Hello everyone, I would like to welcome all the friends, students, teachers, and media present today, as well as those tuned in from around the world. Welcome everyone to the press conference for XPeng AI Technology Day. This is not the first press conference for XPeng, but it is the first time the company is returning to our alma mater, South China University of Technology. XPeng Motors has many alumni from this university. Today I am very happy and excited to be here at one of the best universities in China. I am here to share insights about technology, entrepreneurship, and the beginnings of our dreams. Different times bring different technologies and more aspiring entrepreneurs. I look forward to seeing the entrepreneurs and leaders from the South China University of Technology stepping out into the world in 10 or 20 years.
In 2017, there were more than 300 new energy vehicle companies in China, but only a few have managed to survive. Each company has its own unique identity and market positioning. For XPeng Motors, our identity centers on being a leader in technology-driven intelligent driving. This reputation stems from our background as technical experts. We often jokingly refer to ourselves as technical nerds. However, as technical professionals, we also take pride in our work. We strongly believe that focusing on technological innovation is crucial for creating better, more user-friendly products. Our goal is to expand globally and improve our lives, even if just a little. This commitment to innovation is why XPeng has consistently held a technology day for many years. You can see the 1024 Technology Day of XPeng Motors displayed in the picture behind me. XPeng may be the first car company to host a technology day. For the past six years, we have held various 1024 technology days annually. This date represents Programmer Day, and on this occasion, we aim to showcase the exciting and innovative work we have accomplished in technology over the next 12 to 24 months. We want to share our developments with everyone and ensure that we bring as much technology as possible to our end customers in a timely manner. This is our annual commitment.
This year, for the first time, the 1024 Technology Day has been upgraded to XPeng AI Technology Day. Additionally, we have expanded our focus from China to a global audience. Today we are excited to welcome friends from various countries and media outlets who are joining us for XPeng AI Technology Day. Why has this change occurred? We believe that AI has become a prevailing trend and that the time for its global impact has arrived. Throughout my entrepreneurial journey, I've always believed in a saying: no one can beat the trend. History shows us three major industrial revolutions. The first began with the steam age, followed by the electrical age. The second took shape in the 1970s and 1980s with computers, leading to the rise of the internet from 1995 to 2000 and the mobile internet around 2010. Then in 2014, we experienced a third revolution centered around the internet, marked by advancements in new energy and cloud technology. This era has given rise to various companies, evolving from traditional appliances to computers, mobile phones, and new energy vehicles. I believe even more innovative companies will emerge soon. As you can see, OpenAI unexpectedly gained over 200 million customers annually two years ago, while Nvidia recently overtook Apple's market value, reaching the highest market capitalization in the world at one point. Actually, it doesn't matter what industry you are in; you are more likely to succeed if you follow the trends.
Therefore, in the age of artificial intelligence, XPeng Motors will discuss five key areas today: one, how to transform a car into an AI-enabled vehicle; two, the role of chips as the essential driving force in the AI era; three, strategies for developing and implementing operating systems in the world of AI; and two additional topics related to transportation or mobility: one on flying cars and the other on robotics. Those familiar with XPeng Motors know that the company's vision from the very beginning has been to explore the future of mobility. In the past century, we have witnessed the invention of cars and airplanes, significantly transforming how we travel. Previously, we relied on walking and other slow forms of transportation. I believe that in the next 100 years, humans will likely be able to leave Earth and explore space. Overall, the world of mobility is set to undergo significant changes in the future. XPeng Motors aims to create not only different cars but also innovative modes of transportation and future exploration technologies.
Over the past decade, XPeng has been dedicated to advancing future travel through the development of new energy vehicles. Ten years ago, the penetration rate of new energy vehicles in China was just 0.4%. Now that rate has surged to a monthly maximum of 53%. This significant shift highlights an important truth: fast or slow, technology can evolve at varying speeds, and it's crucial to find the right rhythm and pace for innovation. Looking ahead to the next decade, I believe the future will see a strong integration of automobiles and artificial intelligence. Therefore, XPeng Motors has set a clear goal for the next 10 years: to become a global AI car company.
Let me first share XPeng's vision for the global future. Firstly, unlike many Chinese companies that focus on scaling up when entering international markets, we prioritize establishing a strong foundation. Our globalization efforts began in Europe three years ago, and this year we are accelerating our expansion. Over the past three years, we have approached our growth steadily and methodically, beginning with one country, then three, and now four. During this process, we have learned how a Chinese car company like us can adapt to global markets and better serve customers worldwide. I'm very happy that this year we have expanded to nearly 30 countries in Europe and around the world. Notably, among all Chinese brands that are selling pure electric vehicles priced over €40,000 in Europe, we have become the best seller. This presents an interesting contrast in data. I recently visited France, Germany, and Singapore, where our dealers also represent brands like BMW, Mercedes-Benz, and Audi. Interestingly, they have all chosen to partner with XPeng Motors. I recall that the G9 model is sold for over €80,000 in Germany and 230,000 Singapore dollars in Singapore. To put that in perspective, 230,000 Singapore dollars is approximately 1.2 million RMB in China. We have achieved significant success in Denmark; we ranked second in sales of high-end pure electric models, and we rank eighth in Israel among all EV brands and third in Singapore as of last month. By focusing on the high-end market, we have established a strong reputation for our branding, quality, and technology. This marks the successful beginning of XPeng Motors' globalization efforts beginning in Europe. In 2025, we will expand our presence to more than 60 countries and regions and establish over 300 sales and after-sales service outlets overseas. We anticipate significant changes in our approach. I've always believed that in the realm of globalization, two key principles are essential: first, patience, and second, the importance of collaboration and achieving win-win. Success is not merely about solo winning repeatedly; it is about achieving greater victories alongside our partners, the industry, and the local communities. Furthermore, I believe that the most effective strategy for Chinese companies pursuing globalization is not solely based on cost-effectiveness, but rather on leveraging technological advancements. In 2025, XPeng Motors is committed to enhancing our globalization efforts. To illustrate our vision for globalization, we have asked our director to present a short video for you.
Everyone, in 2020, we landed in Norway and embarked on a global expansion, and today we are already selling our products in over 30 countries and regions.
After discussing globalization, let's shift our focus to XPeng's AI car. First, we need to consider electrification, which is the most crucial aspect of an AI EV. In 2022, XPeng took the lead in China by deploying the 800V super-fast charging platform. Many may not fully understand what this entails. The most critical factor for an electric car is its driving range; it needs to be able to drive long distances, which is a key performance indicator. Compared to other technical platforms, the 800V charging system allows a car to travel farther on the same battery capacity. Since the introduction of the 800-volt system, XPeng Motors has implemented it in various models, including the G6, G9, and X9. Currently, you can see that all mid-to-high-end electric cars in China are equipped with 800-volt technology. I recently attended the Paris Auto Show and noticed that many startups were exploring super-fast charging solutions, which are also based on the 800V platform. Having established this foundation in 2022, I am excited to share a series of future innovations in electrification and rechargeable batteries.
The first one is XPeng's 5C supercharger, which is capable of charging faster than a mobile phone. To put this into perspective, the average mobile phone holds less than 0.1 kilowatt-hour of power, with most holding around 0.05 kilowatt-hours. In contrast, a car typically has a battery capacity of about 60 to 100 kilowatt-hours, which means there is a difference of 1,000 to 2,000 times between the two. Remarkably, the XPeng 5C supercharger can charge a vehicle to 80% in just 12 minutes, which is even quicker than the time it takes for a smartphone to charge. Cars use larger batteries to achieve these faster charging speeds. The 5C's supercharged AI battery, paired with the S5 liquid-cooled supercharging pile, can charge more than 1 kilometer in just one second. What an impressive concept! In the past, a key consideration when purchasing an electric car was the charging time. Initially, charging stations were a concern, and the typical charging pile would take about one hour to fully charge a vehicle. However, with the XPeng 5C supercharger, you can be back on the road in just 12 minutes. This is a significant improvement compared to what we had before, and it usually takes 3 to 5 minutes to refuel at a gas station. With advancements like these, worries about charging speed in future electric vehicles are becoming a thing of the past.
The new battery safety technology is known as the 5C protective armor technology. In the past, many people have expressed concerns about whether electric vehicle batteries are safer than the fuel used in traditional vehicles. Perceptions of safety can vary widely among individuals. However, I believe that while the technology for fuel has remained relatively stable in recent years, the technology for electric vehicle batteries is advancing at a much faster pace. Just to elaborate, the data indicates that the side of the new 5C charging pack can withstand up to 890 kilonewtons of pressure, which is equivalent to the weight of a 20-year-old adult elephant. This design allows for pressure to be applied to the side without compromising safety. The current goal for the bottom protection is to achieve an impact resistance of 2,500 joules. To put this in perspective, 2,500 joules are approximately five times the kinetic energy of a bullet from a pistol, significantly enhancing safety. In addition to the impact resistance of both the side and bottom, XPeng's 5C supercharger will also feature IP68 dust and water resistance to further ensure user safety.
In addition to the points I mentioned about charging speed and battery pack safety, we will soon be launching XPeng's Battery Doctor powered by AI. Traditionally, battery management relied on algorithms and fixed rules. However, with AI-derived algorithms, recent tests have shown a significant reduction in battery degradation after seven years of use. This improved battery management can extend the lifespan of vehicles that have been in service for a long time. As a result, we can expect a 30% increase in battery life, substantially less overall degradation, and potential enhancement in battery performance. Many people are concerned about the decline in battery quality of electric vehicles over time, but I believe this issue will be effectively addressed in the near future.
At the same time, XPeng's energy consumption level demonstrates two main ways to extend the range of an electric car. The first method is to equip the vehicle with sufficient battery capacity, while the other method involves optimizing energy management to make the most of the existing battery, allowing it to travel further and saving money for the end user. XPeng Motors initially introduced the G6 and G9 models, and most recently the P7 Plus. Notably, the P7 Plus and all of XPeng's latest vehicles utilize the same battery design to achieve greater range. Currently, the entire lineup often ranks among the top two or three in various range performance evaluations across China, Europe, and globally, with many of these models frequently taking the top spot. Our company has garnered a strong reputation for its effective energy consumption management in diverse conditions, whether in extremely cold or warm climates, on highways, or in urban settings. Some media outlets have commented on our performance, stating that we mislabel our products because we can achieve extended ranges beyond our claims, which is unusual in the industry. How do we accomplish this? Today I will share some key technologies that enable us to do so.
We are excited to announce a groundbreaking new motor technology called hybrid silicon carbide coaxial electric drive. Let's take a moment to applaud this innovation. Many may not be familiar with this technology, but its logic is compelling. By using silicon carbide in the next generation of electric drives, we can achieve higher conversion efficiency in the China Light-duty Vehicle Test Cycle, or CLTC. However, silicon carbide is expensive, prompting us to ask if we can use less of it while still maintaining high efficiency. XPeng has successfully addressed this challenge, reducing silicon carbide usage by 60% while increasing motor output power by 10%. This is a significant innovation. While it might seem like a small technical detail, it allows us to utilize less expensive materials while enhancing performance. Additionally, this technology reduces the size of the coaxial electric drive by 30%. This reduction will have substantial value. For instance, in the second and third row of seating, you will notice an increase in space equivalent to the size of a fist. As a result, XPeng Motors is now able to offer more room in the back seats in the second and third rows. Furthermore, the hybrid carbon design decreases the weight of the motor by 7.5%. In the future, this hybrid carbon electric drive will be gradually introduced across all of XPeng's vehicles worldwide. We hope this technology will enhance everyone's range and ensure a more dependable driving experience.
We have excellent batteries, reliable charging, and powerful electric drives, but we also need supercharging. Supercharging can enhance the speed and convenience of electric vehicles, benefiting users who not only travel in local areas but also across the country. Currently, XPeng has 1,300 supercharging stations operational. Overall, XPeng's charging network, including both supercharging and regular charging stations, exceeds 1,600 locations, covering major cities in China. Additionally, we're excited to share some good news: XPeng's new S5 liquid-cooled ultra-fast charging station can enable a car to charge more than 1 km in just 1 second. Construction of these stations is beginning, and we invite all users to visit XPeng's charging facilities. Our supercharging stations also feature a new millisecond-level power-off safety protection system. Looking ahead to 2025, we plan not only to expand our supercharging station network within China but also to take our services global. Everyone knows that China's infrastructure is really impressive. China's supercharging and ordinary charging piles are about three times the global total. China has a large number of electric vehicle companies, especially high-end electric vehicle companies, and they are all doing their self-branded supercharging, but there are still very few in the world. XPeng will bring supercharging to Europe and more countries in 2025.
However, the construction of highways in many countries is significantly slower than in China. Currently, we only have supercharging stations available, but these stations may lack sufficient power, and the road conditions can be inadequate. We believe there is a need for a new method of energy replenishment to support our expansion into larger global areas, including vast regions like Northwest China, the cold Northeastern areas, and other countries where electricity supply may be limited. That's why the hybrid extended range is now on the horizon. Over the past few years, we have frequently discussed hybrids, particularly regarding their future development as extended range models. Recently, there have been numerous discussions in the industry about whether XPeng should pursue an extended range hybrid model. From our perspective, creating a hybrid and conventional range extension has been largely unproductive in the past. However, considering the future global development trends, it is valuable to explore different hybrids and range extension options. As a result, XPeng has decided to launch the XPeng Kunpeng Super Electric System. This system will utilize the next generation of electric technology developed by XPeng to address the evolving needs of users in terms of range extension. We refer to it as the Kunpeng Super Electric System for several reasons. First, we call our company XPeng Motors. In Chinese mythology, the Peng is a legendary creature known for its ability to both fly in the sky and swim in the sea. Our vision is that XPeng's future vehicles will embody the concept of one car with two functions. The first function will be pure electric driving, while the second will be the XPeng Kunpeng Super Electric System. This innovative system represents the next generation of extended range technology that XPeng plans to introduce to support globalization and adapt to various environmental developments.
How do we create the next generation of electric vehicles? In the slide, we can see several key technologies about to be launched in the pure electric field: 5C supercharging technology, driven diagnostics, and enhanced efficiency with hybrid carbon. But can we leverage these technologies to achieve a super pure electric vehicle? The benchmark for a super pure electric vehicle is to achieve a pure electric range of 430 km and a comprehensive range of 1,400 km. In comparison, most current industry extended range electric vehicles provide a range of only 200 to 240 km, approximately half of our target range of 430 km. Furthermore, the typical charging speed for these vehicles is around 30 minutes, but generally only allows charging from 20% to 80%. While this is considered fast, many vehicles still rely on AC charging, which results in shorter battery life and slower charging speeds. XPeng's Super Electric System aims to address these issues. It allows users to drive an XPeng super electric car and only need to charge once or twice a week. Achieving a range of 430 km means that this charging frequency is sufficient, significantly reducing the hassle of plugging in for charging every day. Also, many users of hybrid and extended range vehicles report that these cars can be quite noisy, particularly when the engine starts. However, in the super electric machine produced by XPeng, we have developed an ultra-quiet extended range system that is barely noticeable. Our goal is to ensure that when the extended range engine starts, the noise level is only 1 dB higher than that of the electric motor. Many users of extended range vehicles encounter problems such as poor acceleration, inadequate climbing performance, and unsatisfactory highway driving. XPeng Motors addresses these issues by utilizing large-capacity extended range batteries and AI technology to adapt to the current driving environment. For example, if the vehicle is on a regular road, the system maximizes electric power usage. Conversely, on muddy or mountainous terrains, the engine can be started in advance to extend the cycle between engine and electric power. XPeng Kunpeng Supercharging System enhances the performance of its pure electric vehicles, integrating features like 5C, hybrid carbon, safety, and power. This technology enables a range of up to 430 km in pure electric mode, ensuring a quiet yet powerful driving experience. I believe that XPeng's pure electric vehicles and super electric vehicles will greatly benefit customers in China and around the world over the next decade.
I just spent nearly 20 minutes discussing the technological advancements in electrification and super-electrification. Now let's shift our focus to XPeng's autonomous driving. A friend shared an interesting comment with us: XPeng is crossing the river by feeling the stones, but others cross the river by feeling the stones. XPeng Motors started in China 10 years ago, initially focusing on car manufacturing and now advancing into autonomous driving. I recall the development phase of the P7 when we had no clear idea of how a car could drive autonomously on the highway. We were uncertain about camera placements and how to ensure adequate computing power. At that time, we were essentially crossing the river by feeling the stones. We were pioneers in several areas, being the first in China to launch a Navigation Guided Pilot, or NGP, for highway driving, the first to introduce a car equipped with LiDAR for urban environments, and the first in the country to enable nationwide driving without reliance on HD maps. This year, we also achieved the milestone of implementing end-to-end autonomous driving. While we've made significant progress and achieved many innovations, it's important to acknowledge that we've also taken numerous detours along the way. This is the inherent cost of being innovators, but it's a journey we must undertake. Many believe that learning drives innovation at the fastest pace, and I completely agree. We are very confident in the field of autonomous driving because we have been learning to innovate, and others have learned from us. Since last year, numerous colleagues from various industries have approached me, saying, 'How could you have believed in autonomous driving so many years ago? It wasn't until last year that we truly recognized its potential for the future.' Indeed, when you are ahead of the curve, it often feels like you're fighting alone. I am pleased to see that today's Chinese car manufacturers have come to understand the importance of both autonomous driving and assisted driving, which are now widely discussed in the industry. This year, it's not just high-end expensive cars that are focused on autonomous driving; we are also witnessing a growing interest in this technology among low- and mid-priced vehicles.
Today, XPeng will share some of our new insights and visions for the future of autonomous driving. In the past, we had two types of NGP: City NGP and Highway NGP. I want to share information about both and discuss our plans to unify them under one single term: NGP for China and the global market. A few years ago, XPeng Motors was the first to introduce the concept of full-stack self-developed autonomous driving. Since then, the term 'full-stack self-developed' has been mentioned at many press conferences across the industry. We have also put forward a new perspective on the next generation of full-stack self-developed. This includes self-developed large models in the cloud, self-developed models on the vehicle side, self-developed chips, and the entire base of underlying hardware. Among these four self-developed systems, we are introducing XPeng's next-generation NGP, which requires comprehensive full-stack self-development across all four areas.
I am very happy. Two years ago, as a practitioner in the field of autonomous driving, I wasn't sure when we would achieve global success in this area. We faced a significant dilemma. Initially, we relied on people to write algorithms and establish rules, which proved to be an endless process. With a program consisting of around 10 million lines of code, it was still manageable. However, once we reached 100 million or 200 million lines, conflicts among rules started to arise. Additionally, driving logic varies from country to country. For example, the way you drive in China may differ significantly from driving in France. How can we solve this problem? I am particularly pleased with the changes we observed in the large AI model last year. In the past two years, XPeng has gone through various explorations with the end-to-end large model for autonomous driving. While many of these attempts have resulted in setbacks, there have also been moments of success. We have come to realize that the potential of utilizing the large model to transform autonomous driving and achieve global integration is growing stronger. In the picture behind you, you can see XPeng Motors' latest training techniques that utilize cloud technology, local terminals, and extensive amounts of data. This approach facilitates the delivery of knowledge to local terminals by increasing the flow of information. Additionally, the system incorporates two key modules: simulation and reinforcement learning, which work together to enhance performance and ensure stability in the new XPeng autonomous driving technology. This year, we have observed numerous changes in OpenAI practices and the development of the next generation of training models. These changes align closely with the actions we took at the end of last year. The model has transitioned from a single large model to a combination of multiple large models. Here is an analogy that our team spent a lot of time contemplating, trying to capture the difference: in the earlier stages, you worked hard independently. Later, you had the opportunity to run a small model locally with your teacher guiding you step by step. However, when it comes to utilizing a large model in the cloud, it requires a strong network, more knowledgeable teachers, and a substantial amount of data to facilitate your growth. For instance, XPeng's large model in the cloud has 80 times the data and training capacity compared to the car side. Given this situation, what should we do? I would like to share that we will use 10 EFLOPS to help us achieve comprehensive results next year.
After the recent release of version 5.4.0, we noted that it took three months and several iterations to complete the upgrade. We have improved the overall experience by four times, including enhancements in approaching a human driver. We invite everyone to test out XPeng Max model with the new 5.4.0 autonomous driving features. You'll discover that it can navigate all roads in Guangzhou and operates like an experienced human driver. The improvements in human-like driving skills have increased the success rate of lane changes, both longitudinal and lateral, by 50% to 150%. Our goal is to achieve fewer than one takeover per 100 km in all urban areas in China by the second half of next year. This poses a significant challenge, but we believe achieving it will be a remarkable success. I believe that college students present here still need to learn how to drive. Younger friends and perhaps the junior students after you may not require driving lessons, since cars can be quite safe and may drive better than a person. Taking over control of the vehicle once every 100 km is roughly equivalent to an average daily commute of about 70 km for someone going to work. This means you might only need to take over the wheel three to four times a week, and these instances are primarily for gaining experience rather than for safety concerns. By the end of this year, the entire process of AI driving from your home parking space to your workplace or school parking space will be completed. You will only need to activate the system when you get into the car and close the door. After getting out, the car will handle the rest automatically. It is anticipated that within the next 18 months, models currently equipped with XNGP Max will offer a software experience comparable to Level 3+. This is our most important task moving forward. Additionally, in 2026, we will begin exploring unmanned driving in select scenarios.
I am very happy to share something important today. In the past, many cars have been marketed with claims of autonomous assisted driving, but few have clearly explained their capabilities. I understand that the goal of many product managers in this industry is to develop these features for promotional purposes. However, this is not the right mindset. I believe it is essential to prioritize usability. A function should not just exist; it must be well-developed and user-friendly. As one of the leaders in autonomous driving in China, XPeng Motor has officially disclosed its autonomous driving evaluation system. Over the past 10 years, we have traveled to 2,595 cities across the country, achieving nearly 200 million km of simulation test mileage and almost 10 million km of actual vehicle testing. We have also collected a substantial amount of real data from car owners. We hope to release the results of these data evaluations and collaborate with the industry to advance high-level autonomous assisted driving. Our aim is to establish a foundation for complete assisted driving and ultimately unmanned driving evaluations in the future. I am very pleased that Shanghai Auto Research has begun sending letters expressing its willingness to collaborate with XPeng and other companies to improve our efforts. Today, I also urge more media professionals and automotive testing and inspection agencies to join us in establishing an open and transparent standard for autonomous driving testing. This standard can unify the testing criteria for autonomous driving scores across different manufacturers, models, and versions, which will ultimately help users make easier and better purchasing decisions.
I just spent a significant amount of time discussing AI cars. Now I'd like to shift the focus to chips, specifically XPeng's self-developed Turing chip. I would like to share our story behind chips. I believe it takes immense courage to develop chips. The process can cost hundreds of millions of dollars and...
Typically, chip development takes several years with extremely high risks involved. Many chip companies in China are putting in tremendous effort in this area, but some are ultimately deciding to give up. I believe that making chips involves a certain amount of luck. When XPeng developed its chip, it was based on the Transformer architecture. Later, I noticed that all the large models were also built on this foundation. This chip happened to align with the trend of technological change.
In the field of intelligent automobiles, XPeng Motors was the first in our field to propose the concept of map-light, which means reducing reliance on high-definition maps. Later, we introduced radar-light, which involves less use of radars. Today, we are talking about computing power-heavy, highlighting the growing demand for computing power in the automotive industry. In the past, cars didn't require a significant amount of computing power because it wasn't utilized to its full potential. I predict that by 2030, the computing power required for cars may reach around 10,000 teraflops. Currently, the typical computing power is around 500 teraflops. I believe that this demand for computing power will be substantial in the future.
Given such high requirements, even at the server level, a typical card has only 1,000 to 2,000 teraflops. How can we achieve tens of thousands of teraflops? A crucial aspect is that we must develop our own solutions. We need to customize our chips. Understanding that chip design is quite complex, it's worth noting that the most expensive chip in a car can cost nearly 1,000 USD. Consider how many chips are installed in our vehicles. If an AI car has 5,500 chips, the most expensive ones are significant expenses, but they are often not fully utilized.
Companies that produce chips usually cater to multiple industries. They rarely specialize in creating customized chips for a single application. To create versatile and multi-functional chips, significant amounts of waste can occur. A company might only utilize 30% to 40% of its chip capabilities, while the remaining 50% to 60% is available for use in other industries. To address this issue, XPeng has developed its own Turing chip. This chip has been specially customized by XPeng for large models and big data. I'm excited to share more details about it today.
This chip features 40 cores. You typically hear about chips with four, eight, or 16 cores, but the chips used in cars are becoming larger and more complex. The largest local model running on this chip has reached 30 billion parameters, which is impressive. Additionally, we have two independent image signal processors (ISPs) in this chip, enabling enhanced camera capabilities such as managing reflections, rain, and night conditions. These features are all supported by the independent ISP.
This chip plays a crucial role for XPeng, but we know that the chip will be utilized in XPeng cars, flying vehicles, and robots. Over the past 40 days during the tape-out process, we have completed 2,791 functional verifications. Within the entire 60 to 70-day time frame, we operated the NGP on the self-developed Turing chip. I believe that our overall development efficiency has increased by more than three times.
Most chip research and development occurs overseas, while China focuses on upper-level research and development. If any issues arise, it might take us two weeks or even two months to enhance the accuracy. However, at XPeng Motors, our team collaborates closely and puts in significant effort. I am confident that we can resolve issues in about two to three days. I believe that any company aiming to excel in the AI field in the future is likely to achieve vital advancement in chip technology.
I would like to share some insights about the computing power of XPeng's Turing AI chip. In fact, the performance of Turing's AI chip is quite unique compared to standard chips. While other chips focus on broad computing capabilities, XPeng Motors focuses on maximizing performance in specific areas. For instance, in the realm of large models, our effective computing power in one chip can match the combined capabilities of three of Orin X's current chips, which is comparable to the performance of two chips from NVIDIA. Our chip can effectively support both large autonomous driving models and future smart cockpit applications.
What I just discussed might be quite technical. After talking about chips, it's important to understand that the bottom layer of a chip needs a better supporting structure akin to a nervous system. Today, I want to share information about XPeng's Tanghai platform. When XPeng released the G3, the vehicle contained 70 individual components. In contrast, their latest model has reduced this to just seven components, with plans to decrease it even further to five or four controllers in future models.
For instance, a car may have seven airbags, each requiring its own control unit because they are independently controlled. The challenge is how to integrate these various control systems into a cohesive framework that operates more effectively. This concept is similar to a combination of a car's neural network system, central processing system, and spinal system. This year, XPeng Motors collaborated with Volkswagen on the EA 3.5, which is the current version of the Tanghai platform. The Tanghai platform represents an updated version of this technology. It is an AI nervous system designed by XPeng for level four autonomy in unmanned driving scenarios.
This system will be integrated with existing capabilities in the future. I want to quickly share what impact this base is expected to achieve. The communication bandwidth of the entire vehicle has increased by 33 times. Today, XPeng completed a vehicle upgrade in just over 20 minutes. Can we further reduce this time to a few minutes? Is it possible to transfer computing power from one unit to another within the vehicle? Can we also enjoy gaming and perform various tasks? The new platform has enabled us to achieve all of this.
The efficiency of camera output has improved by 12 times, and significant power savings can be made in many areas. For instance, the Sentinel mode can reduce power consumption by 72%. These advancements are the capabilities provided by the new platform for future AI vehicles. We have successfully implemented four safety redundancies at this platform, achieving the highest safety standards in the industry, specifically the ASIL-D safety standards. This has been accomplished through the installation of double security partitions and double security backups.
What I just shared with you pertains to XPeng's Turing chip and its capabilities in the field of AI intelligent driving. Now, allow me to provide a summary. XPeng's next-generation navigation NGP aims to utilize maps, radars, and advanced computing power, making it easier to drive autonomously around the world. This foundation is crucial for XPeng to develop a robust autonomous driving system that can be deployed globally.
I'd also like to share an interesting feature. Previously, the reversing function provided a 360-degree view around the vehicle. However, with the new platform and upgraded technology, our goal is to enable vehicles to achieve a 720-degree perspective, allowing them to view their surroundings from a greater height. This feature is particularly useful for parking, as there are often fire hydrants located above ground in many areas. The 720-degree visual parking capability allows for a 3D perspective, enabling us to identify and avoid obstacles even when they are positioned above the car, such as those on walls.
Tomorrow, November 7, the XPeng P7 Plus will be launched. This marks the first time that we will not distinguish between the Max and Pro versions of the car. The entire series will come standard with Turing AI intelligent driving technology, showcasing the most advanced technological capabilities in China's intelligent ADAS driving.
Let's discuss level four autonomous driving. Many of the students here will experience L4 technology after graduating from undergraduate and graduate programs. L4 represents the advancement from autonomous driving with assistance to fully unmanned driving in certain situations, and potentially in all situations in the future. I would like to share XPeng's insights and progress in this area.
First, we will implement steer-by-wire technology in level four autonomous vehicles. For the past 100 years, the steering wheel has been recognized as the most important and safest component of a car. It is hard to imagine a car without a steering wheel, even in the future. However, the introduction of steering-by-wire will enable level four autonomous driving systems to reduce the size of the steering wheel or even allow it to fold away entirely, potentially making it obsolete in the automotive industry. In the next phase, XPeng Motors will gradually integrate steering-by-wire to more vehicles to establish a solid foundation for autonomous driving.
Second, we are currently developing a new Ultra model, which is an enhanced version focusing on high-level autonomous driving. This Ultra model will gradually transition to a system that features both human-machine co-driving and partially unmanned operation, ultimately evolving into a fully unmanned service known as Robotaxi. At that point, XPeng's computing power will reach 3,000 TOPS, enabling a fully redundant design for both hardware and software to support the Ultra model in China, Europe, and potentially other countries. I believe we are already witnessing the advent of high-level autonomous driving, and fully unmanned driving will certainly be a reality in the near future.
I recently discussed the role of AI in cars, chips, and intelligent driving. Now, let's shift our focus to the next generation of operating systems, which I refer to as AIOS. Before diving into this topic, I want to briefly highlight the evolution of AI models. We begin with the publication of the initial research paper in 2017, followed by Google's BERT in 2018, the release of GPT-3 in 2020, and the launch of ChatGPT at the end of 2022, culminating in the recent introduction of GPT-4. This timeline illustrates the rapid acceleration of AI development.
Previously, programmers wrote code to implement specific functions. Now, we have shifted to creating models that learn from the environment and generate new rules autonomously. This change represents a fundamental shift in methodology. What does this mean for operating systems? If AI capabilities are deeply integrated into the foundational structure of these systems, it will undoubtedly transform the current landscape of operating systems. In my opinion, the evolution of large AI models will give rise to what I refer to as AIOS. I believe that AIOS will become the standard configuration for future AI cars in XPeng's ecosystem.
Apple has recently launched a large language model that can operate on both cloud and local terminals. One reason mobile phones offer limited computing power is their small battery capacity. Typically, a mobile phone has a power supply of about 0.01 kWh or less, while electric cars have batteries that range from 60 to 100 kWh. The significant difference in energy capacity means that mobile phones do not have enough power to support high levels of computing performance. This raises a question about the future of electric vehicles.
Traditional gasoline or hybrid cars have very limited electricity supply, typically ranging from 0.7 kWh to 1.1 kWh. These vehicles depend on both electric motors and internal combustion engines for power. As a result, without a robust electric or advanced hybrid system, these cars will face challenges in supporting a large server infrastructure in local areas. Consider this: you have computers, refrigerators, and color TVs in your home. Without electricity, you wouldn't be able to use any of them. Similarly, electric cars represent the future of AI-driven vehicles. In this context, I believe both fully electric and advanced electric cars hold significant potential for ushering in a new era of intelligent driving.
Today, I'm excited to share that the goal of XPeng AIOS is to pave the way for this new era of AI-driven mobility. In the past, the architecture of smart cockpits in the automotive industry typically consisted of a single computing power paired with a cockpit system. However, looking ahead, I believe that future cockpit systems will incorporate at least two, if not more, chips for the computing power. This advancement will not only support existing systems but also enable the implementation of new AI operating systems. XPeng's new AIOS is set to redefine cockpit systems by utilizing more chips to run a larger operating system.
We are witnessing a remarkable increase in computing power, with enhancement of up to 20 times in AI capabilities. In the near future, XPeng is expected to introduce a large model with over 10 billion parameters into their vehicles for the first time. Why should we equip cars with a large model? The main reason is that it allows for fast interactions while ensuring everyone's privacy. The data you exchange with the car doesn't need to be sent to the internet or to the cloud. This approach can reduce the overall cost of car usage for everyone.
Did you know that many vehicles overseas suffer from poor network connections, and data services can be expensive? We aim to integrate our self-developed GPT-4-like capabilities directly into the cockpit, enabling local operation, real-time perception, privacy protection, and instant communication. In the future, with the merger of robotics, we could even achieve autonomous driving. This represents the cockpit of the next generation AIOS. It will support multiple real-time language conversations, allowing you to communicate efficiently in both Chinese and English simultaneously.
For example, if you encounter fog and need to clear it, the system will automatically assist with defogging instead of just providing a simple guide. One of the recent features being tested is intriguing: when you approach a parked car and want to open the trunk but find it too close to a wall, you can simply tell the car, 'Move forward a little.' The car could then drive forward by about 20 cm on its own, allowing you to easily access the trunk. This isn't a distant dream. I look forward to seeing it implemented soon.
AIOS is capable of enhancing real-time perception, operation, dialogue, and cognitive functions. It will also contribute to local purification in more vehicles. I would like to share four examples of how it can drive changes in different areas: changes in audio, changes in power systems, changes in battery technology, and changes in chassis design.
First, let's talk about the audio system. Most people listen to music in MP3 format, which is typically a two-channel sound source. However, many cars are equipped with 10, 20, or even 30 speakers. XPeng AI technology can enhance this experience by automatically converting a two-channel MP3 song into a panoramic sound format. This allows for a more immersive listening experience. In addition to improving sound quality, the AI can manipulate various audio elements. For example, if you want to focus solely on the background music and eliminate the human voice, the AI can remove the vocals for you. It can also adjust the overall tuning style based on different music genres and implement active noise reduction. These advancements illustrate some of the exciting changes in AI audio technology.
Second, let's consider the capabilities of AI in vehicle performance, specifically regarding flat roads and climbing slopes. The AI integrated into the battery not only enhances battery life and minimizes degradation but can also improve performance over time. The chassis, a traditional core component of vehicles, plays a crucial role in ride quality. For example, we often encounter speed bumps in residential areas that require us to slow down. Additionally, rugged roads can cause a bumpy ride, leading to noticeable swaying from side to side. In the past, drivers had to rely on cars equipped with air suspension, which, although better, still often required manual adjustments for different driving conditions such as off-road or road mode. Recently, XPeng version 5.4.0 has introduced the first stage of AI chassis capabilities, with plans for further advancements in the integration of AI and chassis performance in the future.
XPeng AeroHT has been developing electric vertical takeoff and landing (eVTOL) for nearly a decade, marking the beginning of autonomous flying. Similar to XPeng Motors, this year the low-altitude aerospace economy is starting to gain momentum. Many people here still don't fully understand what the current generation of flying cars aims to achieve. As you can see in the slide, there is a car that can have a two-wheeled aircraft stored in its trunk. When the aircraft is ready to fly, it can automatically fold itself and fit back in the trunk.
In the past, most people were unable to fly planes for various reasons, one of which was the lack of a suitable platform for parking, taking off, and landing the aircraft. We jokingly refer to it as a flying aircraft carrier, much like how fighter jets can take off and land on an aircraft carrier. One key advantage of this innovation is that if you own an aircraft, you can drive it to your parking garage and park it in your designated space, making it much more accessible. The first mass-produced flying car must significantly lower the barriers to its use. However, even if this goal is achieved, I believe there will still be numerous challenges to address.
Today, I want to share with you four common questions that people often ask about this technology: Can this plane be sold? Does the government permit its manufacturing? Is it certified for flight testing? How can I fly it if I don't know how to operate a plane? I want to travel from point A to point B. What kind of approval do I need? What about the flight route? And most importantly, is it safe?
Firstly, in response to the first question, the flight body of the land-based aircraft carrier has successfully completed the acceptance process for the sea trial certificate in China. Today, we hope to obtain the Chinese civil aviation type certification as soon as possible next year. I believe this timeline will not be too far off, especially since the entire country is actively promoting the development of the low-altitude aerospace economy.
Secondly, how do you fly a plane? We introduced a single-lever control system that makes it easy for anyone who knows how to drive to start within just five minutes and achieve a basic level of proficiency in three hours. Additionally, XPeng's assisted driving features in the aviation field can help simplify many complex flight challenges. Jilin and Hunan provinces are currently designated as test areas for the low-altitude aerospace economy, with several cities participating in trials. Shenzhen is also a test city for low-altitude flying in China. I believe that by around 2026, a significant portion of the country will be able to use electronic, unmanned, one-click filing and takeoff systems. At that time, flying cars will become very convenient to use.
The latest XPeng flying car has a user-friendly system that eliminates the need to set complex routes or enter precise GPS coordinates. Instead, you can rely on simple AI to assist with intelligent takeoff, landing, planning, and even returning home with just a single button. To ensure everyone's safety, various flight maneuvers have been simplified. For instance, while a fighter jet can maneuver diagonally, our flying car is designed for more straightforward movements: up and down, left and right, and forward and backward. This design makes flying simpler and safer for users.
In the field of safety, we have made significant advancements. The most critical aspect of flight safety is that it is non-controllable, featuring three levels of redundancy. When one breaks, you will have another two. In terms of power supply, communication, and control, we have implemented dual redundancy systems for all of them. This additional redundancy enhances overall safety. Traditionally, helicopters are designed with only one propeller. If this propeller fails, there is still a chance to address the issue before takeoff or landing. In contrast, XPeng aircraft are equipped with multiple propellers, allowing for safe flight even if one or two propellers stop rotating. This means they can still take off and land safely even in the event of partial failure or issues with related twin rotors.
How can consumers be sure it is safe? We should have the courage to try it ourselves. I've put forth a requirement to the AeroHT team that our senior executives need to collectively fly a distance of 5,000 km in this product, which is equivalent to about 2,300 individual flights, before the final delivery. I'm pleased to note that both founders of AeroHT have flown recently, and I look forward to flying myself in the near future. I'm also excited about the prospect of having more friends join us next year and the year after as they take the opportunity to fly independently while prioritizing safety. I understand that flying for the first time may seem intimidating, but once you get accustomed to it, you will discover a completely different joy in flying.
This flight was significantly better than my first test flight in Dongguan. During that flight, there was no cockpit, and the propellers were spinning uncomfortably close to my hands. We are looking forward to the global public debut of the new flying car at the China Air Show in Zhuhai, China, next week. We also hope to officially begin pre-sales in December. Anyone who enjoys flying is welcome to join us in Guangzhou for a test flight.
Finally, let's discuss robots. You might think that flying cars are purely science fiction and represent the future of mobility. However, I believe that robots can also be considered a sci-fi mode of transportation for the future. Many of my friends ask me, 'You keep saying you want to be an explorer of future mobility. Are robots really a means of transportation?' I respond that traditionally, every mode of transportation has been non-autonomous and merely functioned to move people or goods from point A to point B. However, with the emergence of robots, we will see an increasing number of autonomous means of transportation in the future. This shift is a crucial part of the evolution of travel.
When transportation tools possess autonomous thinking, both robots and driverless cars should share a common goal and be capable of navigating a variety of traffic situations in the future. Driverless cars should be able to communicate and engage in intelligent conversations both inside the vehicle and with their surroundings. This capability is essential for the advancement of future automotive technology.
When I first entered the automotive industry, I was asked how much investment was necessary. Initially, I thought an investment of 10 billion RMB would suffice, but later realized that 20 billion RMB was more appropriate. I have been working on robotics for five years now, and the more I delve into it, the more I understand the complexities and challenges it presents. This field demands significant investment, especially in AI technologies. Recently, I came across a news report stating that OpenAI is preparing to enter the hardware sector. It seems likely that they may focus on robotics.
One of the primary challenges in developing effective robots is the need for extremely powerful hardware combined with sophisticated software capabilities. Over the past five years, XPeng has developed four generations of robots, evolving from quadruped to bipeds, and advancing from single-point bipedal designs to multi-point configurations. Today, let's take a look at a video showcasing the latest development of XPeng's fourth-generation robot.
Today, I am excited to share for the first time that the name of the AI robot is Iron. Welcome, Iron. When discussing XPeng's AI robot system, it's important to note that our robots are designed to resemble humans. Some friends have suggested that our robots don't need to look human; they could simply have wheels to roll around, which might make them faster and more stable. While that perspective has its merits, let's consider it from a different angle.
Creating effective technology falls under scientific research, while developing a quality product is referred to as research and development. Large-scale production of a commercially viable commodity comes later. If we approach this from a scientific research standpoint, we could design robots that aren't human-like and can be customized for specific scenarios. However, when we consider large-scale mass production, we begin to see certain commonalities among successful technologies and commodities that often resemble human features. For example, if you aim to address the issue of user acceptance, designing a robot with eight arms instead of two might deter potential buyers due to fear. Additionally, if your robot is equipped with wheels, you may discover that it struggles to maneuver in various environments.
XPeng is dedicated to creating a life-sized humanoid robot designed to mirror the human form at a one-to-one scale. In considering this project, we must take into account various human capabilities such as the heart, muscles, spine, brain, cerebellum, brain stem, eyes, ears, mouth, as well as sensory perceptions like touch and temperature. XPeng Iron is currently 1.78 meters tall, weighs 70 kg, and has 62 degrees of freedom of movement. XPeng's robot utilizes even more Turing AI chips than XPeng Motors, as we aim to integrate the brain, cerebellum, brain stem, and spine into a single system. This requires significant computing power and extensive models for training.
XPeng will equip the robot with the AI Eagle Eye perception system, which will soon be released on the P7 Plus. This system not only allows the robot to see its surroundings in 360 degrees but also in 720 degrees in three-dimensional space. It's important to note that XPeng Motors will not use LiDAR technology in the development of its robots. The robot must be capable of walking and planning its movements, which requires reinforcement learning. Additionally, it needs XPeng's end-to-end large model for autonomous driving and a substantial amount of data. We have conducted extensive research on the robot's walking capabilities.
Despite this, no company has yet achieved a truly reliable and stable walking mechanism for robots. There are two different videos featuring our robots. Similar to many robots from other companies, the video on the left is titled 'Small Steps.' In earlier attempts, robots did not walk in a natural manner akin to human beings. For instance, if you saw a person walking in an unnatural way, it would stand out to you. Previously, XPeng's autonomous driving relied on humans to write the rules, which limited its ability to drive like a human driver. However, with the implementation of reinforcement learning and advanced algorithms, we have improved the robot's walking capabilities. Today, we are focused on how to make robots move in a more human-like way.
Additionally, earlier robots moved blindly, as if their eyes were covered. We need to figure out how to enable robots to perceive their surroundings. For example, if you hand me a glass of water, I can take it and continue walking. This kind of action requires seamless cooperation. While robot locomotion is fundamental to their functionality, this challenge has not yet been fully resolved. The AIOS developed by XPeng includes capabilities such as conversation, memory, and reasoning. We are excited about the potential for thinking abilities in the future, which would allow us to advance alongside XPeng Motors. Currently, both our robots and XPeng cars utilize the same large models within AIOS, and they have been trained on extensive data sets from around the world.
I have some exciting news to share with you. AI robots are already being implemented at XPeng's factory in Guangzhou. Some roles for the upcoming XPeng P7 Plus, which is about to be released and delivered, are currently being performed by these robots. Allow us to show a short video of the robots in action.
XPeng will open its factory to the public for visits. We also welcome friends to visit and intern at the factory, where they can even try their hand at screwing. In the future, XPeng robot Iron will be used in more roles within the factory, performing tasks such as screwing in screws, patrolling, and more. We also hope that Iron will be available in stores to engage with and assist car owners.
Today, I spent 100 minutes discussing XPeng's new insights on various AI technologies, including AI cars, AI chips, smart driving, AI operating systems, flying cars, and Iron robots. It seems that many next-generation technologies and future product concepts could soon be launched. The best course of action for XPeng Motors is to embrace these ideas and work diligently to bring them to fruition.
After 10 years of dedicated effort, our self-driving cars are continually improving, and I am confident that flying cars and Iron robots will become available to everyone sooner than expected. At the same time, I want to express my gratitude for this era, which enables ordinary people like us to turn our dreams into reality through hard work alongside partners, classmates, and friends. I am thankful for this time, for our motherland, and for the teachers at South China University of Technology who have supported us and fostered the imagination and practice of many XPeng individuals in this field.
Looking ahead, I envision a future travel world with unmanned vehicles navigating cities, low-altitude flights connecting urban areas, and last-mile deliveries fully handled by robots. I believe that in this wave of technological change, XPeng Motors can be a symbol of China's technological innovation, making a mark both domestically and globally, and offering more exciting and innovative products for future transportation.
