robotaxi, and humanoid robot technologies to global markets more quickly. Thank you. We hope that we will be able to increase our customers' performance and our customers' performance. We will now turn the call over to management for our customers and customers. I believe XPENG is entering a period of accelerating momentum across multiple businesses in the second half of the year.

Both domestic and overseas vehicle deliveries are expected to reach new highs. We are also expected to be among the first globally to achieve a scaled mass production and commercialization of advanced general purpose humanoid robots. We also expected to be among the first to deploy advanced data technologies in overseas markets, and we will establish new business models around the physical AI ecosystem, creating greater value for customers and shareholders worldwide.

Thank you, everyone.

With that, I will now turn over the call to our VP of Finance, James, who will walk you through our financial performance for the second quarter of 2026. Thank you, Xiaofeng.

Now let me provide a brief overview of our financial results for the second quarter of 2026. I will reference RMB only in my discussion today, unless otherwise stated. Our total revenues were $19.74 billion for the second quarter of 2026, an increase of 8% year-over-year and an increase of 51.5% quote-over-quarter. Revenues from vehicle sales were $17.05 billion for the second quarter of 2026, an increase of 1% year-over-year and an increase of 55% quote-over-quarter. The quarter-over-quarter increase was mainly attributable to higher vehicle deliveries.

Revenues from services and others were $2.7 billion for the second quarter of 2026, representing an increase of 93.9% year-over-year and an increase of 32.6% quarter-over-quarter. The year-over-year and quarter-over-quarter increases were primarily attributable to the increased revenues from, first, technical R&D services rendered to the Volkswagen Group due to the successful achievements of certain key milestones, and secondly, parts and accessory sales. Gross margin was 20.7% for the second quarter of 2026, compared with 17.3% for the same period of 2025 and 20.6% for the first quarter of 2026.

Vehicle margin was 12.1% for the second quarter of 2026, compared with 14.3% for the same period of 2025 and 12.1% for the first quarter of 2026. The year-over-year decrease was primarily due to the production generation transition.

R&D expenses were $2.91 billion for the second quarter of 2026, representing an increase of 32.1% year-over-year and an increase of 0.3% quarter-over-quarter. The year-over-year increase was mainly due to higher expenses related to the development of new vehicle models and AI-related technologies as the company expanded its product portfolio to support the future growth.

SG&A expenses were $2.5 billion for the second quarter of 2026, representing an increase of 15.2% year-over-year and an increase of 32.5% quarter-over-quarter. The year increase was primarily due to higher marketing and advertising expenses The quarter increase was primarily due to the higher commission to the franchise stores and higher marketing and advertising expenses As a result of the foregoing, loss of operations was $1.14 billion for the second quarter of 2026, compared with $0.93 billion year-over-year and $1.87 billion quarter-over-quarter. Net loss was $1.34 billion for the second quarter of 2026, compared with net loss of $0.48 billion year-over-year and net loss of $1.7 billion quarter-over-quarter.

As of June 30, 2026, our cash position was $40.48 billion. To be mindful of the length of the earnings call, I would encourage listeners to refer to our earnings press release for more details on our second quarter 2026 financial results. This concludes our prepared remarks. We will now open the call to questions.

Thank you for your question. This is Xiaofeng speaking. In terms of robot capacities versus automobile capacities, from our perspective, we do think that there is quite a large difference between those two. I think that when it comes to the capacities of the supply chain of robots, it is rather broad and deep. But in our company, we emphasize on multiple areas of full-stack research and development. And I think when it comes to about the challenges of capacities, in the early stage, it is about quality, and in the later stage, it is about sales.

We will first from the market and the material market and the material market, and will be able to increase the supply of the network. So, according to the product development, the material and the supply of the product and the supply of the product, we will gradually increase the supply of the product. So, we will see the demand to the market and the market. In terms of mass production for 2026, we expect that by year end of 2026, we will see mass production at scale kicking in. For IRON's product, we believe that we will see the commercialization of the robot itself first starting with our stores. And in 2027, we will see that the commercialization will take place in the different areas of our own self-operated scenarios and rolling out as well as picking up the speed to external commercialization and user-case scenarios.

In terms of R&D, and I think that we are looking at starting from the mid and the second half of next year in 2027, we will pick up the R&D development, and the mass production units will be at multiple of several thousand units per month, and further picking up the speed. Thank you. I believe that these capabilities of real communication will see more in different areas of the industry and more efficient. Thank you.

One final part that I would like to supplement, which is that for the IRON robot deliveries, and this will mainly be rolled out in the areas of retail and services, both in China and abroad, and same as our automotive business. The delivery of our robots and sales will be authentic and genuine data and the figures that we will share. In terms of the quality of our robots, the services that it can provide, I believe that versus the other peers out there in the market when it comes to either the shopping assistance perspective or the intelligence level, we will definitely be better and stronger than the others, as well as to be able to be used in a more Thank you for your share.

My second question is about the unique economics and margin profile. So what is the estimated unit cost for the mass production of iron? And what go-to market pricing strategy does management intend to deploy? And what is the anticipated gross profit margin trajectory, especially after the full-scale ramp? That's my second question. Thank you.

Thank you. I will now turn the call over to management for management. Thank you for your question. With respect to the master-produced robots at hand, I think that we are looking at from the perspective of innovation. quality, capacity, and all of these for both our hardware and software. We're looking at doing the R&D research all in-house, and when it comes to the supply chain of all these parts, actually over 85% of the supply chain partners that we work with actually overlap with the existing supply chain partners for our automotive business. I believe the cost of our robots, IRON robots, competitiveness will definitely be leading in this area. At the moment, in terms of the pricing for robots in the market, generally speaking, it's about 2.5 to 3 times of the bomb material.

So, for IRON, given that this is a general purpose robot, there is a very limited supply in the market, and I do believe that for our growth margins of the hardware will definitely be better than the existing automotive business. In the meantime, not only that, we are relying on the sales of the hardware, there will also be sales of our different models and the software services, subscriptions, et cetera. These, we believe, will all bring in profits for our business.

对,以上就是我的对这个问题的答复,谢谢。 Those conclude my answers to your question, so thank you. Thank you.

So in our system, we will see, Thank you for your question. And in order to answer this question, in our industry, for instance, many people will say that in robots, generally all they need is one brain and one large model. That will be enough. Perhaps this is possible many years later, but I don't think that is viable as of now. In terms of the different large models, and there are different types. For instance, we have the superfast models, and those are operating at 100 frames per second or even several hundred frames per second. There are these medium-speed models, large models, which are operating at 10 frames to 20 frames per second. And there are also the slow ones and slow large models.

We call them the thinking large models. They operate at one frame per second. When it comes to the VLA and VLM, and for instance, those are the ones adopted in our automated business. I believe that there are similarities, and for instance, the currently VLA adopted in the automotive business when it comes to the roaming in non-planned roads, that will be quite similar to the roaming or moving around of the robots. For robots, on the other hand, I also believe that some of the thinking capabilities of robots, such as on the open platform, next year can also be put into use for our XPENG's automotive business. So you can see that there are definitely some synergistic commonalities there.

In the meantime, there are also some unique perspectives and points of the robots, such as the different modes for safety. And for instance, there are the models of safety, such as data privacy and safety about prevention of folding, and data safety about the lack of running out of electricity, etc. So as you can see, there are commonalities and all of these features we are developing under the large XPENG's system altogether. And even so, if we look at the further underlying system, and there are many other areas that are quite dissimilar. For instance, the generation of the model, generative models, as well as the mimicking and the simulation models, etc. So those are the similarities.

Thank you. I will now turn the call over to management for prepared remarks. of the overall architecture and structure, and these are also the ones that we do share across the two different parts of the business.

That's my second question. If it's with other software companies, then the small companies in the data collection, training, and the switchboard, etc. What differentiated advantages does XPENG have in the past? Robot data collection training and closed loop iteration.

It's a very important thing. We have 10 years of electric cars, and we have been making our research. We will now turn the call over to management for prepared remarks. I think that the value of low-duty data today will make the whole speed of the speed of the data. We need to use high-duty data to create the data-based method of using the data-based method of using the data-based method of using the data-based method. Thank you. Thank you. That is a great question. And I do believe that in terms of the physical AI and in the future, and data is of course the key. Many people say that as long as you have enough data and that will help with the integration of the services It is a necessary condition However it is not yet the full condition And what we would see is that for XPENG and what we are good at is that we have much better data and the training of the data we also have a higher quality of the data For XPENG, for instance, we have been in the area of autonomous driving for over 10 years.

Within not only our R&D for the past over a decade of experience, as well as the data that we have collected, we are absolutely leading in the industry among our peers. In terms of the robots, it is the same. For the robots and management, the robots data management, data training, and data quality, all of these are being developed under the same ecosystem at our company. And I believe that apart from the hardware being different in terms of the collection of the data and to all the way to the application of the data at our company for our two different lines of business, it would be the same. For IRAN, again, once these products become mass-produced and launched into the market, not only that, we will be further continuing to collect real-world data, as well as the human-demonstrated data, which are both sets of high-quality data, and this will further drive the development and R&D of our products.

And this is different from the low-quality data, which are not helpful at all. So I believe that by having all of these, we'll actually be able to create a flywheel of the high-quality data and R&D that we are conducting and continue to contribute to the development of our products. That's all my answer for your question. Thank you.

Thank you, Xiaopong. That's all my question.

Thank you. We will also turn the call over to management for the XPENG INC. So, the first thing I think is that the most important thing is that we have to put our machine in the whole network as four big things. We think that in China and abroad, the system of trade and the world is the most important thing to us. We will open up our island, and we will open up our SDK, and support the development of the app. So these are the things we will do next. So through the business of business, we will see that the company can share the whole entire network, and will create a new online sales. So our customers are looking for the direct-to-line and direct-to-line customers.

I believe this will be a very valuable, new and new and new and new companies. Thank you. Thank you very much for your question. For XPENG's robots, yes, indeed. When it comes to commercialization, we have gone down a different route versus other competitors. Many other competitors, they are focusing on breaking into the market by ways of entering into factories, home usage, and mainly for to-be business. What we are looking at is that we are focusing on the large-scale seats as well as the small and medium bees, and that is, so to speak, we start entering into the market with business commercialization cases, followed by industries and home uses at a later stage with smaller SKUs.

The reason we have picked the salespersons and tour guides, etc. , it is because that we believe both in China and abroad, and there are four major comprehensive capabilities of our robots that are very helpful and would be able to be reflected very well in these sectors. And the four comprehensive capabilities are as follows. Number one is the main body and the main hardware of the robots itself. Number two, the environment. Number three, the business that it provides. And number four, the emotional values that it brings. Therefore, starting from commercial usage cases and starting from the smaller type of business and the smaller and median type of business industries that we go in. And later on with IRON, of course, when it starts opening up the market, and we will also open up SDKs to enable secondary development as well as further expand its user cases.

With these commercial use scenarios for collaboration, and we believe that this will also open up more channels for XPENG's robot business, not only for offline sales, but as well as for online sales, so that our customers would be able to see the use cases for our robots, and not only for the big customers, but as well as for the small and medium business. So this is our thinking in the regard and which is different from our peers.

The second question is about the latest progress on the company's self-developed death stress hand. What overall design approach has been adopted, and how does it compare with peers in terms of performance and cost? Thank you, 小分钟.

Thank you. We have 21 degrees. Our size and size are the same. We believe that our team is the same. We are not using the business of management. We are not using the business of management. We don't think it's a good choice. So we don't need to use high-quality energy or high-quality energy. We need to use more and more. We even need to use the power of management for management. We need to use the power of management for management and quality. And we hope to bring the digital access to the management of the management of the system. Thank you.

Thank you very much for your question. And yes, indeed, the dexterous hand is an extremely important part for robots. And for our robots, we have one set of hardware, one set of software, as well as three sets of different perceptive systems. In terms of the specific mass production plans, we will be communicating with the analysts by year end. We are not only just conducting the R&D of dexterous hands, and we have also invested greatly into the manufacturing of the processing of dexterous hands, as well as the equipment in this regard. The dexterous hand for our robot, we have 21 degrees of freedom, as mentioned earlier. In terms of the size of the product of the dexterous hand, it is the exact same size as the hand of an adult.

And when it comes to the load-bearing capabilities as well as grasping and gripping, we believe that our dexterous hand is in the leading position versus the other peers. I know that in the industry that people often talk about another type of hand, and we do not think that is the smart choice to go for. We are not focusing on the force or the accuracy itself of Dexter's hand, and what we are focusing on rather is on striking a good balance of safety, reliability, easy to maintain, and cost. And for instance, our Dexter's hand also has a very nice set of skin that is very similar to the human hand.

Your next question comes from Nick Lay with JP Morgan.

Thank you. Thank you very much for your question. Given that the signal wasn't coming through very clearly, and I could only hear some keywords of your questions, so I'll try my best to answer your question based on what I have heard. Number one, with respect to the Turing AI chips, hence this has already been deployed to our L03 vehicles for the one start, which are launched to the overseas market as well, and all of these will have the ultra version. For VLA, this will be deployed in L03 models across all different markets. And in the meantime, we're also catching up and working with the compliance and local laws and regulations and localization work, testing. All of these are being done at the same time in tandem.

And in the meantime, we are looking at, for instance, about the subscription service of the software for our customers, and such updates will be announced and shared with you all in due course. We have also established another BD team, and this team is actively discussing with our partners, and with respect to the VLA usage, or even further expanded into other areas. Thank you.

Hey, Tina, it's Brian. Let me address your question. With regard to the financial outlook of the robotic business, I think it's a bit too early for us to comment. I would say we are now focused on the milestones that Xiaofeng shared, which is reach SOP for our robots for volume production capability by the end of this year. And also start deploying first in our internal scenarios and gradually offer to external customers starting, I would say, the first half of next year and gradually ramp up from that.

So that's our goal. I would say it's too early for us to provide a volume prediction guidance. On the probability, we anticipate the product of a human and robot will achieve much higher gross profit potential compared to the automotive business. In fact, I would say the hardware is already much higher than the automotive business. In addition to that, we think there will be significant opportunities to add on future AI model training, upgrade, software capability related revenues, which is much higher margin as well. So given the high profitability, I would say, expectation, as well as I would say much smaller investment and capex requirement for robotic business, I would say once volume rampant is achieved, I would say probability can become reality much faster than the auto business.

So that's our projection. And also to answer your question regarding the potential separation of the businesses, at the moment, actually, the businesses are operating together. In fact, we have not started any separation of the business. We obviously, according to the announcement you saw, we actually have a period of 18 months that gradually allows us to achieve a separation.

But in the meantime, what we've got to be focused on is still achieving high degrees of synergy because we talked about leveraging the capabilities in AI, the capability of advanced manufacturing, powertrain, supply chain, so forth. So actually the two businesses can both achieve high efficiency and greater capabilities.

So with that, I think the near-term expectation is it will still be mostly viewed as together as business. And also as the volume production and also as the commercialization scenarios become more clear, we'll probably think about more likely separations. But in any event, given the ownership structure, this business will be 100% consolidated. It will not impact our financials going forward, even though the business may start to separate based on the plan. So I think in short, we still see the group consolidating all the financials of the robotic business, and at the same time, we'll really think about the most efficient and also most synergistic way to run the business.

Thank you.

Thank you once again for joining us today. If you have further questions, please feel free to contact the XPENG's IR team through the contact information provided on our website or the piece of financial communications. This concludes today's conference call. You may now disconnect your line. Thank you.