XPeng (Minutes): Expected sales to more than double in 2025, profitability in the fourth quarter
The following is the minutes of the 4Q24 earnings call for $XPeng(XPEV.US) . For the earnings interpretation, please refer to XPeng: Did the guidance collapse? Can XPeng still turn things around? .
I. Review of Core Financial Information
II. Detailed Content of the Earnings Call
2.1 Key Information from Executive Statements
1. Development of Intelligent Driving Technology:
In 2025, AI large models will bring a disruptive experience to autonomous driving. In the first three months of this year, we have seen breakthrough progress in AI technology, which reinforces my confidence in XPeng's strategic and technical direction regarding AI large models in the physical world. We have trained cloud-based foundational models with hundreds of billions of parameters, utilizing over a hundred million kilometers of real driving data for cloud training, which is unprecedented in the field of autonomous driving in China in terms of model parameters and data volume. On this basis, we enhance the model's generalization ability and capability to handle long-tail scenarios through reinforcement learning, and deploy distillation, pruning, and quantization techniques in the vehicle-side model, resulting in more than double the accuracy of the entire model.
In the second half of 2025, XPeng is expected to be the first in China to provide intelligent driving experiences equivalent to L3 autonomous driving in terms of software capabilities and user experience; in 2026, we plan to mass-produce models that support L4 autonomous driving in low-speed scenarios. Starting in 2025, the penetration rate of urban intelligent driving among new car buyers in China will exceed 10% for the first time, showing a non-linear growth trend. XPeng is well-prepared in terms of technology, cost, and model lineup.
2. Domestic Product Layout: Starting from the P7+, all new products from XPeng, including the newly launched 2025 models G6 and G9, will be equipped with industry-leading pure vision intelligent driving as standard, without the need for optional purchase.
The upcoming MONA M03 max model will also lower the threshold for urban intelligent driving to the level of 150,000, allowing more young users to easily enjoy the fun brought by technology. I believe that by integrating the most advanced AI technology with extreme platform efficiency, XPeng can achieve AI equity across the widest price range, making it affordable for a broad user baseIt's also doing well.
In 2025, we will have significant new models and facelifts launched every quarter, including the latest generation of AI technology integrated into the new model from XPeng. On March 13, the 2025 G6 and G9, which have undergone a comprehensive upgrade, have been well received by users, with orders continuously rising, exceeding our original plans. We are also very pleased that the explosive increase in foot traffic after the new models were launched has begun to drive significant growth in demand for our entire range of models.
We plan to start deliveries of the new G6 on March 21, and we are working closely with our supply chain to enhance production capacity, allowing more users to pick up their cars as soon as possible. Following this, models including XPeng's MONA M03, G7, and X9 will see the launch of their max versions or facelifts. I believe that through a series of combinations, including M03, G6, P7+, G7, G9, and X9, we will form a strong product lineup that will comprehensively cover the diverse needs of young users and family users looking for quality upgrades, allowing us to significantly increase our market share across multiple segments.
As we ramp up deliveries of new and facelifted models, I believe XPeng's sales will steadily transition through the product upgrade period in the second quarter, continuing to grow and reach new heights. In the second half of this year, we also plan to deliver several new products, including pure electric and Kunpeng super electric products. The vast majority of our models will support dual energy, significantly expanding our reachable market size through super electric products. By the end of 2026, XPeng will span the price range of 100,000 to 500,000, with a more complete product layout across all mainstream segments from compact to large vehicles.
In addition, our new model development will adhere to extreme platformization, integrating multiple underlying technologies such as AI driving, AI cockpit, electronic electrical architecture, and AI power to achieve higher efficiency in R&D, iteration, quality, and stronger scale effects.
As AI intelligence accelerates, over the next three years, XPeng will focus on refining and differentiating based on clearly defined product technologies and global cooperation. Through this three-year product cycle, we aim to achieve rapid sales growth, leading scale, and sustainable profitability.
3. Overseas Market Expansion: In 2024, XPeng's overseas vehicle sales will exceed 20,000 units, ranking first among emerging Chinese electric vehicle brands in terms of export volume, and also leading in the export of mid-to-high-end electric vehicles. The overseas business goals for 2025 include doubling sales, establishing over 300 sales and service outlets globally; setting up new overseas R&D centers, and conducting overseas research and testing for Turing AI intelligent driving. The goal is to bring XPeng's AI intelligent driving capabilities to more countries and users by 2026 and beyond, ensuring compliance with local regulations.
4. Chip R&D Progress: We expect to start mass production of our self-developed Turing chip this year. This chip is suitable for cars, flying cars, and humanoid robots, designed for rapid deployment in global markets and diverse scenarios. The Turing chip, combined with XPeng's self-developed software and hardware, will allow XPeng to fully leverage vehicle scale advantages, with a model architecture that leads the industry by a generation, and the effective parameters used by vehicles being dozens of times greater than those of competitors5. Strategic Growth Direction: XPeng's strategic plan has three main growth trajectories. First, focus on AI-enabled vehicles, equipping cars with cutting-edge AI technology; second, expand into global markets, actively exploring international markets outside of China; third, commit to the research and development of humanoid robots to seamlessly integrate with the automotive industry.
In January of this year, deepseek R1 achieved significant success with its AI model in the digital world. Interestingly, this aligns with XPeng's technological roadmap for autonomous driving from multiple angles. We see large models merging across different industries, and XPeng is also exploring the boundaries of these technologies in collaboration with outstanding AI technology companies from various fields. This excites me greatly, and I believe AI will drive significant transformations in the automotive industry over the next few years, including accelerating the realization of L3 and even L4 autonomous driving. At the same time, I see smart cars developing in conjunction with embodied intelligence; smart cars are likely to be precursors to humanoid robots, and future smart cars will certainly also be vehicles with embodied intelligence. My goal for this decade is to transform XPeng into a global automotive company, becoming a beloved smart technology brand among global users.
6. Sales and Profit Goals for 2025:
By launching more competitively leading products and expanding globally, I believe XPeng's total sales in 2025 will achieve more than double the growth compared to 2024, significantly expanding our market share in the smart electric vehicle industry in both China and globally, becoming a widely recognized AI automotive leading brand among consumers. Meanwhile, through the continuous improvement of our overall operational capabilities, both past and present, we will continue to reduce costs and achieve greater economies of scale. We expect the automotive gross margin to continue improving this year. We are very confident that while we persist in significant investments in AI research and development, we will steadily move towards profitability for XPeng in the fourth quarter of this year.
2.2 Q&A
Q: As XPeng achieves the popularization of artificial intelligence in China, will autonomous driving or assisted driving functions become standard, and with the upcoming upgrade to L3 autonomous driving systems, will this really widen the gap between leaders and laggards, or will competition remain in a stalemate?
A: In the past few months, we have witnessed a significant transformation in the automotive industry, with many companies announcing their entry into the autonomous driving field. This transformation is profound, as the consensus on viewing new energy as the future development direction has been reached among automakers. This is not only a revolution against traditional fuel vehicles but also a tremendous advancement for the entire industry. Looking back at 2020 to 2021, many automakers had already recognized the importance of new energy, and now we stand at a new starting point, witnessing the industry's rapid advancement towards autonomous driving. The development potential of autonomous driving technology is enormous; it not only brings significant value to users in high-speed driving and parking scenarios but is also advancing towards full-domain, point-to-point autonomous driving from parking space to parking space. This progress will bring unprecedented convenience and value to users. Last year, XPeng launched its AI intelligent assisted driving system, including full-scenario intelligent assisted driving functions. We believe that this series of innovations will endow future AI cars with more unique capabilities, such as human-like driving intelligenceLooking ahead, Level 3 autonomous driving will become a key turning point. This is not only due to the enhancement of technology, including significant improvements in hardware, software, and model capabilities, but also because the market's recognition of the value of high-level autonomous driving is reaching a critical point.
When 10% of users in the market begin to consider high-level autonomous driving as an important factor in their car purchasing decisions, it will trigger profound changes in the industry. XPeng has already prepared in multiple ways to differentiate itself from the industry in the next phase. First, XPeng has made significant investments in full-stack self-research capabilities, forming an end-to-end closed-loop capability from vehicle hardware to chips, cost control, and engineering implementation. This capability allows us to utilize data more efficiently and achieve rapid iteration. Second, the horizontal integration of autonomous driving with cockpit systems, chassis control, and power systems is generating significant synergies, and self-research companies have a natural advantage in this regard. Third, the unity achieved on the AI automotive platform allows us to far exceed our competitors in terms of rapid data iteration capabilities. Finally, from the very beginning, XPeng has taken into account the characteristics of the Chinese market and the efficiency and cost of globalization, providing us with a first-mover advantage and brand value in entering international markets. I believe that as high-level autonomous driving technology develops, leading companies will continue to widen the gap with laggards.
Q: How do you view the investment synergy of XPeng in artificial intelligence applications (such as autonomous vehicles, humanoid robots, and electronic pets)? When can we see these projects contribute value to automotive products? Will these projects lead to significant increases in R&D and other operating expenses in the coming quarters?
A: The company's long-term vision is to become an explorer of future mobility. More deeply, we are actually exploring the ecosystem of future mobility, which means we hope to combine energy with AI to create a new way of transportation. This will not only include cars but also flying cars or humanoid robots. Over the past decade, we have invested not only in smart driving cars but also in humanoid robots. In the next ten to twenty years, we believe that a truly intelligent or AI-enabled automotive manufacturer will definitely produce or manufacture embodied intelligent new mobility tools or automotive manufacturers. This includes cars, flying cars, or humanoid robots.
In this field, if you have very early preparations, you can definitely create a lot of synergies across different areas. For example, we are currently developing Level 3 capabilities for autonomous vehicles. This physical world AI model is certainly compatible with the Level 3 capabilities of humanoid robots. The only difference is that the data they collect to train the models is different. However, in many other areas and aspects, there are many capabilities that can be shared. Additionally, when we look at other aspects of end-to-end AI, global sales networks, and distribution networks, there are also many synergies between different types of mobility tools when we consider manufacturing capabilities and production capacity networks. I believe that by 2026, you will see the outstanding results of our synergies. As we mentioned in our preparation speech earlier, we believe that by 2026, we can actually mass-produce flying cars. By then, you will be able to see that XPeng will be the first in China, if not the first in the world, to actually mass-produce flying carsQ: More and more automotive companies are adopting end-to-end models to develop autonomous driving systems. Will the technological gap between leading and lagging autonomous driving companies narrow? XPeng started using self-developed chips this year. Besides cost control, how does this chip enhance XPeng's artificial intelligence autonomous driving capabilities?
A: In the era of advanced autonomous driving or intelligent driving, the competitive landscape of the automotive industry has changed dramatically. Any participant wishing to stand out in this competition must possess comprehensive full-stack self-research capabilities. This not only means mastering software technology but also requires hardware development and supply chain integration capabilities to establish close cooperation with top suppliers. Companies with abundant resources and comprehensive technical strength will have a greater advantage in this competition, and the gap with their competitors will further widen. More critically, we have a strong data closed-loop iteration capability, allowing us to efficiently utilize massive amounts of data to continuously optimize our products. Combined with our current robust profitability and significant economies of scale, I believe this will make us stronger in market competition than ever before. In addition, our self-developed chips not only have the capability for practical deployment in different markets and various scenarios but also bring us customization advantages and cost benefits, further consolidating our leading position. The significance of self-developed chips goes far beyond this. They enable us to achieve energy savings and consumption reduction, significantly enhancing the effectiveness and efficiency of our products while improving collaboration efficiency with suppliers and other participants. This self-development capability allows us to fully leverage data resources for efficient and rapid iterative upgrades. This is also why we invest heavily in self-research. We are confident in maintaining our leadership position in competition because it is difficult for new entrants or existing competitors to surpass us based solely on one or two advantages, which is very different from the digital competitive landscape.
Q: Will XPeng start selling humanoid robot products in 2026? Based on the current development progress, which aspects of humanoid robots is XPeng satisfied with, and which aspects need further progress before the official launch?
A: The research and development progress of humanoid robots is going very smoothly. I believe we can now not only achieve basic lower body capabilities but also develop leg, arm, mouth, and some very simple but necessary intelligent capabilities for humanoid robots. We refer to this as agent capabilities combined with humanoid or semi-human brain intelligence capabilities. However, this is very, very challenging. I hope to share more updates with you in the near future when we have more information.
Q: Does the mass production of humanoid robots in 2026 refer to large-scale production for the commercial market? What challenges does humanoid robot manufacturing face? How is XPeng's humanoid robot R&D progress?
A: The mass production of humanoid robots mentioned for 2026 refers to large-scale production for the commercial market. Manufacturing humanoid robots is more challenging than producing smart cars because it requires comprehensive integration capabilities. A true humanoid robot not only needs hardware but also requires real data and simulated data of human activities, making such comprehensive integration difficult to achieve. Currently, many peers in the market focus on the lower body movement capabilities of humanoid robots, which is very limited, only reaching L2 level, while a humanoid robot without L3 level artificial intelligence capabilities is basically uselessEven the most basic humanoid robot models require many conditions, including jointed arms, mobility, and flexibility. XPeng's humanoid robot research and development is progressing well, currently not only possessing basic lower body capabilities but also leg, arm, mouth, and some simple yet necessary "brain" capabilities, namely the so-called agency capabilities and humanoid or semi-humanoid brain intelligence.
Q: Can you provide guidance on the gross margin for the first quarter? What is the fair value gain related to the Didi acquisition? How is the financial improvement of long-term investments?
A: First, regarding the gross margin guidance for the first quarter, we saw an improvement in automotive gross margin in the fourth quarter compared to the previous quarter, mainly due to cost reduction measures and the expansion of economies of scale. We expect to continue consolidating a double-digit automotive gross margin level through cost optimization in the future. As production scale continues to expand, manufacturing costs will also benefit further.
Q: We noticed a significant improvement in long-term investment income in the fourth quarter compared to the third quarter, and the book value related to Didi has also increased significantly. Can you explain why there was a revaluation gain of 200 million yuan in the fourth quarter?
Adjustment related to the Didi transaction: You may recall that when we completed the transaction with Didi, we embedded an exit mechanism linked to the delivery volume of its right-hand drive vehicle platform. As we launched the Mona product line, we internally decided and reached an agreement with Didi to shift our strategic focus to C-end customers rather than the B-end market. Upon evaluation, the first milestone of the transaction could no longer be achieved, so we relieved the related liability obligations, and this accounting treatment has been reflected in the financial statements.
As for the long-term investment income part, we made revaluation adjustments based on the valuation of the invested enterprises. The changes in the fourth quarter mainly stemmed from the equity method accounting impact of XPeng Huitian, as well as valuation fluctuations of other investment portfolios. However, overall, the financial impact of long-term investment revaluation is far less than the liability relief adjustment related to the Didi transaction.
Q: How is XPeng's expansion in international markets, and what are the future plans and expectations?
A: International market expansion is a key pillar of XPeng's growth curve. Last year, XPeng's products received positive feedback in several new markets, and by the end of last year, products were sold in over 30 countries worldwide, ranking among the top three electric vehicles in some categories in certain countries. This year, the global platform will continue to expand, with the number of international stores expected to double from about 150 to over 300, and the number of countries selling products will also nearly double. Based on these expansions, we expect international sales to achieve nearly double-digit growth year-on-year.
At the same time, the company will invest in teams and international infrastructure, and this year will build overseas R&D centers to recruit local talent focused on international technology development. We will also explore local production solutions in some regions, such as the announced collaboration with Indonesian partners to explore local manufacturing solutions.
Q: Is expanding MPI to 100 kilometers by the end of this year the final goal, or is there a higher target? How do you view the possibility of car manufacturers or insurance companies bearing some responsibility for driving systems?
Regarding the frequency of autonomous driving takeovers, this is actually a capability dimension that needs to be viewed dialectically. As the system covers a wider range of scenarios (such as including complex environments like communities, parking lots, and side roads), and as the user base grows larger, the takeover rate data may show an upward trendThis is because ordinary users, for reasons of security or efficiency, will have different takeover behaviors compared to the safety-oriented takeover logic of professional testers. The current industry situation is that all L2+ level advanced driver assistance systems still have a significant gap in taking over once every hundred kilometers, including some L4 technical solutions.
We believe that the realization of L3 level autonomous driving requires a collaborative breakthrough relying on next-generation algorithm models, computing power platforms, and massive driving data. The ideal future state is that users drive hundreds to two or three thousand kilometers on average per week or month, needing to take over only once or twice, which will substantially promote the widespread adoption of social-level autonomous driving. XPeng's goal of a takeover every hundred kilometers is based on the expectation that actual driving data exceeds industry testing standards; this is only the initial stage of L3 capability, and our ultimate goal is to achieve takeover intervals of hundreds or even thousands of kilometers.
Regarding the construction of a responsibility identification system, XPeng is actively promoting relevant explorations: during the recent National Two Sessions, we submitted policy recommendations calling for the acceleration of the introduction of L3 level autonomous driving regulatory standards and the clarification of accident responsibility allocation mechanisms. We believe that China will lead the world in the construction of autonomous driving regulations, and the relevant responsibility identification system is expected to be implemented in the near future.
Q: How should we view the business scale of XPeng AI cars, smart products, and humanoid robots by 2030?
A: This question. At the current stage, in the next 10 years, by 2034, we hope that half of the company's sales will come from China and the other half from overseas markets. The company has internal plans and goals for 2030, but they need to be kept confidential. Regarding humanoid robots, it is difficult to predict when the turning point in technology, business, and mass production will arrive. By the end of 2026 or 2027, there may be a clearer understanding of the emergence of mass-produced humanoid robots. Currently, the development of humanoid robots remains unclear, likely at the development stage of new energy vehicles in 2010 or 2011.
Q: The European government has recently introduced new policies to encourage foreign investment in the European automotive market, especially in AI autonomous vehicles. Given XPeng's partnership with Volkswagen, are there new opportunities to explore?
A: XPeng has previously formulated an autonomous driving strategy, is establishing overseas R&D centers, and is pushing autonomous driving technology to global customers, currently preparing to build infrastructure. In terms of cooperation with Volkswagen, both teams and engineering teams are working hard to promote the mass production of the G9 platform cooperation and electronic electrical architecture (EEA) cooperation. The company believes there is potential for cooperation and industrial logic in potential overseas markets
