Li Bin talks about R&D management: innovation means facing a lot of failures|36Kr Interview

Text丨Peng Suping

Editor丨Li Qin

As the "CTO" of NIO, Li Bin plans to hand in a high-scoring R&D answer sheet on the flagship sedan ET9.

This year, NIO launched the ET9 sedan with a pre-sale price of up to 800,000 yuan - this is an executive flagship sedan with a length of more than 5.3 meters and almost all the latest R&D achievements of NIO, with a pre-sale price of 800,000 yuan and is expected to be delivered in the first quarter of 2025.

Since 2020, NIO has spent a total of 27.4 billion yuan on research and development, and in the past two years, it has released more than 10 new cars and upgraded software and hardware systems, and ET9 and the new technologies it will be equipped with are even more important investments: self-developed intelligent driving chips, 900V high-voltage architecture, 46105 large cylindrical batteries, and intelligent chassis systems...... Li Bin said that ET9 covers "17 world premiere technologies and 52 class-leading technologies".

The 900V high-voltage architecture and intelligent chassis system can be regarded as the integration and upgrade of NIO's capabilities in related fields in the past.

In the past two years, the electric vehicle industry has been charging at a rapid pace, and many 800V new cars have been launched this year, while NIO has further raised this threshold to 925V, with a peak charging power of 600kW.

In the early days of its establishment, NIO developed its own electric drive system, and this time it further iterated technologies such as motors and silicon carbide power modules, and also extended its self-developed tentacles to upstream battery cells to create a more outstanding experience.

According to reports, Wei's self-developed 46105 large cylindrical battery, a single energy density of up to 292Wh/kg, internal resistance of 1.6 milliohms, only half of other large cylindrical batteries, its assembled battery pack will be 120KWh, can support 5C high-voltage fast charging.

Zeng Shizhe, vice president of NIO battery system, said in an interview that the upper limit of safety and function realization of large cylindrical batteries is high, and the industry's understanding and application of it is still relatively early, so NIO has room to improve to form some platform advantages.

However, as he said, the technical peak of large cylindrical batteries has not yet been climbed. An engineer from CATL told 36Kr that to achieve 5C fast charging, it is not only necessary to solve the problem of internal resistance, but also heat dissipation, and the heat dissipation efficiency of the large cylindrical serpentine tube is inferior to that of the large surface heat dissipation of prismatic batteries.

Previously, NIO had reached a cooperation with Honeycomb Energy to jointly develop large cylindrical batteries, but according to 36Kr's latest understanding, the implementation progress of the cooperation is still unclear. Tesla, the pioneer, is still making poor progress in the mass production of large cylindrical batteries, and the engineering verification of 5C fast charging is also blank.

The intelligent chassis released this time integrates steer-by-wire, rear-wheel steering and fully active suspension technologies, which makes the ET9 large, agile and extremely smooth.

NIO showed a video at the press conference: a three-layer "champagne tower" was placed above the front of the ET9, and the car started and drove over a continuous speed bump, and the champagne was not poured or spilled. NIO named this chassis "Tianxing Chassis" to promote its "smooth and comfortable driving like an airplane in the stratosphere".

It should also be pointed out that the technology of full steer-by-wire and full active suspension has not yet been mass-produced in the industry. Some practitioners believe that the mass production of fully active suspension is expected to take at least three years.

Chips are similar to batteries, which is the result of NIO's deepening layout in the upstream of the industrial chain in the past two years, and this is also the first time they have been put on the car.

The intelligent driving chip, named "Shenji NX9031", will be manufactured using a 5nm automotive process with 500 100 million transistors. However, the specific computing power has not been announced, Li Bin said that the goal is to replace 4 industry flagship intelligent driving chips (Orin X) with 1 Shenji NX9031.

It is worth mentioning that the ET9 will be equipped with two "Shenji NX9031" chips to achieve computing redundancy.

Safety redundancy is an important design concept of ET9, which has a backup design for seven key systems, including drive, braking, steering, perception, computing, communication, and power distribution, and the backup system can work immediately in the extreme case of the main system failure.

The delivery time of NIO ET9 is set for the first quarter of 2025, which means that the technology released by Li Bin from batteries to chips and even chassis systems will be mass-produced in a year. With more than 40 billion R&D investment, it seems that it is only a "kick in the door" to reap the results. Li Bin also summarized the R&D management experience to 36Kr, he said that all R&D directions are formulated around a fundamental principle: whether it can create value for your target users.

In the process of R&D, he believes that the team cannot be strict enough not to fail, "Innovation, in fact, means that you have to face a lot of failures. As long as you do it seriously, the company will give you (the opportunity), no one is a god, and many of the decisions I make here are also wrong, how can you ask everyone to do exactly the right thing?"

But at the same time, he emphasized the bottom line of R&D management: to seek truth from facts, solve problems with problems, and not cover up problems, "My requirements are very simple, I can't cover up problems, and I can't accept covering up problems." ”

Technology investment is about to start branching, and Li Bin also looked forward to the company's profitability node: when the monthly sales of 30,000 vehicles and the gross profit of 20% are realized, NIO China's main business can basically break even.

The following is the content of the communication between NIO executives such as Li Bin, CEO of NIO, Qin Lihong, President of NIO, Bai Jian, Vice President of NIO Intelligent Hardware, Zeng Shizhe, Vice President of NIO Battery System, and other NIO executives and 36Kr and other media, 36Kr has made some excerpts:

High-end brands should continue their genes

Q: The ET9 will not be delivered until 2025, and many people don't understand why the information about this car was released so early.

Li Bin: Judging from NIO's past experience, it was not released too early. We are a big hit on NIO Day every year, and you can't wait until next year, and it will be listed in the first quarter of '25, which doesn't seem to be appropriate.

The research and development progress of the model is still much faster than everyone imagined, this car has been in development for two years, it is a car with a completely reshaped soul, brain and body, but the design gene has not been subverted.

Has Porsche subverted its design DNA? No. Has Mercedes-Benz subverted its design genes? No. What kind of people will subvert their own design genes? Kia, Hyundai, old subvert, generation after generation is different. Should we make Kia and Hyundai? I'm not saying they're bad, they're very successful, just that every company has its own trade-offs.

The time is about the same, of course, it can also be sent in February and March of the next year, but is it really that anxious? As an administrative flagship, it still needs a long warm-up period, and there is still a lot of content to talk about.

Q: Everyone is surprised by the design of the ET9, which is not the exterior design of the executive flagship that everyone thinks of. How does NIO communicate with potential users to make it easy for them to accept?

Li Bin: Design is divided into styling and interior. From the perspective of styling, there are three goals: first, when we defined styling two years ago, we took into account who the user is, who is the innovation-led, more conservative mentality, and will definitely not buy our car; second, the intelligent electric vehicle technology has brought a breakthrough in the boundaries of many things, such as ET9 The ground clearance of 200 mm cannot be achieved without a fully active suspension, and with a full active suspension, the advantages of high seat order are reflected.

Subversive, it is impossible to make everyone like. For an executive flagship, the times have changed, unlike before, everyone wants to suit straight, today may be business casual, how to have a new sense of decency and calmness in today's business scene, the scale in the middle is very difficult to grasp, we think ET9 is good.

Self-developed products are efficient in staying ahead of the curve

Q: What is the computing power of the self-developed intelligent driving chip, and what are the advantages compared with NVIDIA's chips?

Bai Jian: From the perspective of computing power, our support for NIO's intelligent driving is sufficient. But computing power is enough to solve the problem of food and clothing, and we now have to consider some things to be well-off.

I came from a mobile phone, and the mobile phone is very concerned about power consumption, because battery life is very critical for mobile phones. But before the car didn't care so much about power consumption, for example, the automatic driving system of the NT1 model, I didn't feel it even if it was fully turned on 24 hours a day, because the power consumption was too low. But now it is different, the computing power of the autonomous driving system is getting higher and higher, and it will have a demand for power consumption.

So we introduced power gating in the nervous system, each block is specially designed to be turned on and off individually, and more importantly, it can be pulled up and hibernate instantly, the car is a high-safety, high-reliability system, this must be done, and any point of user experience cannot be at the expense of safety.

The other is our exploration of latency, ISP (Image Signal Processor), and reliability. Intelligent driving experience is getting better and better, latency is very critical, now a camera is 30 frames per second, and there will be more in the future, if you are still slowly and leisurely using data throughput, and then to transmit, that is actually squeezing the time of the algorithm. So if you don't invest enough in this area, no matter how strong your computing power is, it's not a utilization rate.

There's also redundant security as mentioned yesterday, which may be difficult to achieve in the industry in the next few years. I will make FIT (failures per 1 billion hours) to hundreds, but no matter how low this value is, we can't guarantee that it won't make mistakes, once it fails, can I make my backup system pull up and take over in milliseconds, this is not just a chip system, it requires a full set of operational systems and AD algorithms, and we must do it together to achieve it.

Q: What do you think of the trend of cockpit integration? Will you make cockpit chips?

Li Bin: It's not really a technical problem, it's a definition problem. The higher the integration, the more efficient it is to handle core tasks, so the cockpit chip is not such a high priority at present.

But our CCC, which is the supercomputing platform on the car, has a very high level of integration, and we really make it a central brain, and whether the chips are put together or separated, this is actually a matter of definition.

Bai Jian: The integration of the cockpit is to use a chip to run the functions of AD and the cockpit, but that is based on the fact that now AD and the cockpit are originally two part lines, and there is no integration of software. However, in the ET9 architecture, the CCC cabin is originally integrated with each other, and its computing power can also be called to each other.

Cross-domain calls are not easy to do in the industry. It's not just a matter of technology, it's also a matter of organizational integration. In many cases, the development of AD and cockpit is completely two ToB teams, but in this R&D system of NIO, under the scheduling of SKY OS, it is now integrated. As for whether to use one chip or two chips, the perception from the user's point of view is not strong.

Q: ET9 is equipped with three lidars and two self-developed chips, will the hardware of the next generation of AD be so standard?

Li Bin: Many people tell me that NIO should push a single motor and remove the lidar, but from the perspective of the NIO brand itself, everyone must still understand the value of computing power and sensors to users. In the confrontation of many things, such as the pure visual solution, as if it is better than the solution with Lidar, the pure visual solution is of course the foundation, but it does not mean that Lidar is useless, it is from the perspective of cost, and is unwilling to use Lidar, which is a problem of configuration trade-offs, not a technical problem.

Don't put these opposites, the key is what is the benefit of each sensor, and don't have negative benefits. We had ET7 at the time, and we had a certain amount of redundancy even visually with the layout. When do you say that its role will be played? It will only be played in very few cases, but the advanced nature of technology will solve the problem in a very few cases. I often take airbags as an example, you can't say that 99.999% of the time the brakes and seat belts work, and the airbags are optional.

Q: NIO has been using square cells before, why did this generation choose to use large cylinders?

Zeng Shizhe: We have been researching for a few years to define some directions for the next generation of platforms. There are some very key points, the first is safety, the second is that I hope that the battery monomer in addition to its own function of providing energy, can also bring more different functions, improve the feeling of the whole vehicle, and later also see some advantages of the large cylinder, in a more large-scale effect of the large single product, the large cylindrical battery will solve the problem that we have not been able to solve.

Why can't you see the big cylinder so strong? It's because no car factory can provide the volume of a large cylindrical production line, and then everyone's needs are different, 3C, 4C, 5C...... Therefore, we make our own large cylinders, and we have the opportunity to expand the platform of large cylinders, and we can also optimize the application of the benefits that it can bring.

Why is it difficult for large cylinders to do 5C charging? The key point of 5C charging lies in its combination, then we have basically tested the large cylinders of every battery factory on the market, and then we have also tried all the samples provided to us by battery cell factories, I think everyone's application and understanding of large cylindrical batteries is still in a relatively early stage.

Q: There is a saying in the industry that it is difficult for large cylinders to achieve 5C, what kind of technology does NIO use to achieve 5C capabilities?

Zeng Shizhe: Wei's self-developed 46105 large cylindrical battery cell has an internal resistance of 1.6 milliohms, and other large cylindrical cells released in the industry are generally close to about 3 milliohms, which is basically halved. There is still a lot of room for breakthrough in the battery itself, including the 150-degree semi-solid-state battery we made before, everyone said it was unlikely, but we forced it out.

Because a lot of people just use the supplier's stuff and get together, he doesn't pay attention to the details, we pay attention to the details, and we will find that there is still a lot of room for improvement, the choice of materials, the way of welding, the way of connecting parts...... After these improvements, we can form some platform advantages in terms of energy density.

In the end, the car is a comprehensive competition, and NIO hopes to enter the top 10 in the world

Q: As CTO, what have you done on NIO's core technologies over the years?

Li Bin: There are many layers of technology, and what I am good at is large architecture. When I was in college, I took the system analyst exam, and I believe I was the only one among the liberal arts students who had ever taken the system analyst exam, and I was also the first certified system engineer at Microsoft in China. The title of my dissertation as a systems analyst at the time was rapid prototyping, how to quickly build a prototype of software to speed up iteration - this is something in my bones, and I care more about the architectural logic of the technology, and how it should be related to each other.

For example, I will look at the value of mobile phones from this perspective, we call it panoramic interconnection, the experience of car machine and mobile phone interconnection, the logic behind it, the underlying architecture, I look at the technical architecture at a higher level, the cross-domain technical architecture, the full-stack technical architecture. But you talk about how to write specific code, it's true that I haven't written code for a long time, and I shouldn't devote my time to writing code anymore.

If you look at the past few years, I talk about a change in the organizational structure, and you can see that we have a broader vision than before. At the beginning, we divided the company into three parts: intelligence, electric, and automobile, and then I didn't think it was right, because this is equivalent to three mountains, and there are too many places to connect with each other and the inside, so a few years ago we made a relatively big organizational change and split these pieces into several parts. In this way, the communication between departments does not need to climb the big mountain, just climb the small mountain, and the advantages of collaboration and collaboration brought by flattening are much better.

Q: How to evaluate the direction of NIO's R&D, and how much redundancy is there if R&D failure is allowed?

Li Bin: First, whether it can create value for your target users is fundamental. For example, in the automotive field, we believe that the sense of security is also a very important value for the executive flagship;

Second, pay attention to systematic efficiency, which is not a local, short-term cost and efficiency, but a global, life-cycle cost and efficiency. When we invest in R&D, we must have our own logic, just doing it for the sake of doing it, in fact, there is no point;

Third, from the perspective of short-term execution, we will also evaluate the balance of all aspects of the company's operation. R&D belongs to the CTO's decision-making logic, but manufacturing belongs to the supply chain, and R&D is resolutely done, but manufacturing does not have to be done by yourself.

In the peer, each company has to make up its own lessons, and the final competition is comprehensive competition. In the past two years, we have been building charging and swapping facilities, and other car companies are also building them now, so we have been doing full-stack self-research, and other car companies may also be doing it.

Different companies still have different genes, and we have been resolute in doing this since day one, and of course it will also involve the internal hidden dangers that you mentioned. Of course, NIO has a fault tolerance rate, I said within the company, as long as you do it seriously, the company will give you (opportunity), no one is a god, and many of the decisions I make here are also wrong, how can you ask everyone to do exactly the right thing?

Innovation, in fact, means that you have to face a lot of failures, it is impossible to say who can do what at once, then this matter is too simple. NIO has had many ups and downs in the past few years, but the key is to learn something, the organization has grown, you have to seek truth from facts, and solve problems when there are problems. But you can't cover up the problem, my request in the company is very simple, you can't cover up the problem, and I can't accept the problem of covering up.

Q: In the past, Li wanted to say that there will be 4 car companies left in the future: Tesla, BYD, Huawei, and Ideal, but he did not mention NIO.

Li Bin: NIO actually rarely says who is not on this list. The automotive industry has never been a winner-takes-all, let's look at the two companies that are doing better today, BYD and Chery, which is ignored by everyone. In 19 years, these two companies have encountered a lot of pressure, BYD once only had four or five hundred thousand units, a few years ago, everyone may not have imagined that it has the opportunity to 3 million units a year today, and it will be a matter of two or three years. Chery has accumulated a lot of experience in the past two years, but he has been engaged in globalization for more than ten years.

The automotive industry is a very long competition, we said that we will enter the finals in 25 years, but I have never said when the finals will end, because there is no end to the finals, let alone the automotive industry, there is no end to doing business.

I think that around 2035, Chinese brands will account for 5 of the world's top 10 auto groups. But which 5 companies, each person has a different answer, and we strive to be one of them.