Editor: Mark
Produced by: Red Star (ID: redplanx)
Header picture: He Xiaopeng picture
Beijing is sweating profusely today, with the highest temperature reaching 41 degrees, which is about to melt people.
I made an appointment with Liu Long, a friend in the automotive industry, at 10 am. I was going to talk about smart cars and went to the mall to experience Xiaopeng Motors .
1. The touch screen era has become a "past tense"
When I walked into the mall and got in the car, a huge "iPad" came into view.
Author: Why are there only two screens? Don’t current electric vehicles all have four or five screens?
Liu Long: This is related to the concept of Xiaopeng . In fact, Xiaopeng is very disgusted with multi-screen and firmly opposes multi-screen. In Xiaopeng's view, there is no need for so many screens in the car. This is an electric car. The number of screens equals the amount of electricity, and the input-output ratio is not high.
Nowadays, everyone is playing with the interior of electric vehicles. For Xiaopeng, these are not important. Although it can attract users in the short term, there are no barriers.
The central control screen looks like a "large" iPad. It has both systems and software applications, but in fact it looks similar but is different in spirit. If you just add a large central control screen and some mobile applications, it will be an Internet car. In Xiaopeng's view, this definition is too superficial. What Xiaopeng values is the intelligent capabilities of the entire vehicle.
In the early days, everyone thought that the large central control screen was the best way to interact with smart cars, but now this way of interaction has basically been eliminated, at least according to the trend. Maybe this is why Xiaopeng is disgusted with multiple screens. A reason.
Author: Why, this way of touching is quite humane.
Liu Long: If you are sitting quietly, the touch method is indeed optimal, but the driver needs to focus on driving.
Initially, everyone believed that touch screen was the best way of interaction. Later, everyone thought that AI vision was the best way of interaction. Now, everyone thinks that intelligent voice is the best way of interaction.
If you include the earliest mechanical touch control, the interaction method has reached 4.0, but at this stage, touch screen still occupies a large interactive market. For Xiaopeng, the most expensive thing is the support of horizontal and vertical screens.
Nowadays, when everyone talks about smart cockpits, they all think about adding entertainment functions. This is actually traditional Internet thinking.
For cars, we should "do subtraction" instead of trying to do everything. I personally think that in the future, voice interaction should be able to control everything except the steering wheel and brakes.
This is the best way to interact.
(Picture of Xpeng Motors’ intelligent driving system)
Author: I heard that the car system often freezes and the screen goes black. Are these problems caused by the system or the hardware?
Liu Long: Xpeng Motors’ current cars and machines are all Android systems. There are also Linux, QNX and other systems on the market. But in China, Android has the richest ecosystem, including music, games, social software, etc.
If you choose a platform like Linux, the number of available applications will be very small, and sometimes car companies need to develop it themselves. Tesla has encountered this problem, because Tesla uses Intel chip (X86 architecture ), so it is more suitable for running Linux systems.
Because the application ecosystem of Linux is not yet complete, in the cockpit area, domestic users still prefer Android systems.
Author: Since smart voice is the best interaction method in the future, how to solve the interference of various noises, such as wind noise, tire noise, power system noise, interior noise, etc.?
Liu Long: Regarding the issue of sound source localization and noise reduction, everyone is working hard to do it. Of course, it is still very difficult, but the interior of the car can still be classified compared to the home environment, so it is relatively easy to deal with.
In terms of intelligent speech technology, Zhao Hengyi was previously in charge. Zhao Hengyi came from Spichi , so except for core self-developed things such as semantic parsing, semantic anti-noise, and natural language understanding, other non-core things were outsourced to Sibichi. Must Chi.
Noise is also the core competitiveness of the electric drive system, which is headed by Liu Minghui, Vice President of Power. Of course, how to increase the usage of speech recognition requires deeper integration with cars.
2. Putting things right
In 2016, autonomous driving began to bloom in China. Pony.ai, WeRide , AutoX, and Momenta were established one after another. He Xiaopeng also felt this wave.
So under the persuasion of investors, he officially joined Xpeng Motors in 2017 and served as chairman. The main reason is that He Xiaopeng thought one thing clearly. There is not much temptation to simply make electric cars and , but if intelligence is added, He Xiaopeng instantly felt that this is a market worth hundreds of billions of dollars.
Author: I remember that Xpeng Motors mentioned a concept called "three electricity and one screen" before, but why didn't it mention it now.
Liu Long: In fact, Xiaopeng also developed step by step. The concept of "three electricity and one screen" was gradually updated and improved. "Three electricity" was refined into electrification, and "one screen" was refined into intelligence.
Xpeng may think that "one screen" is too narrow, and at most it can only represent the concept of a smart cockpit, which has nothing to do with smart cars; and "three electricity" cannot fully represent electrification, because electrification is an overall concept.
Therefore, Xpeng Motors’ promotional slogan has been re-upgraded, focusing on the two main lines of electrification and intelligence.
In order to reflect intelligence, Xiaopeng has successively established an autonomous driving center, an Internet center and an AI product center, which mainly serve data, algorithms and user interactive experience, and support these three new R&D centers in vehicle integration.
Author: So what does Xiaopeng think of the G3 product? Looking back now, I feel that it is very weak. Compared with the current sales data of P7 and P5, it is too low.
Liu Long: I think in the early days, everyone made a mistake. It is believed that the mid-to-high-end market only belongs to joint ventures, and domestic OEMs have no chance. This may be due to too many discussions with fuel vehicle companies.
Another reason is that it is taken for granted that the sales volume of A-class cars is greater than the sales volume of B-class cars .
So everyone positioned it as a compact city car. Of course, this was also related to the environment at the time. On the one hand, it was because of the battery life problem, which was only 300 kilometers at most. Even the new car manufacturers at that time believed that low battery life could sell better.
On the other hand, because users have such a demand, many manufacturers, including Xpeng Motors, want to cater to this market and hope to increase sales. The prices of
models are not expensive either, all under 200,000.
The actual situation is that the demand for this small positioning is not much. The unsatisfactory sales of G3 also caused Xiaopeng to reflect. Then it began to put things right and re-formulated the company's strategy. It should try its best to follow the intelligent, mid-to-high end and avoid other detours.
So you can see that even when the Series C financing was very difficult, Xpeng did not lay off its intelligent project team.
Author: I have also heard Wu Xinzhou say that during the C round of financing, all parts of the company were controlling costs, including the autonomous driving department, but Xiaopeng's protection of autonomous driving was still very strong.
(Xiaopeng G3 picture)
Liu Long: At that time, Xiaopeng faced two paths, one was to be acquired, and the other was to break out on its own.
At this time, Xiaopeng Motors president Gu Hongdi began to say that BAT was not interested in electric vehicles, but they were very interested in user data and operational data, because these related to food, clothing, housing and transportation, which were very important to BAT Technology Company. Said it was attractive.
Xpeng has also begun to rethink the company's product strategy, whether to continue to make a high-end SUV or to make a sedan. In fact, everyone internally believes that it will make a high-end SUV.
However, Xiaopeng believes that everyone is making SUVs, which is already a red ocean, so the P7 is ultimately positioned as a mid-to-high-end electric coupe.
From G3 to P7, Xiaopeng has thought clearly. Making an electric car with a high cost performance (under 200,000 yuan) will earn some attention today at most, but it will have no future.
To make the vehicle smart, we must ensure that the vehicle has a certain profit, and the price cannot be set too low. Otherwise, it will not be conducive to the company’s long-term R&D investment, and there will definitely be an imbalance in some aspects (smart costs). Scruples. Therefore, Xiaopeng set the price of P7 between 250,000 and 300,000. In order to highlight the performance of the coupe, Xiaopeng lowered the battery cell height by 30 mm.
Making a cheap electric car will not produce intelligent capabilities; making a very expensive car will not produce scale, and there will be no later data operations.
If you only make electric vehicles, you may be able to enter this market through marketing, but once this market is really opened, traditional OEMs will swarm in, and the competition will be extremely fierce. If you are not differentiated, you will have a very difficult life. .
At this time, you have to fight for your channel capabilities, marketing capabilities, and . Although you can still be alive, it will be difficult to become bigger, let alone stronger.
This concept was not until 2021 that many traditional OEMs realized that in addition to electrification, intelligence is the soul of future cars.
Author: Then switching from one model to another requires redesigning the chassis and structure. For Xpeng at that time, it should be very stressful and very expensive.
Liu Long: For Xiaopeng, there is no way. Tesla has also made this mistake. Premature personalization has led to very low production efficiency, and its continuous innovation in technology has led to a decline in production efficiency at the beginning. The advantage of the
platform strategy is that it can use a large amount of existing technology, the development risk is relatively small, the quality is easy to control, and a large number of parts can be used, which can also increase the bargaining space with suppliers.
But at that time, Xiaopeng had no choice.
For this reason, Xiaopeng has found a solid conceptual support for himself. He does not want to develop vehicles according to the characteristics of the platform. Why does he say this? Although the benefits of the platform are that it is scalable, low-cost, and reusable, it is easy for the platform to be reused, resulting in slow innovation or even daring to innovate.
Why many car manufacturers only make one model? Maybe they want to make more use of the scalability of the platform and reduce the cost of research and development.
3. Opinions on friendly businesses
Author: What do you think of Internet car manufacturing, especially companies like Xiaomi , Baidu ?
Liu Long: In fact, Xiaopeng can be regarded as one of the earliest representatives of Internet car manufacturing.
I think that if the Internet makes cars or Internet people make hardware, they will basically have a narrow escape. Internet-made cars often ignore the power of the supply chain, underestimate the ability to sell in person, and place too much emphasis on online sales.
Internet companies and traditional automakers each have their own advantages and disadvantages: Internet companies focus on R&D and operations, while traditional automakers focus more on manufacturing and sales. If Internet people ignore this logic, they will find that sales cannot increase no matter what. If sales cannot reach scale, it is of course impossible to operate and generate subsequent value.
So my advice is that in the early days, we still focus on selling cars. The first thing is to make high-quality cars reliably, open up the user base, establish the brand, and then think about cars. s things. And judging from current market feedback, consumers still value the traditional value of cars more.
Whether it is a car built by the Internet or people from the automotive industry, everyone must not have shortcomings. From design and R&D, to supply chain production and manufacturing, to final after-sales operations, everything needs to be connected.
Therefore, I often see that some companies have been established for several years, but their sales seem to have no change. It is because they have put too much effort at a certain point and have not planned and built a good system for car sales in other aspects.
They may have done a good job at a single point, but because the system was not done well and perfected, and they did not wait for large-scale batch delivery, they basically failed.
The last point is also the most important, and it is also the most easily overlooked aspect of Internet car manufacturing, which is that mature technology and traditional craftsmanship should be respected. Internet cars are not equivalent to concept cars with weird shapes. Models and components must be designed to meet engineering constraints.
(Xpeng Motors Management Team)
Like Xpeng, it initially borrowed a large number of Tesla’s open source patents and independently developed a set of very safe liquid cooling technology while utilizing 70% of Tesla’s three electricity and one screen suppliers. , and gradually established the entire supply chain system. I think for Internet car manufacturing companies, the difficulty in delivery is whether the "three electrics" can be done well and whether the intelligent driving software is stable.
Therefore, mature technology should be used, and it must be used. Because of a serious safety incident, these Internet car manufacturing companies may be wiped out in half.
Author: Regarding the service system of Internet-based cars, such as NIO , how does Xiaopeng think about it?
Liu Long: When talking about Weilai , we have to mention Haidilao . The Haidilao system is very successful, but it also has great limitations. For example, it is a very good system among Chinese people, but it is not clear whether it is a very good service system abroad.
I can talk about the service system of traditional cars. Why is there an offline 4S store model but no online sales? First, they are afraid of online price impact. Once unified prices are opened online, it will inevitably impact 4S stores; second, it is difficult to close the data loop with offline data alone. Although 4S stores will also generate a lot of data and provide it to the OEMs, but Much of this data is secondary data.
But there is one thing we cannot skip. As a big-ticket commodity, cars are a must-have link from understanding, to test driving, to decision-making and purchase. If you skip this link, the conversion rate will be close to 0.
For Internet car manufacturing companies, the investment in infrastructure construction expenses, basic research and development expenses, basic sales network, and basic maintenance network is very large. But there is no way to pass these costs on to consumers.
Internet car manufacturing, including traditional car companies, will definitely follow a similar path to ours, including establishing intelligent centers and user operation centers, but the overall cost is very high.
Author: Xiaopeng mentioned before that smart car operations are greater than manufacturing. What do you think of this?
Liu Long: I agree, but this has to be based on cars. Without hundreds of thousands of cars, talking about operations is empty talk.
Without matching user operation capabilities, future automobile products will not be called smart cars at all. They will just be covered in the shell of a traditional car with an added function.
Xpeng is still user-centered, making some functional upgrades and OTA, but these operations are based on the user's way of thinking to solve scenarios, and are not done for the sake of showing off skills.
The core of smart cars is driven by operations, OTA, and data. Now slowly, many people are starting to think about this matter.
This leads to a problem. Originally you just wanted to be a software company, but because it requires a lot of underlying hardware for support, and after the software and hardware are combined together, the whole vehicle hardware architecture system is needed to combine, so you will find that the more you do, the better. Deep, it eventually became a car factory, so everyone finally chose to build cars.
Not only Internet giants, but also L4 companies have this impulse, but they are trapped by resources and financial strength and do not dare to do so.
Author: Regarding the cooperation model of Huawei empowerment, what do Xiaopeng think?
Liu Long: Regarding Huawei’s cooperation model, Xiaopeng also did it in UC, but it died miserably later. The first challenge is that each partner will have different demands. It is very painful for you. To complete its things, you need to constantly adapt. Each partner will be very unhappy in the future.
Lao Yu has also been complaining recently. Ten thousand alpacas are rushing in his heart.
The second challenge is that, taking the Chinese Internet as an example, when the sales of electric vehicles increase, the price of software cooperation will be very important. The Chinese Internet is basically free or low-cost. If a software costs tens of thousands of yuan, this logic is not right. For a car company, if it is unreasonable, it must find a reasonable price. Unreasonable prices have no life cycle.
The third and most important thing is user experience. In the wave of China's Internet opening up, most of them are opening up the bottom layer, but you generally don't have access to the underlying technical personnel. So I may be more optimistic in the short term, because these car companies do not have this ability, but I am not optimistic in the long term.
Author: What does Xiaopeng think about the change from oil to electric , or the development of extended-range electric vehicles?
Liu Long: I think it won’t take long, maybe 2-3 years, for extended-range electric vehicles to be slowly abandoned by the market, just like touch screen interactions.
The reason why Xiaopeng did not make a fuel vehicle at the beginning was based on an intuition: the simplicity and purity of electric vehicles are precisely the genes that make it easier for popular products to develop. This intuition has gradually been recognized by the electric vehicle circle.
Making pure electric vehicles is the same as making in-vehicle systems. Pure electric vehicles are designed with the battery as the center. The layout of the entire battery and the structure of the vehicle must consider the needs of the battery, including safety, energy density and BMS system, etc. Only in this way can electric battery life and software power consumption be improved.
When developing a pure electric vehicle, when developing the chassis, the main consideration is how to make the battery layout more reasonable and efficient, without having to consider other factors, which is something that extended-range electric vehicles do not have.
The traditional car architecture has no problem for fuel vehicles, because there are very few control algorithms and and relatively few control functions. But for smart cars, it is very difficult.
There is an essential difference between the new chassis platform and the method of modifying a fuel vehicle chassis and then adding a battery to turn it into an electric vehicle. This is because the modified design has major flaws. It requires consideration of how to place the fuel tank, engine, and transmission. box.
These two are system-level thinking.
The range extender has certain flaws. I think both pure electric and fuel cell are more promising than the range extender. As pure electric vehicle battery technology continues to develop, the energy density of batteries can now be made higher and higher. In the next 2-3 years, battery life anxiety should be resolved.
4. Globalization
Author: What do you think of Xpeng’s globalization strategy, especially the globalization of G9?
Liu Long: I think Xiaopeng’s globalization strategy has several foundations.
Xpeng believes that 2023 will be the first year of commercial payment for intelligent driving systems. We can classify the period from 2020 to 2022 as the intermediate intelligence stage, and smart cars are still gaining momentum.
Whether it is Xpeng Motors or Tesla, everyone is in the first stage now. At this stage, the main task is to collect user data, conduct product iterations, maintain operations, and build infrastructure.
From 2023 to 2025, smart cars will usher in a period of explosion and subversion of traditional functional cars, rapidly developing from mid-level intelligence to high-level or full-level intelligence.
At this stage, Xiaopeng must seize the wave of global intelligence. The United States is Tesla’s home base, and China is Xiaopeng’s home base, so Xiaopeng plans to compete with Tesla in Europe, even if it cannot defeat Tesla. La, you can also eat a lot of meat as the second child.
(Xpeng Motors G9 picture)
The second factor is Xpeng's new electronic and electrical architecture, a new generation of central computing EEA architecture, which is more like a computer and can perform OTA quickly. The previous architecture needed to take into account some upgrades to ECU. This time, it is completely updated directly by the brain. OTA is becoming less and less difficult and more efficient.
Facing global users, Xpeng can quickly obtain users' operational data and update it efficiently. In this case, Xpeng's cars can quickly adapt to localization no matter where they are in the world.
Especially the use of silicon carbide devices has greatly improved charging efficiency and battery life.
The third factor is that in the next three years, there will be an intersection between the decline in hardware costs and the increase in software development costs. That will be the time when the automotive landscape will truly be disrupted and changed. Software will begin to become important and become a major part of the automotive landscape. The company's main source of revenue.
This is also related to everyone's hardware arms race. More and more core hardware companies are beginning to enter this industry.
At this stage, Xpeng’s autonomous driving team has nearly 1,500 people, but the personnel cost is also very high. It is also very difficult to control the cost of the autonomous driving team, so more people need to use Xpeng Motors. Smart software.
Just as Yu Chengdong said, Huawei's current intelligent driving team's annual research and development costs are about 1 billion US dollars, and large-scale commercialization is necessary to maintain the team's expenses. Only global sales can support such R&D support.
This is also in line with Xiaopeng's philosophy. In the long run, a technology-based automobile company should have high gross profits. In the short and medium term, it is wrong to have too high gross profits in the short term.
Author: So what does Xiaopeng think of Robotaxi?
Liu Long: Whether it is autonomous driving or urban navigation systems, just like barrel effect , there must be no shortcomings. If there is a shortcoming, the entire system may not be able to run.
In the future, there will be a small number of leading players in this field, and of course more funds and resources will need to be invested. Another group of players will provide supporting equipment to OEMs to achieve commercialization.
This trend is obvious now. But in 10 years, autonomous driving will no longer be a high-tech technology, and everyone will start to compete more with each other in terms of operational capabilities. Just like 10 years ago, the big data team and the cloud computing and team thought they were awesome, but now they have become the basic technology department of a company.
A salesman suddenly stuck his head in. The two brothers have been chatting for 2 hours. We are going to have dinner. Do you want to continue chatting? If you want to continue chatting, we will match you with a salesperson.
Liu Long said with a smile, we have almost finished chatting, we still have dinner, let’s go. The two of them stood up, patted their butts, and walked straight to the box on the fifth floor.
