110 minutes is almost 2 hours- enough time for a couple to watch a movie, giving them a topic to discuss for the rest of the evening; for white-collar workers to complete a thorough strength training, feeling the soreness in their muscles for the next few days; for food enthusiasts to indulge in a feast and perhaps create a food review video.
If you give this length of time to Li Xiang, he would share with you the “36 strategies for making a car”, and you may need quite some time to verify and digest it.
Some people say that Li Xiang is trying to teach “me” how to make a car, which is enough to illustrate the immense amount of information conveyed at the Spring Media Communication Meeting. It includes Li Xiang and Ideal Automobile’s understanding and foresight concerning the technical aspect, as well as their profound understanding of brand, supply chain, and consumer demand.
We have summarized the content shared by Li Xiang while making slight adjustments while ensuring authenticity and completeness.
We understand that the development of human civilization over the past few thousand years has had two main influences. On the one hand, energy, from humans’ use of wood for fire, to the subsequent use of fossil fuels, and to today’s large-scale use of renewable energy, is a process of human civilization’s progress. On the other hand, the development of information technology, from humans’ differentiation from other animals, to the ability to communicate with language, to the use of measurement units, to the advent of the printing press, to today’s computers and artificial intelligence, is also progressing rapidly.
Intelligent electric vehicles are one of the few industries that cross both energy and cutting-edge information technology fields, and this opportunity is crucial for all of us.
From the perspective of users, there are three levels of energy demand. The first level is the convenience of acquiring energy. We often see different electric vehicle users giving completely different evaluations on forums, Weibo, and Douyin. Users with charging stations will say that electric vehicles are fantastic and they cannot go back to gasoline cars, while users without charging stations, including those who frequently drive long distances, express regret in buying an electric car. So, why are there two different opinions? It is because the convenience of use is different. For any user who has a charging station, their entire way of thinking is drastically different, and this is the first point.4. The second point is the cost of obtaining energy for users. The cost of obtaining energy actually has two aspects: one is whether the product I bought is more expensive, so the battery price is involved. The other is the usage cost, and of course, the usage cost of electric cars is very low. I think this is why more and more consumers will choose electric cars in the current economic background over the past few years, because the money saved each year can buy an iPhone or a new iPad, which is key to saving money for the entire economy after the slowdown, and this is the second user’s demand.
What is the third user’s demand? The user’s demand is for comfort, environmental protection, and zero pollution, but it is different from what the government calls carbon neutrality and carbon emissions. For example, when an electric car is stuck in traffic, it is more comfortable without a gearbox, which everyone acknowledges. What do users think environmental protection is? For example, if we drive a car ourselves to pick up children from school in the summer, if I turn on the air conditioner and there are smells everywhere, that is pollution, but if I drive an electric car and turn on the air conditioner, it will be good.
The comfort and smell of a car with the air conditioning on for a gasoline car and an electric car are completely different, so I think that these demands are genuine. The better these demands are met, the higher the popularity of electric cars will be, because ultimately, it is about gradually replacing the roughly 200 million passenger cars currently in use.
- In terms of the government, there are three demands. The first is the independent controllability of core technologies. When we are developing rapidly and getting stronger, being balanced by foreign countries is inevitable, so the independent controllability of technology is critical in the competition between major powers.
The second is the security of energy supply. Whether it is our oil production or lithium carbonate production, it is different from our actual shipping and usage. We produce more than 60% of fuel cells in the world, but our lithium carbonate storage accounts for less than 10%. We use more than 20% of the world’s oil, but our oil reserves are probably less than 2% to 3%. Therefore, there is a problem of energy supply security behind this.
The third is the demand of the government. “Green mountains and clear waters are better than gold and silver.” This is about carbon peaking and carbon neutrality. This has also created China’s leading position in the entire electric vehicle industry, the entire industry chain, including motors, electronic control, third-generation semiconductors, new battery technologies, and the entire photovoltaic industry chain, all of which are based on the government’s core vision.7. As far as a car enterprise is concerned, we face only two challenges: the first challenge is the problem of difficult charging, and the second challenge is the high cost of batteries.
Although reducing prices has not helped companies to win, it can strike against competitors. Therefore, the consumer’s attitude to purchase is very obvious, and almost every enterprise has only completed 4%-5% of the full-year goal set after two months.
The price of lithium carbonate has dropped significantly, from 550,000 yuan per ton to the present 390,000 yuan per ton. We judge that the overall cost will certainly not drop to the original 40,000-50,000 yuan per ton, and the price of 200,000 to 300,000 yuan per ton is a long-term and stable price.
We believe that there are three paths to solve the problem of difficult charging:
In fact, battery swapping is a very good way. First, the experience of battery swapping can approach that of refueling, and secondly, many problems that users worry about can be solved.
On the other hand, battery swapping can make the cost of buying an electric car lower than that of a gasoline car, and the whole experience will be completely different.
With battery swapping, there will also be more batteries needed and a risk of large fluctuations in battery prices. Therefore, we did not have much money at that time, and we were thinking about a problem: Can we put it directly on the product side, without relying on the service side? So, we made an extended-range electric vehicle.
We thought about how to combine the user experience with this travel efficiency, so the most important choice we made at the time was to make a large battery (starting with 40 degrees). Today, all the cars we sell can achieve more than 200 kilometers of CLTC cruising range.
What is the third path? In fact, today there is another problem with charging piles, that is, the vast majority of enterprise charging piles do not make money. Because this business can be easily calculated, achieving about 6 to 7 charging orders per day is basically a balanced revenue and expenditure. If 10 charging orders can be achieved, the profitability of a charging pile will be better than that of a gas station.
However, nowadays, on average, only 2 to 3 orders are basically losing money. The method commonly used by everyone, including enterprise financing, is to split the cost of a charging pile that can only be used for 5 years into 10 years, thus reducing the overall profits and losses. The principle of charging piles is similar to that of restaurants, which is a turnover rate.Calculate this, if we can reduce the user’s waiting time to within 20 minutes, then they won’t leave the car. If we shorten it to 10 minutes, we can fully charge it with enough mileage, and the user experience will be basically the same as that of a gasoline car.
To achieve this, there are two main factors. The first core is to use a high-voltage platform. Only a high-voltage platform can achieve such a charging speed. At the same time, the second advantage of a high-voltage platform is that when we use silicon carbide with a high-voltage platform and better aerodynamic coefficients, the battery cost can be significantly reduced, and efficiency can be significantly improved. For example, the best-selling medium-sized SUV currently has about 100 degrees of electricity, which can be achieved with a range of 600 kilometers. However, with the 800-volt high-voltage platform, better aerodynamic coefficients, silicon carbide, and overall efficiency optimization, we can achieve the same cruising range with about 80 degrees of electricity, and the vehicle will be lighter. After the vehicle is lighter, we can reduce the use of aluminum, and the cost can be decreased by about 30,000 to 40,000 yuan. This is also related to our own production of silicon carbide modules and three-in-one motors.
After all these things are done, compared to today’s traditional 400V mainstream electric vehicles, in the same size and driving form, we can reduce the cost of parts of a car by about 30,000 to 40,000 yuan. Therefore, you will be able to see our 800V high-voltage platform electric vehicles, because 4C still has certain selectivity, and we can probably achieve the same price with mini-extended-range electric vehicles. This may be different from what you imagined because people today generally think that 800V silicon carbide is a more expensive price, but it can actually save a lot of costs.
Whether we make extended-range electric vehicles or high-voltage pure electric vehicles, the core purpose is to solve the charging problem. When we use 4C batteries, we can charge 400 kilometers in 10 minutes. When we use 2C batteries, we can charge 400 kilometers in 20 minutes, which is the core.
On the other hand, if we want to launch electric vehicles, charging piles are products, not services. The charging pile is the product itself. If we provide 4C, but there is no complete network of 4C charging piles, then it is actually equivalent to buying a 4G mobile phone, but still only having 2G network. So, we believe that it belongs to the product itself, and you don’t have to worry about the speed and determination of our charging pile deployment because it is not as expensive as you might think.
- At Ideal Auto, we have four artificial intelligence algorithm teams, each dedicated to solving problems in four areas: 1. The intelligent cabin team established last year, which specializes in multimodal algorithms; 2. Intelligent driving; 3. Smart factories; 4. Intelligent retail.
As for the third team, “Smart Factory,” the reason why we use artificial intelligence is that on our production lines, there are hundreds of inspectors in each work process every day, and they can all be replaced by artificial intelligence algorithms. In addition, these algorithms can help us solve quality problems. From a quantitative perspective, the cost of quality per car is about 50% of the price of a luxury car brand like Mercedes-Benz, BMW, or Audi. Therefore, if we measure quality using money, in the case of products with the same price, our quality is significantly better than these luxury car brands.
Furthermore, these algorithms can also help us break down various accidents. Our insurance is the cheapest among all competitors, even lower than that of gasoline cars. As a car priced at over 300,000 RMB, the commercial insurance part for renewing the second year is only 3,300 RMB, and including compulsory insurance, the total is 4,300 RMB.
- Why do we need such a system? Actually, it is related to my experience with Autohome. Almost all products were successful, except one that failed miserably, which was automotive e-commerce. Moreover, many start-up companies in the automotive e-commerce and used car e-commerce industries all failed.
There is an important point to consider. In the entire business chain of automotive e-commerce, 90%, or even 95%, occurs offline. In the whole offline experience process of storage, inventory, logistics, and delivery, we did not make any improvements at that time. The cost was even higher than traditional car dealerships. We only spent an additional 5,000 to 15,000 RMB on CPS costs, and did not create any other value. The same problem existed in new retail that was popular in the past, such as community group purchases and ride-hailing apps. Whether it is Uber in the United States or Didi in China, their current market values are far lower than their financing values at that time. The core reason is that in the past, it was Software 1.0 and could not solve problems in the physical world. What is Software 1.0? It is rules made by humans and programming done by humans, which, after programming, are used by humans. Whether we use the Internet, APP, or other terminals, it’s the same principle. The main core is that it runs in the digital world, but we cannot change what happens in the physical world. We can easily get an Uber ride in the United States, but the total cost of ownership, or TCO, during the ride from point A to point B has not decreased at all. It is even higher than that of a traditional taxi because of higher idling rates. Therefore, these platform companies will face a problem of losses if they want to acquire a high growth rate and share. The only way to make money is to exploit drivers, for example, Uber charges 30% in the United States, and 25% in China, while rental companies only charge 20%. Therefore, they now charge drivers more than rental companies did.Retail e-commerce is different – when this expensive mall is moved to a warehouse, the overall cost drops by 35% to 40%. The entire industry’s 35% to 40% has been divided into three parts. Firstly, consumers take away a portion, which makes retail prices cheaper. Secondly, the platform takes away a portion, creating trillion-dollar companies like Alibaba and Amazon. The third part is these brands, which use e-commerce in new retail ways.
Therefore, at that time, we were thinking about a question – what can change the physical world? What ways can change the physical world? We found that only AI can!
- We divide AI into two categories. The first category is based on long-term analysis and observation, reaching a humanoid state.
There are roughly two ways in which humans learn in the physical world. One is behavioral learning, where the cerebellum is dominant and the brain is secondary. We use our brains to think and train ourselves, but once training succeeds, the cerebellum can handle tasks. For example, when we drive a car or perform some tests on a production line, it often does not require complete thinking by the brain. It can react quickly and timely. This is work dominated by the cerebellum with the brain as support.
The other category is the popular Open AI, such as ChatGPT, Ideal Companion in the car, and NOMI of NIO, etc. Their way of working is cognitive learning. It is dominated by the brain with support from the cerebellum. When you give a complex command, it has to go back to the cloud for processing, and then after training it can execute with relative certainty. However, if it’s done in-car, it can provide results through the GPU or BPU. These are two different ways. One advantage of Open AI is that it’s very difficult to train the brain, but you’re not afraid of it making mistakes. For example, if ChatGPT talks nonsense to you, you won’t mind too much. However, if it is based on the first category of AI, which is the cerebellum’s processing of action execution, once it makes a mistake, it could cause a traffic accident or someone’s injury. The consequences of these two types of errors are completely different.
- First, we need to understand how humans learn to drive and how we work in the physical world in a behavioral way. The first is three-dimensional perception. Humans use two eyes and two ears in conjunction with the brain to obtain and process information about the entire three-dimensional world. Second, we need to make the car judge and decide based on what it sees and combines it with its own state. For example, if it sees a car in front of it, it will brake and slow down. If it sees an accident or a stone in front of it, it will judge that it needs to steer clear of it. Third, we need to execute through the cerebellum and nervous system to determine how to turn the steering wheel to avoid it, how much force and speed to use, or how to apply the brakes and with what force. Fourth, it is the result and feedback, and feedback is actually training. That is, if the entire operation experience is very smooth and safe, I will gradually train and reinforce my instincts.# 15. Autopilot has gone through two stages: behavioral learning and cognitive learning.
In the first stage, due to limited sensors and computing power (only two TOPS on Mobileye with a front camera of only 2 million pixels), the perception could only recognize 2D pictures instead of 3D. Using AI only to identify 2D images, it could draw rectangular boxes that look like cars for cars and boxes that look like people for pedestrians. Using 2D range-finding, the closer something is to the center of the picture, the closer it is to the vehicle. However, due to various vibrations in driving, there are sometimes recognition errors. To solve this issue, a cost-effective solution is implemented by using a millimeter-wave radar to perform range-finding and by feeding back the distance between the car and the object or person for each image. This 2.5D method still dominates the operation of 99% of cars on Chinese roads, even for high-tech hardware systems such as the Orin platform and Tesla’s AP and FSD systems in China.
In the second stage, all current commercially available vehicles make decisions based on transparent rule-based systems instead of AI, even if they do use algorithms. These algorithms are white box in nature and are designed and coded by humans. Thus, decisions like how to handle cars cutting in, how to react when cars leave the lane, how to react at different distances, etc., are all predetermined and written into the algorithms by humans. Today, the driving experience still heavily relies on rules written by humans rather than AI algorithms. Therefore, the first stage requires more testing scenarios and more rules to be written by humans. It is concluded that accidents caused by situations beyond predetermined rules are the drivers’ responsibility since the autopilot system is only an assistance system.The third part, as with the entire execution control, is fully rules-based. Therefore, when encountering a congested road section, if following too closely, drivers may feel a bit dizzy as the brakes are more sudden and the acceleration faster to prevent other cars from cutting in. Cars that are driven more comfortably will have a smoother acceleration, but are more susceptible to being cut off in traffic. Major automakers are constantly adjusting to find the balance between the two.
The fourth part is about feedback and growth. This is divided into two main parts. First, I collected more videos and hired a calibration team to manually calibrate based on these videos. At this stage, whether doing visual or speech AI in the industry, our internal joke is, “the more manual labor, the smarter the AI”.
Another part of the training is that we can adjust rules, whether decision-making or execution rules, to make them more comfortable and safer for more people in more situations.
At the current stage, there is essentially no relationship to autonomous driving. It is just driver assistance, even L2 is only slightly improved.
The essential change in this (L2) stage is due to Tesla’s FSD and its senior director of autonomous driving algorithm, AK (Andrej Karparthy, who joined Tesla in June 2017 and left in July last year), who was recruited by Musk from Open AI. Only after AK joined Tesla did its intelligent driving undergo a fundamental change. Prior to joining Tesla, AK learned 3D vision with Fei-Fei Li at Stanford.
The efficiency of the FSD chip is actually quite high since the FSD is a dedicated BPU. Although it has only 144 TOPS, its effective computing power is basically the same as two Orin chips. Its effective computing power is about three times that of a GPU, so the 144 TOPS FSD is basically the same as the performance of two 508 TOPS Orin X chips. At the same time, Tesla had already upgraded to 360-degree cameras. It started using large models to do BEV 3D, which became a 3D vision. The process between the pixel occupancy and pixel movement of the entire 3D vision and the network is similar. One benefit is that the car begins to observe the world in a way similar to people’s observation.
Of course, it still has some differences with humans. The disadvantage is that its pixel is currently only 2 million pixels, while the following cars can achieve 8 million pixels, and human eyes are basically 150 million to 200 million pixels. Its advantage is that you don’t need to look back, you are always at 360 degrees.
With such three-dimensional perception, the judgment and decision-making will be different. The entire judgment and decision-making can be divided into two levels: one level will be white box, that is, programmed, used for following traffic rules and obeying the law. The other level is from three-dimensional perception, to judgment and decision-making, to execution control, forming an end-to-end system. We call it shadow learning. It can clearly see how humans do, what kind of judgment humans make, and what kind of execution they carry out, and this entire process can be recorded.
I believe this is a fundamental change. We call it shadow learning. It is a learning process, and it requires your terminal, including today’s GPU or dedicated BPU, to have a closed-loop ability to obtain the entire process.
The entire feedback growth level is also different, and it will feedback this structural data to the supercomputing platform for training. Although there are algorithms, some algorithms are white boxes, which we believe belong to stage 1.0 or 1.5. The algorithms that are completely black boxes, which we believe are 2.0, are a fundamental change.
Learning on the car side, putting enough samples on the cloud for training, this cloud will do automatic annotation, automatic classification and automatic training, which means that there will no longer be so many people and so much manpower needed. In fact, the number of Tesla algorithm team members is very small, including the Open AI algorithm team members who are also very small, compared to the huge difference in the number of four little dragon AI companies in China, because the automation of large models is completely different.
When the machine is learning, it is not just about right or wrong. Only programming can be right or wrong, and the core is to look at the quality changes. At this time, the training will be carried out on the entire supercomputer side, and it will not be directly deployed to the car after training. In the past, there were also simulation systems, and now it will enter into another simulation method, which will return to the car to do shadow verification, which means validating with a large number of human operating behaviors to see if the quality of my algorithm has been improved. After verifying that the algorithm’s quality has been significantly improved, it will be deployed to the vehicle, because it’s safer.“`
At present, intelligent driving competition in the auto industry consists of three parts:
How to lower the cost of the vehicle’s computing platform and sensors to make it possible to equip all vehicles with sensors, a camera alone is not enough for training, so it is the overall ability of the vehicle. Tesla’s cost is currently very low, including sensors (radar, cameras) + computing platform, Tesla costing $1500 and the ideal being $4000.
Who has more end-to-end closed-loop data? Obtaining only a part of the data is not enough, complete end-to-end training is necessary, and training for different countries and scenes is different and cannot be simply transferred. Enterprises need to sell enough vehicles with sensors and computing platforms to complete this part.
The more expensive and challenging part is the training of the entire large model. Since last year, the US has restricted the development of Chinese training chips and restricted Nvidia from selling high-bandwidth training chips to China. Even with training chips, whether we use GPUs on the vehicle side or GPGPUs in the cloud, it is not the most efficient way for large models, as it only has about 20% efficiency, meaning that theoretically, 80% of the cost is wasted. Therefore, companies like Nvidia need to develop their own BPUs and D1 training chips and use data flow to train large models. After building the entire system, the overall cost can be roughly equivalent to 1/6 of Nvidia’s A100, and the scalability becomes higher. I think this may be the part that AI companies must face, solve, and invest in when they truly want to do AI in the future. This is what we see in the progress of AI development.
We believe that the real competition in the future will appear at the bottom, in the AI operating system.
The AI operating system requires hard real-time, which means that it must be executed at a certain time and place in the physical world. The on-board computer cannot afford delays, as a 0.5-second delay could lead to a fatal accident. At that time, the entire system chain must be a real-time operating system. When running various new large model algorithms, the efficiency of such an operating system becomes critical.
However, it has not reflected in sales. In fact, at this point, we need to find a precise user experience value point in order to fully explode.
At that time, I believed that smartphones could rapidly develop as long as they met three conditions. The first was the 3G network, the second was the launch of the App Store in July 2008, and the third was the OTA technology provided after the release of the iPhone4 in 2010.
Now, when will intelligent electric vehicles truly become intelligent? The terminal is not intelligent, it only extends the functions of the phone. Why extend phone functions? Because Android has built a very good ecosystem, consumers are not using Android for the sake of using Android, but for the ecosystem behind Android, which includes the map ecosystem, entertainment and audiovisual ecosystem, and a large number of application software developers. However, I believe that the era of change truly belonging to intelligent electric vehicles will start with the true 2.0 version. I predict that this point in time, especially for mid-to-high-end cars, will occur in 2024. It will be the era of city NOA based on large models and BEV technology. As of now, Ideal Auto has done relatively well, with usage rates of over 13% for assisted driving. Despite the limited high-speed scenarios, the use rate of our assisted driving has already exceeded 50% in high-speed scenes. However, it is impossible to use in urban areas. I think that a technology will only become something users cannot live without if it is used by the user every day, and the mileage usage rate is consistently more than 60%.
- What will the city NOA look like? In fact, it does not need to become fully autonomous, but it can help the driver to improve the experience when commuting to and from work during congested traffic. More than 60%, even 80% of the cars in the city can be driven by this function, and the safety is not a problem. Unless someone violates traffic rules, even if there is an accident, it is probably just a small scratch. At this point, a huge change will occur. Just as consumers who buy a 20-30 story building, whether it has an elevator or not is fundamentally different.
I think that this year, using Orin computing platforms and other companies, they will all deliver NOA for testing based on large models around Q4. Because it requires rich training, not just putting the computing platform on it. I personally think that by the end of this year, most of the top companies will be able to achieve the level of Tesla at the end of 2011. By 2024, everyone can generally achieve the level of Tesla in late 2022 or early 2023 in North America. I think that after this, at least mid-to-high-end cars, if they cannot provide city NOA, it will affect consumer purchasing decisions. At this point, starting with mid-to-high-end cars, we will completely enter the era of intelligent and automatic vehicles, based on software 2.0 and intelligent electric vehicles. This is our forecast, otherwise we can only sell non-intelligent electric vehicles forever.
This is a simple prediction we’re making, we’re not going to provide city NOAs based on the old way of doing it – city by city. We’re not doing it that way. When we provide city NOAs, we’ll provide this new function of NOAs that covers all cities based on a large model.
- As a 0 to 1 company, the most important thing is the brand because our brand needs to answer questions both internally and externally.
First, we need to answer internally because we haven’t taught our consumers who we are, where we’re going, and how we’re going to get there.
On the other hand, we must provide clear answers to consumers about who we are and what kind of value we provide. I believe this is the most important thing for a brand.
What is our company’s mission? Why did we start this company? In fact, it started from the moment of registration. Our company is called “The Car and Home.” We hope to change the two most important spaces with technology in artificial intelligence and renewable energy – one is the car, the other is the home. So our mission is to create a mobile home and create a happy home.
Second, what is the company’s vision? The company’s vision is what kind of company we hope to become in the medium to long term. We made a plan for 15 years, until 2030, to be committed to becoming a world-leading artificial intelligence company. Whether there are three or five leading companies in the industry, we hope that we will be one of them. We must have the same capabilities as Tesla and Apple, whether it is product functionality, platform application or system capability, in order to stand firm.
Third, what are the company’s values? The company’s values are our unique abilities that we possess and that continuously help us succeed. Generally speaking, a good company value is not just about making up for what is lacking, but after a certain validation period, we can extract what we actually possess. At the time when Qin Zhi joined in 2007, Autohome had already made it to industry first in terms of product library, second in terms of forums, and third in terms of information. But Qin Zhi asked, “Why were we able to achieve this in such a short period of time? What is behind it?” So we found all the veteran colleagues from the beginning, and we wrote on the blackboard why we could achieve these things. Each one had a deep story behind it. At the time, we decided on three values: put consumer interests first; do the right thing, not the easy thing; achieve 60 points before going for 100 points.Actually, it’s the same when we come back to here. At the beginning, when we were doing the ideal ONE, many people asked why we needed to make a family car, why we needed to make a six-seater, why we needed to introduce OKR, and why we needed to use the Toyota Business Philosophy (TBP). Behind these questions lies our thinking. We concluded that the core values of the Ideal Automotive are to surpass the needs of our customers and to create the most outstanding products and services. What we emphasize is that we must go beyond the needs of customers, because the automotive industry has a very long life cycle. Just meeting needs can easily lead to outdated products being replaced.
What does it mean to exceed customer needs? In 2016, if we had done research asking customers if they wanted a range-extended electric vehicle, they would have said our choices were meaningless. Then, we asked if they needed four screens, and customers said they wanted more buttons. Did customers need a six-seater vehicle? At that time, there were no six-seater SUVs, only MPVs with two middle seats. Customers would say why not buy a five-seater or seven-seater and spend less money for one less seat. I think this is what we need to see – customers don’t know, but they may have these needs.
Until the customers drove the six-seater SUV and found that getting in and out was so easy with their elders and children. When they used multiple screens, the front passenger would find that the long-distance travel experience had become so good. After using the extended range, customers found that it could be as convenient as a fuel car, but the entire driving experience was electric, so more and more people would follow.
Do the right thing, not the easy thing. Why do we see that many things need to be thought through first? Because the automotive life cycle is too long, in the Ideal Automotive, we have to make deep insights and thinking before we take action. Doing the right thing is more important than taking action immediately. Taking action immediately is easy, and many of our wrong choices are caused by taking action immediately.
Solve all problems in a collaborative way. Why do we need OKRs and IPDs? Because collaboration is everything. It’s not only important to cooperate with oneself but also to cooperate with all partners. Products need to cooperate well with technology, manufacturing, and service to be effective. This is our core value.
On the other hand, who are the consumers that we are targeting? What value do we provide?
Our core positioning is a luxury electric car brand for families, which includes two key elements: luxury, like BBA, is only for users with a price tag over 200,000 RMB. What does family mean? 79% of users with a price tag over 200,000 RMB are family users. It’s like how Mercedes-Benz is committed to luxury, BMW is committed to driving, and Volvo is committed to safety. We are not the pioneer in the electric car industry, so we need a distinct image to do this. Of course, I also believe that doing a family car is what we are good at, because the Ideal Automotive Founder team has an average of three children in their families. So, we know very well about the six-seater and the scenario of traveling with multiple people, we can understand it from personal experience. As we could not find any products that could meet our own needs, we first chose to make this product ourselves, and all partners trusted this thing.# The entire brand philosophy
The entire brand philosophy, both inward and outward, must be clear and communicated every day in order to unify all resources and energies, as well as to make everyone aware of how to make
conceptual choices. In this industry, where choices are crucial, it is vital to know what we want and what we don’t want. If a company lacks direction, clarity or makes bad choices, you will find that
many people are pursuing their own interests rather than producing what users need or what is of value to the enterprise. Our brand philosophy is essential to diagnose every aspect of our capabilities and face the challenges and competition we must overcome as a smart electric vehicle company, from zero to one, or from one to ten.
As a top-notch product in the industry, we deliver over 300,000 to 400,000 vehicles to our customers every year, and must ensure its safety, maintain its resale value, and provide superior driving experience, which sets us apart from many fast-moving consumer goods as an industry jewel with exceptional professionalism.
Our self-observation and self-diagnosis
How do we come up with our self-observation and self-diagnosis? We created a cross-constructed model, starting with the brand. As we all know, there are over 300 new car makers, yet only about 1%
forms a basic qualification in branding. Most of them lack basic knowledge. They often talk about one thing today, such as use for family, and another tomorrow, for instance, 3-second acceleration, and their logic is often messed up. Consequently, it has impacted on their products quality and business in all aspects. This is where most brands fail. You can see that the internal management of failing companies is all messed up, and each team member has their own ideas. Yet, we have made our ideas so clear that even the workers at the Starbucks in our factory district know what we are trying to do.
The second point is products. Once we have a brand, we must have products, as a brand is the head while products are the body. I believe that in the automobile industry, products are divided into
three levels, which are no secret, only that many people do not respect their common sense.
The first level must be safety. We see safety as a baseline. We can do as much as we can to ensure safety, and I can confidently say to everyone that L7, L8, and L9 are the safest cars in the world and far exceed all collision requirements. For family users, it’s not just one person’s safety, but the safety of the whole family. This includes the design of the entire structure. Many things behind the industry are secrets. For example, it only costs a few hundred dollars to use high-strength steel and hot-formed steel on the A-pillars, but many companies want to save the money. When the media disassembles our cars, they can see two collision prevention beams at the front of the car, two more under the wheel arches near the A-pillar, and collision prevention beams at the rear, all made of the best aluminum structure. It only costs about 1,000 yuan more, but some companies will insert foam plastics just to compete on cost. In addition, all our cars’ front seat belts are explosive pre-tensioning, which can secure you in any situation. All our cars come with fully-equipped airbags, including head and side airbags, which is an added cost of only 2,000 RMB. However, most companies are unwilling to spend this money. Safety is a bottom line, and it’s not subject to any budget constraints. It must be done directly. It’s not about meeting the collision test requirements on one side and ignoring the other side; both sides are equally important because my wife sits on the right side and she is as important as anyone else.The second aspect is product value. It includes space, comfort, and configuration, which can be divided into two parts: software and hardware. The core of the software part is that when we are able to self-develop the central domain controller, the intelligent driving controller, and the whole software backend, there is no problem with software. Whatever the user’s needs are, they can be satisfied. Even now, we allow users to act as their own product managers, providing them with a master task. Apple’s “Shortcuts” can only call for app software functions, while our “Master Task” can call for hardware and software together. Is it self-developed? It depends on how many functions the master task can call. There are some functions that the supplier does not allow you to access. Only by self-developing can you call for all functions and combine them together. The most difficult part is the hardware aspect. The hardware of a car is like having 52 playing cards. The more functions you add, the more you may need to remove. It doesn’t mean that you can increase the cost just because you want to do that. For example, many people wonder why L7 does not have linked seats like the Mercedes S-Class and BMW 7 Series, where the seat cushion automatically extends forward when adjusting the backrest. It is because there is another kind of frame structure. To ensure the safety of the structure while using the linked method, whether it is the BMW 7 Series, Audi A8 or Mercedes S-Class, the rear seat backrest cannot be folded down. When a family wants to transport something, it becomes very difficult. At this time, you have to make a choice: are you serving the boss or serving the household users? This kind of relationship is easy to manage.
If you want to achieve the acceleration of a car in just over 3 seconds, you may have to sacrifice some other parts, such as energy consumption in medium and low-speeds and the size of your motor. At this point, you may realize that hardware is a trade-off, and how to make this trade-off should be based on your brand and user base, not just for the sake of the existence of the R&D team. At the same time, a complete verification system must be established. At IDEA, it is impossible for me to change anything alone. It’s a very rigorous process, quite different from what you imagine. We need to have PEA and requirement analysis. Our requirement analysis includes how many cycles are needed for the R&D, how competitors do it, what value it generates after going online, and how this value is calculated. Both software and hardware require a complete PEA analysis. Many employees from traditional car manufacturers who come to us to develop products realize that traditional car manufacturers’ products cannot surpass ours, because the work we do is much more comprehensive than that of traditional car manufacturers, from requirement analysis to verification, from SOR (supplier offer request) of parts to the verification of individual components, and even to the verification of whole vehicles. Our entire process is very complete, and no one can skip any stage without performing a full analysis. R&D teams, including the supply chain behind them, must be involved in product development at an early stage. We need to go to suppliers and R&D teams to obtain all key details before we can proceed with developing a new feature. This is the key to good product development: knowing how to make trade-offs and establishing a complete verification system.### Third, it’s about the special brand appeal.
We want to create a mobile home, so we can’t just skim the surface. If we want it to be like a home, we need to have the biggest and most comfortable space. If we want to make it intelligent, we need to have more screens because everyone at home has a screen. Otherwise, users won’t perceive the value even though you think it’s good.
On the other hand, if we want to have intelligent driving, we must use laser radar and dual Orin. We will make NOA standard on every car, this is the core. We must make excess investment and configuration. Otherwise, users won’t feel the value.
These are the three levels of the entire product: safety level, product value level, and product appeal level. You must believe that consumers can fully perceive these things. It’s not fake at all. Just don’t focus on non-target user groups. Real consumers can understand and identify with this. This is how we deliver a good enough product to users.
26. Why is gross margin so important?
It’s still related to the automotive industry. If you want to be an automotive company that stays at the table, you basically need billions of dollars in revenue. If you want to become a world-class enterprise, you need tens of billions of dollars in revenue. You can’t rely on financing forever. When you’re small, financing can cover tens or hundreds of billions of dollars in revenue, but as you grow bigger, how can you rely on financing？
Through the development history of the automotive industry in recent years, we can see that even when cars are the most difficult to sell, such as during economic crises when financing is hardly possible, even large companies such as General Motors and Chrysler went bankrupt when car sales dropped by roughly 40% in 2008. Besides revenue, too much leverage was used, including breaking up companies like Delphi and carrying all the leverage by themselves. Therefore, almost 40% of the decline in sales fell on them.
Gross margin is the sale price minus the cost of sales. The cost of sales includes the BOM cost of the car (bill of materials cost), the manufacturing cost of the car, the corresponding cost allocation of the car, the transportation cost of the car, the software cost of the car, and taxes. Employee expenses in the store are called sales expenses, which are outside the gross margin. The remaining part of the gross margin is for responding to sales costs. So the gross margin equals sales revenue minus cost of sales. After subtracting the cost of sales from sales revenue, the remaining money is what a company can invest, which is the part that enhances its ability.As an intelligent electric vehicle company, we believe that a healthy threshold is twenty points, with our research and development investment being mostly above ten points, our sales management expenses being well-managed at seven to eight points, and also bearing certain risks and capital investments, such as building factories.
We believe that twenty points is relatively healthy, and currently, Tesla is above twenty points, we maintain a stable above twenty points, and BYD is also above twenty points. BYD’s sales network is not directly operated, and if the dealer expenses are included, the gross profit margin of BYD cars is also above twenty points. We believe that this is relatively healthy because how can we invest in research and development if we don’t make profit. Automotive industry plans must be made for a period of over five years, and gross profit margin is a crucial point in these plans. Of course, the higher the gross profit margin, the more difficult the enterprise management, and the lower the gross profit margin, the easier it gets, and negative gross profit margin is equivalent to giving away money, which is the easiest way. What we see is that twenty points is the minimum gross profit margin for an enterprise’s long-term healthy development. You can see that Tesla has been losing money for so many years, but starting from the Model S, Model X, and even the Roadster, the gross profit margin has been steadily above 20%. I can spend a lot of money, but always insist on a gross profit margin of over 20%. Tesla had only dropped to eighteen points something in 2018, which was the year when the Model 3 was difficult to produce. Although Tesla has lowered its prices significantly this year, he said at the annual report meeting that he still maintains a gross profit margin of over twenty points this year because the enterprise can only develop healthily in this way, which allows us to make long-term investments in the enterprise, and enables us to be fearless and determined when investing.
After having the gross profit margin, how should we spend the money? This has two parts. The gross profit margin represents our ability to generate funds. The first is research and development investment, and according to different stages, reference should be made to intelligent phones and Tesla.
The second layer is platform-based research and development. After the new L series, we made the most important decision in 2019, which is to fully devote ourselves to platformization. Traditional platformization only means sharing the chassis, but our platformization is more extreme, basically done in the way of the iPhone. We divided the entire platform into 4 parts, including range extender electric platform, high-voltage pure electric platform, intelligent driving platform, intelligent cockpit platform, and auxiliary platform such as electronic electrical architecture platform. They are completely universal. Why is our OTA efficiency high? Because we only need to match one car, not four or five cars. This avoids what happened in the early stages of smartphones. If you give the Apple iOS operating system to Nokia, it wouldn’t work because it basically goes crazy matching thousands of different screens and processors.
- Who has been doing platform transformation in the past? It is not traditional automakers but rather super suppliers. Almost every PHEV system for European automakers is a supplier’s solution, including autonomous driving. However, the range extender system we desire cannot be made by suppliers, so we have to do it ourselves. This includes the range extender, all controls, and intelligent driving.
The results of outsourcing to suppliers are evident for all to see, especially when it comes to large-scale model training, which is simply impossible. A supplier’s L2 assisted driving is likely to require an investment of 500-1000 million US dollars, and if you want to do large-scale modeling, you will need an investment of 1-2 billion US dollars. They may sell it to you for 100-200 million, and continue to provide you with parts. This is their business model, and the benefit is that they can find 20-30 companies to share the costs. However, if we want to do it ourselves, it will cost billions of dollars to do an excellent job on the autonomous driving technology and to completely master City NOA, which also includes training. Therefore, in the early stages, the ability to invest in such platforms is critical and it depends very much on your overall gross margin and financial structure. Do you have the determination to invest continuously?
However, once we scale up, the efficiency will be different. Our update speed will be faster, development efficiency will be higher, supply chain management will be easier, user experience will be better, quality will be better, and costs will be reduced. To be honest, L7, L8, and L9 are all the same except for the body and seats, just like buying iPhone 14, iPhone14 Plus, iPhone 14 Pro, and iPhone 14 Pro Max, all of which have the same components. Platform development basically accounts for 7-8% of our costs, far higher than traditional automakers’ R&D expenses.
The third issue is system development. How to continue to reduce the cost of our platform has a deeper layer, including whether we are using someone else’s system or developing our operating system and the inference chip behind it. If we do the inference chip ourselves, we can achieve a cost similar to Tesla’s, because we have the algorithm in our own hands, including the entire training platform and training chip development.
After research and development, the ability to deliver is crucial as the automotive chain is too long and not only complex but also has a long chain and a long cycle. The cost of trial and error is extremely high, to an unbelievable extent. If a product is delivered incorrectly, the company’s valuation drops by 80%, which is normal. Because everyone is playing the game under the same rules, we must deliver products to consumers, not just develop them. This requires many abilities to be heightened. First is the enhancement of business capabilities, which includes all services and sales-related. Some people say that marketing is not important, but it is very important. Why is it important? Because you have so many people and you have spent tens of billions in R&D costs for the past three years, working tirelessly to make a product. Marketing decides whether you deliver 100% value to consumers or 30% value to consumers. So this is not just the money for marketing, it is the money for the entire research and development. However, most companies choose to give a two-to-three-fold discount and easily sell the product so that they can’t even explain the most basic product. I think this is a very serious problem, which also includes the entire user service process. For example, when we sell a product, we use a central management method, but now that we are selling more products, we are starting to have “provincial governors”.29. We will go further than traditional car manufacturers, not only to the level of regions, but also to the level of provinces, where the governor can allocate all key resources. The challenge lies in how we can effectively manage our direct sales model despite three major obstacles.
The first challenge is cash flow, which has traditionally been put pressure on dealerships. The second challenge is speed of store openings. Finally, we need to avoid the “big pot meal” problem and construct an incentives model.
All of these require us to constantly improve our capabilities and systems. In February, all store employees supported by us achieved a sales average of near seven vehicles per capita. If we exclude management personnel, the per capita sales will exceed 10, which is about three times that of our new energy counterparts. This means there is an opportunity for our employees to earn the pay of three people for the work of two, thus making them hard to lure away. This also applies to our charging network, where we are looking at the strategic necessity within the entire structure, rather than calculating it individually. It is a must-have and must be built to a certain level, such as building all service stations within a province along a particular highway, instead of building them sporadically. This is an upgrade in terms of business capabilities, which also includes supply capabilities. One reason for our fast climbing speed is because of the three electric systems, which is also the most critical barrier. We have divided all parts of the car into four categories. The first is traditional parts, such as rearview mirrors and bumpers. The automotive industry has an annual production capacity of over 20 million vehicles, which can be used for both electric and fuel-powered vehicles, so we do not need to produce them ourselves, as we can rely on suppliers instead. The second category is new electronic and chip related components, such as our domain controller. China has a very good OEM system and procures about 40% of chips worldwide. The manufacturing efficiency is high, and companies such as Foxconn can produce 200,000 to 300,000 sets of domain controllers, which is not a complex issue to resolve. Everyone talks about the shortage of chips, but there has never been a shortage of Qualcomm chips and computing platforms, nor of Nvidia’s chips and computing platforms, or of Horizon’s chips and computing platforms. The shortage is always in the small functional chips, which China is very mature in. It is the world’s strongest electronic OEM factory.
The third category of problems is more complicated and concerns the three electric systems. For example, when we developed the L series, we needed three electric motors, two drive motors, and one generator. If we sell 300,000 cars a year and order 900,000 engines from any particular engine factory, they would not be able to cope, as that would mean building 10 factories. If we failed to purchase them, they would be left with the stock. Interestingly, when these suppliers face a new force, such as ours, the number of units they actually produce differs from the number we requested. They have their own analysis system and will only produce the amount that they believe we can sell. When automakers only sold 10,000 cars in the past but told suppliers they would need to produce 30,000 vehicles in the future, the suppliers did not believe them. At the end of last year, all suppliers believed we could only produce about 150,000 cars this year. It was only after we proved we could produce that amount that they raised their estimate to 250,000. Behind this situation, the three electric systems, including range extender, are produced by us ourselves, otherwise our suppliers simply cannot keep up. Therefore, we built our own range extender factory in Mianyang, and our L-series inverter-motor is produced near Changzhou. We produce our own silicon carbide modules for 800 volts and have experience in producing our silicon carbide motors. This is all part of our supply capabilities. In addition to factories in Changzhou, we will build a factory in Beijing that produces purely electric vehicles. Supply capabilities need to be planned 24 months ahead, otherwise it will be too late. This includes acquiring land, building factories, recruiting personnel, trial production, and basically requires 24 months. Cooperation with suppliers is similar. If we approach a European manufacturer who produces only 30,000 to 40,000 air suspension sets per year, and talk to them for a year, they might only provide us with 60,000 sets at most. However, we need hundreds of thousands of sets per year, which is why we have to work with suppliers in China to discuss building production lines and factories, ensuring that second- and third-level components are in place. The supply chain needs to be very well done, something we didn’t do well in the past year, but in some ways, our forward deployment has brought us great help.30. However, there is a problem with the supply chain, which is that the self-sufficiency rate cannot be too high. All companies in the world have experienced this, such as Toyota separating Aisin and Toyota Textile and sharing them with others, and General Motors separating Delphi. We know that many component suppliers are also separated. Because if you do full self-sufficiency, it means that when sales decline by 40%, you have to bear all the leverage and costs in the 40%. At this time, we are particularly afraid of encountering major economic problems such as an economic crisis. Therefore, we believe that a self-sufficiency rate of 30% is relatively healthy, and we can cooperate with suppliers. Even if we develop ourselves, we can still hand it over to suppliers for production, rather than producing it ourselves. This is the supply capacity part.
The third biggest challenge is organizational capability. Because cars are too complicated, especially smart electric cars. We are a car company, but we are also an internet company. We are a software company, an artificial intelligence company, a business retail company, and a manufacturing company. So the scope of work for all new forces in car making is even more extensive than traditional car manufacturers, such as new forces generally doing their own direct sales system and taking care of dealers. Therefore, organization has become the biggest challenge. Also, because Qin Zhi joined and brought excellent organization to Autohome from the beginning, we have attached great importance to organization. Organization determines our efficiency and whether it can bring effective positive feedback.
- We believe that organizations will encounter three different types of problems. The lowest-level problem is laziness, which is a personal problem. The next level is inertia seen from a management perspective: This is how I did it when I made a fuel-powered car, and this is still how I do it now. For example, what problems did we encounter at the time? Because the people we recruited at the beginning were doing cars worth 100,000 yuan, they also did it that way when doing cars worth 300,000 yuan. What did we do at the time? We bought a bunch of really good cars and let everyone drive them every day. Do not just take out the ruler and go after getting in the car like in the original manufacturer. The car must be driven every day for more than 3 months before discussing how to adjust the suspension. It is not right to adjust the driving experience of a 300,000 or 400,000 yuan car from a full-size or medium-to-large SUV to a compact SUV. We need to try and see how the BMW X7 is calibrated and how the Mercedes-Benz GLS is calibrated. We need to break their inertia. The highest level is if we want to solve business-level problems, the biggest challenge comes from ignorance, which is not knowing what you do not know. The complexity of this becomes even higher.So, how do we solve this problem? When we talk about professionalism, we are facing three levels of the entire business. What does professionalism mean? I can do things well by managing myself properly, which is called professionalism. What is management? It means that I can not only do things well, but also lead a team well, whether it is a project or a department. At this time, we see two things, one is how to do things well, and the other is how to use people well. The next point is even more difficult, which is called management. To do things well, there are two resources outside of it, one is how to use people well and the other is how to use money well. Because at the management level, we see costs every day, so we compete for costs every day. If the cost is low, we can allocate more resources. But when we reach the management level, it is different. We see resources. How to use resources effectively and optimize our own resources through creating more revenue is different.
So, I think the operation of enterprises is constantly improving from 0 to 0.1, from 0 to 1, and then from 1 to 10. The problems we face are also different. What are our challenges? How do we turn laziness into proactivity? How do we turn inertia into necessity? How do we turn ignorance into cognition? There are two ways, and the best companies in the world all do it this way. The first way is the process. What is the process? The process is the product that a company needs to operate internally. Actually, all processes in the world are designed to solve these three types of problems. For example, when we solve specific business problems, we have professional processes, such as PEA and user needs analysis. If we use this tool well, we will become very professional. Good processes don’t tell us how to do it, but they tell us not to skip these things. When we need to do user needs analysis, we need to look outside to see what the current state of this function is, analyze and find data, and what the actual user looks like. We need to plan the product, what it looks like, what functions it has, how long it takes to develop, and how much development costs. We need to plan for operation, so that it doesn’t become a zombie product after launch. It needs to generate value, have activity, have usage rates, and there is also a review process, which is how to turn these things into our own capabilities, product capabilities, and team capabilities. The core is not to skip these important things and do easy things because everyone just relies on experience to do it. I think this is its benefit. As long as we follow this thing to do, we will become very professional. We will find that it is always providing a cycle, that is, the better the product we make, the more satisfied the users are, and the market position will naturally become proactive.
32. What are the most common problems that companies face? Providing low-dimensional processes and tools, only providing employees with a very professional tool to manage results, is irresponsible for the company. Because the corresponding abilities and dimensions are not provided at all, why should the organization be upgraded? Including management tools, after having a business plan, how to analyze what kind of people are needed, you need to first see what kind of organizational structure and employees are used by peers in the market when doing this business, even what kind of salaries are given to each position before planning. Design what kind of position according to your own business process to meet the business plan, and then HR will give you much clearer work. Because, why did this candidate appear here, what are other companies like, how much reward should he get, and what kind of specifications to recruit are all very clear. So, don’t skip these things, so that everyone can train each manager with very good human resource management skills.
33. Going up to the management level, we have learned a lot of first-tier processes from large enterprises, such as IPD, DSTE, IPD is integrated product development management, IPMS is integrated sales service management, ISC is integrated supply chain management, and these processes affect the management results. How do we allocate money and human resources effectively and measurably? Why measurable? The purpose of measurability is twofold. First, in the process of doing it, ensure that the resource is really in place, rather than applying for it temporarily. That is, if we determine to do this, we must really give these resources, and you must use them. The second is that you can effectively review, that is, the difference between my initial planning, the use of these resources, and the final actual results. Doing well will reinforce and tame it into an intelligent ability like intelligent driving, and doing poorly will improve it next time and do it within the plan. We must first solve the problems of these capacities.
Therefore, we will find that behind all the world’s most advanced processes, there are exactly the same as the processes, tools, IT systems, and various methodologies we talk about, all of them are pre-solving problems at each level, solving laziness, inertia, and ignorance problems. Whether it is the tactical level we just mentioned, our own product demand analysis, including our use of Toyota work methodology; going up to the talent structure planning (LTP) we use, the talent structure planning, and the first-tier process like IPD, IPMS are all the same, just used to solve problems at different levels and in different professional fields. So, at this point, I told the team a question, the difference between biology and biology is the difference in information acquisition ability and information processing ability. I think the biggest difference between organizations is the difference in tool use capabilities. With three-dimensional tools, we can easily crush companies that only use two-dimensional tools. This is the core behind what we see.34. As we all know, process transformation and organizational change are not easy tasks. Only after you have sorted out these things can you discover their necessity. Only by doing so can we solve the problems, improve the efficiency of the supply chain, motivate the sales team, and turn each store manager into a business operator rather than a professional. When facing different businesses and different levels, we know how to adapt. The benefit of this is that when we were in the 0 to 1 stage, we focused most on management. Therefore, when we look at the enterprise with a scale of thousands of billions or even tens of trillions of revenue, we can see from a two-dimensional world to a three-dimensional world. What do we see? We see that they are complex and rigid. But when we upgrade our ability to the three-dimensional world, we realize that those things are simple and rich, which is different from what we see in the two-dimensional world. Because the principle is just like this, all the process tools we do first step how to do external perception, we need to know the opponents and what various counterparts are doing, and what are the challenges they encounter. We need to be familiar with ourselves before we can enhance our cognition. The second stage is to set goals and then make plans, not just doing business, but also looking at whether the organizational ability and supply ability are sufficient. Because when we are doing recognition, we can see the gaps, so we should not just lower our heads and do business. How to allocate resources effectively, such as human resources and financial resources, is the second step of the process. The third step is execution and coordination. It is about how to use resources with certainty and how to coordinate effectively. When you find that the first two steps are done very well, coordination will inevitably arise, because no one does not want to do efficient things, things with clear goals, and things with clear resource allocation. The fourth step is that all these good processes and tools will be reviewed, and the capacity we achieve will be settled to the team, the process and the company’s products. This ensures that the next step to do something will use this as a starting point, rather than a random action.
- Another issue to solve is human resources. The most important problem that human beings need to solve is their mental state. Ideal Auto Group actually started from Automotive Home. At that time, Qin Zhi introduced seven habits of highly effective people. Some colleagues from Automotive Home were learning it. After we arrived at Ideal Auto Group, we delved deeper into it. Therefore, we directly cooperated with the company that possesses seven habits of highly effective people to tailor courses for ourselves. Many friends may have read the book “Seven Habits of Highly Effective People” a long time ago. These seven habits of highly effective people actually correspond to the three levels of profession, management and operation to train ourselves. It is a very effective process methodology system. Its first three habits train us to become adults, ensure that we are in a positive state, know what we want, build goals, and effectively arrange our thoughts. The benefit of this is that we can manage ourselves first, handle our relationship with ourselves, and then become adults, so that when we have problems, we do not look for external reasons, but look for internal reasons. Based on this, everyone is willing to cooperate with you and work together with you. This is the first step of the seven habits of highly effective people.The second aspect is how to become a qualified manager. It is our obligation to understand others, whether you are leading a project or a team. As a manager, we must not say “why are my employees different from me”, we say that the value comes from their differences. If everyone was the same, managers can just do the work themselves. Therefore, understanding the team, every member of the team, and everyone cooperative is our duty, so that we can help everyone to overcome their weaknesses and complete the common goal. Only then can we become a qualified manager.
In addition, at the higher level of solving business problems, the complexity becomes higher. I talk about this in every internal training because this world is too complex, and it is crucial to know the “why” behind it. And then, we will be united.
The three corresponding abilities, when we become adults and have professional process tools, we can definitely solve the problem of laziness. When we become qualified managers, able to understand others and handle relationships well, we can surely do good management and solve the problem of inertia by seeing the necessity. When we become leaders, able to let more people support and understand us, and create success together, we can become qualified business managers, use business tools to solve various ignorant problems, continuously improve the team’s cognition, become the team’s ability, and become the team’s harvest. Therefore, this is the core concept of our organization.
Finally, let’s end today’s sharing with this picture. With such a cross, not only can we diagnose ourselves, but we can also see the advantages and mistakes of our competitors and how to avoid them. We can learn from their advantages, including brand, product, gross margin, research and development, including the difference between our product research and development and that of others, the difference between our platform research and development and that of others, and the difference between our system research and development and that of others. It also includes our entire delivery capability and our business capability compared to these advanced enterprises, our supply chain capability, and the difference between us and Tesla and BYD. And our organizational ability, compared with those leading companies with tens of billions or even hundreds of billions of revenue, then we just need to learn from them.
In the past three startups I have experienced five stages, and now that we want to challenge tens of billions of revenue, we have gone through five stages. The first stage is that we always feel that we are different from others, and we never look outward, we are hiding in a well. In the second stage, when we look outward, we don’t understand those complicated processes, pre-research analysis and so on. So, the second stage is called looking outside but not understanding. In the third stage, many enterprises will find some senior consultants or companies to help us understand in order to solve problems. So, in the third stage, we gradually understand. In the fourth stage, we immediately do it after understanding, which is a problem. Why? Because it’s not like you can recruit a few employees from a good company and start doing it. As we often see, those most advanced processes are either done by multinational overseas companies, or by a group of people in that company in that stage. So, the ability of people who plan roads, repair roads, and operate roads is different from those who just drive on the roads. Often we find a group of people who just drive on the roads to help us repair the roads, or we have no planning and start repairing the roads, so this is the fourth stage that we will fall into many pits. In the fifth stage, we know very clearly which consulting companies and even which team this company used for the planning, repair and operation of this road in the past. We recruit them and slowly learn to plan and repair roads ourselves. Therefore, at this time, the team will be driving up there and transporting value efficiency will become higher.## Q&A Session
Attendees: Li Xiang, Chairman and CEO of Ideal Automotive; Ma Donghui, President and Chief Engineer of Ideal Automotive; Li Tie, CFO of Ideal Automotive; Xie Yan, CTO of Ideal Automotive
Media Question: I have attended several auto companies’ press conferences recently, and I feel that many auto companies may not hold press conferences after the epidemic, at least not as expected. However, I benefited a lot from Li Xiang’s sharing today. I have two questions. First, in January this year, everyone knew that the entire car market, including the new forces, did not perform well. However, Ideal Automotive was an exception, with a delivery of 15,000 vehicles. I want to ask whether we are satisfied with this contrary growth internally? And what is the reason for this contrary growth? The second question is for Mr. Li Tie. We saw from the Q4 financial report that we still maintained a relatively healthy gross profit margin of 20.4%. This is actually a milestone. I understand that it should be the first single-season profit among the new forces, because our operating profit is positive. I want to ask, what is the reason for this? Another question is about sustainability. Will the profit be short-lived or a real turning point? For example, can we maintain such a good phenomenon in the first quarter of 2023, and what is the expectation for the whole profit turnaround? Thank you.
Li Xiang: I’ll answer the first question. I think the first quarter is really difficult, and there are two reasons why the first quarter is difficult. The first reason is that the sales in this year are overdrawn in December of last year, because there are two situations in December. On the one hand, there is a subsidy rollback for new energy, and on the other hand, the 50% purchase tax reduction for fuel vehicles is cancelled. Also, in cities such as Shanghai, where we face a situation where new energy vehicles are no longer eligible for license plates, so many vehicles that should have been sold in the first quarter are concentrated in the fourth quarter of last year. For example, in December last year, we sold more than 20,000 vehicles, including more than 1,000 units purchased in advance by users in Shanghai.
The other reason is price reduction. Because once the price is reduced, Sun Shaojun said it well, price reduction may not help you increase sales, but it can hit other companies, because after the price reduction, everyone will watch their opponent’s company and certainly reduce the price. In the past, there were seldom head companies in the automotive industry, such as Tesla, that would lower prices. Head companies in the mobile phone industry always increase prices, never lower them. Tesla caught us off guard. Including what people said about Tesla’s price reduction, will it wait until March and April to lower the price? Musk said at the end of December, and it was lowered in early January. This speed is very fast and it was lowered globally. These two factors have affected the development of the entire auto market.I think there are areas where we have done well and also lucky. Our competitive edge lies in the strong product competitiveness of our products within each price range. We calculate the competitiveness of the entire product because we believe that sales equals market share, and market share equals product strength divided by product price multiplied by NPS. The stronger the product strength, the higher the sales. Of course, when comparing with others at this level, the higher NPS, the better the sales. After selling more cars, NPS will naturally improve. On the other hand, the lower the price, the better. If we want to increase sales without changing product strength, the best way is to reduce the price.
In the second generation of products, our product strength was significantly ahead in this price range, which is one of our strengths.
On the other hand, Tesla will actively reduce its price range to 200,000 to 300,000 yuan, so there will be fewer intersections with us because Teslas priced above 300,000 yuan sell less. Once Tesla’s main model is brought down to 250,000-260,000 yuan, everyone will consider Model Y to be a car priced above 200,000 yuan, so purchases priced above 300,000 yuan will become less and less. In the past, however, Model Y was priced above 300,000 yuan. This actually opens up space for us, and it has attracted many new customers to enter the market. Many customers say they want to compare cars priced similarly, which brings many users to switch to us. This is where we do well and also lucky, but I don’t think the luck will always be with us, and we still need to be prepared for everything.
Li Tie: I’d like to add something about the first question. From my point of view, we still have room for improvement, including our delivery capacity. If our business, supply chain, and organizational capabilities are stronger, based on our current product strength and the customer value our product brings, we should be able to have a greater market share.
Why was everyone concerned about January? Because companies always have difficulties when the market is under pressure. I think the entire market will be under pressure this year. When global shipments are down, head enterprises expand their market share. Look at Apple’s computer and phone market share, they are both increasing, which is what Li Xiang said about product value and customer value.This year, we will enter a phase where we pay attention to market share. At the beginning, we entered this market from scratch and operated under unified management of the group. We obtained nearly 9% of the market share in the SUV market of 300,000 to 500,000 units sold last year. Next, we aim for 20%. What does 20% mean? It means becoming a Leading Brand, with a ranking of 1, 2, or 3, which may vary from month to month. I think our next step is to first become a Leading Brand, so the user value brought by our product strength is very important, especially in this economic situation, including the price reduction expectation mentioned by Li Xiang just now. I think it is a solid foundation for us to gain market share. I think this year will be under pressure, whether it is passenger cars or the entire new energy market.
About profitability, I want to share that the past three years were our validation period from 0 to 1, and the five years before that were spent on R&D. We did not have the opportunity to face consumers. The past three years have validated our products and many assumptions, including the business model. We are all following the road of Tesla. At the early stage, we need to quickly understand the product logic and find out where the product value lies for users. In fact, enterprises need to pay attention to nothing but sales volume and Free Cash Flow (FCF), that is, staying alive. What is FCF? It is the operating cash flow. With the increase of car sales, are we still generating cash flow every quarter? It is the operating cash flow. We still need to invest in factories and R&D. If we deduct these two, if FCF is still positive, it means that no matter how the market changes, we are not afraid of financing, because we have our own blood-making ability, which is what we have been focusing on for the past three years.
We will go public in 2020. You can see from our financial report of the past 12 quarters that only in Q3 of last year, which was during the ideal ONE EOP phase, our FCF was negative. So, we basically believe that this FCF is already in the blood of every manager and employee in our company, which I think is not common in Chinese companies.
This year, we will start to focus on becoming a company with a scale of RMB100 billion or RMB200 billion. What should the level of our management be to achieve this? Starting now, in addition to having 20%-25% gross margin, can we achieve a 5% operating profit margin?Our operating profit is simply the gross profit minus research and development, sales and management expenses. Interest expenses, investments, and taxes are mostly unrelated to the daily operations of our managers but are rather associated with the company as a whole. Our operating profit and long-term operating profit are essentially the stock options we give to our employees, which is an uncontrollable variable. For example, when we gave out the options, the stock price of the day was the actual value, and we just have to go with it. Therefore, we first focus on what we can fully control, which is the long-term operating profit of up to 5 points.
In each quarter’s financial report, Tesla not only discloses the indicators of the quarter but also displays them on a trailing twelve-month (TTM) basis, which better reflects the past trend of financial indicators and can eliminate the impact of seasonal factors to observe long-term trends. Basically, if any indicator of Tesla performs well for three consecutive quarters, they will start pursuing other financial indicators to manage the company’s operations.
We are still in our early stages and need to strike a balance between research and development investment and the pursuit of financial indicators.
Media Question: Mr. Li, the current penetration rate of new energy vehicles has reached about 25%, and some believe that the growth rate of the industry is likely to be in a downward trend in the future. Car companies will face increasing pressure to rely on sales to generate revenue. Do you think the pressure on Li Auto is high? Also, how do you view the future survival status or competition pattern of new forces in the automobile industry in a situation of high costs and intensified competition? Thank you.
Li Xiang: I think the pressure is definitely high. The greatest pressure in this industry is that the top two companies are capable of engaging in price wars, which cannot be seen in other industries, such as the smartphone industry. In addition, people tend to overlook a certain issue, which is that consumers have some basic common sense and cognition. As an example, if we, as a Chinese brand, sell a mid-sized B-class SUV for more than RMB 300,000, there will be absolutely no sales. This is a critical piece of common sense.
Tesla and BYD have always placed their prices in a reasonable space, not overly cheap, but rather within a reasonable range that consumers recognize. Many people ask why Li Auto doesn’t reduce prices to less than RMB 300,000. The same question applies to consumers who will not accept a mid-sized SUV from a reasonably good brand for over RMB 200,000.I will privately tell you a story. When we were preparing for the launch of the Ideal ONE, our most concerning competitor was not the Highlander as most people would have thought. Instead, it was a conglomerate that released three mid-large size SUVs on the same platform: Cadillac’s XT6, Buick’s Enclave, and Chevrolet’s Trailblazer. We were afraid that they would price their cars at different levels, with Cadillac selling for over 400,000 RMB, Buick for over 300,000 RMB, and Chevrolet for over 200,000 RMB. If they did, the Ideal ONE would have been in a difficult situation. However, we were lucky that Buick Enclave was priced at just over 200,000 RMB when it was released. We observed the user’s psychological and cognitive states on autohome.com and knew that if a mid-large size SUV was sold under a decent brand, it could not be priced at over 200,000 RMB, because customers would assume that there were problems with the car and that it was made with corner-cutting. This was their perception. Another cognition that existed between users was that if you bought a car that was so big and spent so much money on it, you probably had some financial problems. These were common perceptions. So, sometimes, it was not better to sell cars at a lower price, nor was it better to sell them at a higher price. You can draw a curve that is accurate from 100,000 RMB to 250,000 RMB within any car level, and BYD’s curve is the most accurate. Tesla’s curve is also precise after they adjusted their pricing. Once you exceed 250,000 RMB, it is difficult to sell cars regardless of whether they are high-priced or low priced. Therefore, we strictly followed this pricing curve to construct the entire product system. When we release an all-electric vehicle, we will also follow this pricing curve, and we won’t sell the car at higher prices, even if it is an 800V model made of silicon carbide. This is actually a fundamental issue that many companies face. If we sell mid-size cars for over 300,000 RMB, we won’t be able to sell them. If we sell mid-large size SUVs for over 200,000 RMB, we also won’t be able to sell them.
Media Question: I have two questions. Firstly, based on last year’s profits, I want to confirm whether you predict making a profit in the first quarter of this year and even for the entire year? You talked a lot about lowering prices just now. If we can maintain a gross profit of over 20% and the operating profit margin is over 5%, could the extra profit be shared with users? Is it possible to reduce prices? Secondly, we recently learned from the end-users that Ideal has some cars in immediate stock for sale. Why is this happening? Are there any issues with the orders? Could you please explain? Thank you.Li Tie: At the Tesla event this morning, we talked about exploring some long-term values. We hope that this year is the beginning of a continuous and stable profit, which is our internal goal, and we also believe that it is what a 1 trillion company should do. We talked about operating profit first, and there are many restricting factors below.
In the long run, we have always focused on efficiency. You just talked about gross profit, and I think Li Xiang mentioned it too. In fact, it is mostly due to competition. As long as we can get the quantities we want and drive our production capacity, Elon’s biggest purpose of decreasing prices in China is to maximize production capacity utilization. I think this is a comprehensive decision, not just for the benefit of users, but also for the win-win problem of the enterprise and users.
Li Xiang: The second question is about the current car issue. We have been working on the current car issue since last year. Unlike when we sold the Ideal ONE, it was an incremental market, and we created this market. Between 300,000 and 400,000 yuan, there were no six-seater cars, and if people had this demand, they would come to buy from us, and even the sales of the Toyota Highlander decreased. After we introduced multiple products and entered the 400,000-500,000 yuan market and provided two cars between 300,000 and 400,000 yuan, we encountered a problem. We will compete for the stock market of BBA. The quantity we add will be the amount they reduce. At this point, we found that almost 100% of users who buy BBA cars have current cars at this price. A considerable proportion of the user groups want current cars when they come. Therefore, we started from last year to ensure that about 10%-15% of the quantity is current cars. These users are not so concerned about what color and configuration I must have, but more concerned about current cars.
We will not place current cars in the store. We will place them in each region and allocate a certain proportion of current cars according to the needs of users in each region. This will be our long-term practice, because we must grab market share from the stock market of BBA, especially L7. Basically, if our L7 grows, models like the Audi Q5 will definitely decline and cannot escape. Because this market itself has no growth, it is actually a zero-sum game, and what we gain is what others lose.
Li Tie: Investors have the same problem. Li Xiang just mentioned that we used to operate centrally with one flag throughout the country and achieved a 9% market share. Now, if we want to achieve 20%, overall, we need to sink the management and operation focus to decide where the budget should be spent. We manage all cities by province because provinces are completely different from each other. Also, we need to get at least a 20% share in each city, and our share in some cities has reached 30%. This requires different strategies. We now want to reach 20%, and some cities need more than 30%. We need to see how many small losses there are and add them up to achieve 20%. In fact, the more we go to first-tier cities, the more we will find that store managers complain that they lose many customers every day because they come in and ask if there are current cars, and if not, they leave. You cannot challenge everyone’s common sense, although you tell them that Tesla is also like this, but anyway, some people just leave when they come in. So we give them this right to choose.Media Question: I have two questions. The first question is about how do you see the position of Li Auto in China’s ADAS market? We briefly discussed this earlier and I hope you can provide a more specific introduction. Regarding software 2.0, what are the specific developments so far starting from ADAS business? The second question is about the 30% proportion mentioned earlier. We may have established joint ventures for supply chain security for components such as range extender, silicon carbide, and electric drive. But in terms of self-development, continuously taking over core components, what are the specific strategies we will have in the future? Thanks.
Li Xiang: Let me answer the first question. I think doing urban NOA and high-speed NOA in this large model, including high-speed NOA in the large model, must be solved after urban NOA is completed. I think this exam question is very easy, open-book exam, and there is only one exam question. Everyone’s bragging is useless. So how do you measure the value to the user? It is the proportion of user usage time, that is, how much proportion of driving time do users use navigation-assisted driving, which is the real key. This is the same as DAU in the Internet, including the user time of the video website. It is the key. I think no one has started taking this exam yet today. Basically, everyone will start taking this exam by the end of this year.
Ma Donghui: Let me answer the second question. First of all, it needs to be clarified that self-development and self-manufacturing are two dimensions. Just now, Brother Xiang also introduced that in terms of platform technology development, we have range extender, high-voltage pure electric, intelligent space, intelligent driving, and EEA electrical architecture. Whether it is range extender or the five-in-one motor you mentioned just now, they are all core technologies of our range extender platform. In the future, we will also have some self-developed technologies in the pure electric platform, such as silicon carbide function modules, and three-in-one motor. However, supply and self-development are not the same. We will divide our entire supply into several dimensions according to the attributes and categories of the materials, such as leverage materials, strategic materials, conventional materials, bottleneck materials. For different materials, our supply strategies are completely different. Our self-development and self-manufacturing, or self-development and joint venture companies, are mainly to ensure that our core technologies can land and ensure the safety of our supply.Just now, we mentioned that the proportion of self-developed technology must be controlled at a reasonable level, about 25%-30%. One aspect is to ensure that the technology can be applied to the market, and the other is to ensure full competition with the external market, ensuring the vitality of supply and R&D. In addition, the direction of our self-developed core technology is related to our platform technology.
Media Question: I want to ask a question about intelligent driving. As Brother Xiang just mentioned, our Orin’s support for Transformer is not very good. How can we solve the problem of fully deploying large-scale model algorithms in the future? Will we introduce new hardware platforms? If there is a difference in the capabilities between the two generations due to hardware, how can we solve this problem? Another question, as Brother Xiang mentioned earlier, the logic of a car owner who buys a car over 300,000 yuan may be different from that of one who buys a car under 300,000 yuan. We want to expand on a larger scale. When we focus on the market below 300,000 yuan, will the concept of our household brand be diluted? How will we develop the logic of new products? Will we continue the logic from Level 7 to Level 9 or open a new logic?
Xie Yan: Let me answer the earlier question. Orin’s Transformer’s operational efficiency is not very high because it is a universal GPU architecture. However, its efficiency can still be increased through optimization. We cooperate with NVIDIA to perform a large number of optimizations to improve its performance and actual TOPS calculation. It’s like having a platform that is a bit weaker, but putting in more effort. This is the method we want to utilize going forward.
Li Xiang: Delivery of NVIDIA’s next-generation chips for cars should not be until 2025. However, the algorithms can still be extended, and the fundamental sensors are generally directly reusable. In terms of algorithm applications and model training, these processes can be extended even further. As for the second question, I think the challenge of cost is greater in the 200,000 to 300,000 yuan range. Our current scale of about one to two thousand vehicles per month cannot support competitiveness in the 200,000 to 300,000 yuan range. So, we will delay entering this market until 2024 when we can produce 30,000 to 40,000 vehicles per month, which will be more effective. But when we do enter this market, we will not reduce our product competitiveness. We will only reduce the size, while continuing to leverage our key technology and platform advantages, which will lead to higher long-term efficiency.Media question: I have a question about the pure electric planning for the future. Xiao Kang and I were just discussing when the deployment of the 800-volt ultra-fast charging network will begin. Currently, 800 volts are not scarce, but high power is scarce. Behind high power is actually the enclosure of the city area, including negotiations with high-speed service areas, your battery capacity, and the position of the car. These are all difficult. The other ideal that has always been emphasized is to seize profits in advance. For the pure electric W series, the profit margin may be higher, but the absolute quantity is relatively small. Can the gross profits of subsequent operation of pure electric products support the expansion speed of the entire ultra-fast charging network? Or do you think that you can strategically lose money and deploy ultra-fast charging first? I actually want to hear your strategy in this regard.
Li Xiang: I think the cost of ultra-fast charging stations is not as high as everyone imagines, and it also has a five-year depreciation period. In fact, the cost is much lower than everyone thinks and cheaper than building a factory. I can give you a calculation. If an ultra-fast charging station on a high-speed road can reach about 640 to 800 kiloamperes of electricity, a total electricity capacity of 480 kW with three 250 kW chargers will be installed at 3+1 stations. If we use only 480 kW, the entire ultra-fast charging station can only reach a peak of around 800 kiloamperes. If we build 3,000 ultra-fast charging stations by 2025, the total cost will be 10 billion yuan. Spread out over a year, for a company with annual revenue in the hundreds of billions, costs are not as high as everyone thinks.
However, we think an effective way is to use slow charging at home, fast charging in cities, and ultra-fast charging on high-speed roads. In the city, 200 kW chargers are sufficient, and building too many ultra-fast chargers in the city still has high requirements for the property and surrounding environment. We believe that building ultra-fast charging stations near gas stations, such as on the ring road, is a good way to build ultra-fast charging stations in cities. Where there is a gas station, there should be an ultra-fast charging station. However, if we want to build a 480 kW ultra-fast charging station in an office building, the challenge for the property is still very high, and even demanding fire protection requirements like a gas station may be necessary. Therefore, although we can make the power higher, the capacity of the power grid, including that of the city and the high-speed road, is another measure. Our better plan is for using slow charging at home, fast charging in the city, and 200 kW fast chargers, instead of the old 60-70 kW fast chargers, and ultra-fast charging on long-distance journeys. We have confidence in the coverage of long-distance travel. Because the country has been encouraging companies to directly build ultra-fast charging stations since last year, and our built ultra-fast charging stations will also be shared by other 800-volt cars to ensure that every other car of our competitors can run at high efficiency. In the long run, we believe that the ultra-fast charging network will not lose money at normal charging prices, but if we can earn more money, it still depends on our operational abilities. However, at least not losing money is completely achievable.Question from Media: There is a question about autonomous driving. Ideanomic stated that the early bird test of city NOA (Navigate on Autopilot) will be open in Q4 this year. Li Xiang also shared earlier that the coverage of all cities will be achieved in 2024. I am curious whether the NOA feature will become a standard for all vehicles in Ideal City in the future, like it is now. If yes, will it create an imbalance between the investment and the return of the software development? Another question is about the biggest technical challenge when transitioning from highways to urban areas. Will it follow the same path as FSD by integrating highways and urban areas?
Li Xiang: I can give a straight answer. The technical roadmap for city NOA is identical to that of FSD. The key is how to build a high-quality, efficient training system for big models, because it heavily relies on training, with human intervention being auxiliary. For now, only AD Max can run the big models needed for city NOA, as AD Pro was designed for high-speed NOA from the beginning, with limited performance.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email firstname.lastname@example.org.