In recent years, when interviewing executives from car companies, two old topics have come up again and again: chip shortage and battery shortage. At the same time, some industry jokes have emerged, such as the one from He XPeng, “Whoever gives me a chip, I will invite him for a drink.”
Many people wonder why car companies are so short of chips now. Li Xiaohe, the global vice president and general manager of the New Energy and Drive Systems Product Line at NXP Semiconductors, one of the world’s largest automotive chip manufacturers, explained at the 2022 media communication meeting: “It’s not that the total amount of chips is less, but the number of chips in each car is increasing by a geometric progression.”
According to Li Xiaohe, when our car goes from L2 to L4, our electronic demand will double; when we upgrade from a gasoline vehicle to an electric vehicle, the electronic demand will also double; when we add all the functions and become an informationized new energy vehicle, the electronic demand may increase by 3 to 4 times or even higher.
From Li Xiaohe’s words, it is not difficult to see that with the emergence and development of intelligent electric vehicles, the production capacity of automotive chips has encountered great challenges. In fact, according to Li Xiaohe’s statement, it is not just a chip supply problem, but the entire automotive industry chain is accelerating its restructuring.
Here are some Q&As from this communication meeting for reference (partially deleted):
Interviewees: Li Xiaohe (right below), global vice president and general manager of the New Energy and Drive Systems Product Line at NXP Semiconductors; Zhou Xiang (left below), general manager of the Automotive Electronics Market in Greater China at NXP Semiconductors.
Q: The combination of chip companies and the automotive industry chain may be different from before. It may not be combined with Tier 1 and automakers. So, from the perspective of a chip company, how do you view this trend?
Li Xiaohe: Nowadays, the automotive industry chain is in an environment of changing integration, and we almost have very good interactions with all customers in the industry. Talking about the development trend in the new energy industry, we see that the industrial chain of the automotive industry is becoming shorter and shorter from the perspective of the whole car factory. In the past, the linear model of Tier 2, Tier 1, and the car factory was adopted. Now, many new energy vehicle companies have adopted the form of direct self-research and vertical integration. Even some new energy vehicle companies include battery and even battery cell manufacturing in their overall supply system, and battery manufacturers often have their own battery management system. The industrial chain of the automobile is becoming shorter, and at the same time, the research and development cycle has been reduced from the original 4 years to 2 years. Therefore, the response speed of the entire chain has become faster, and more and more car factories choose to cooperate directly with battery or chip manufacturers in a collaborative manner. This cooperative approach not only simplifies the design but also speeds up the design process. In our view, this is a very effective R&D approach. Of course, there are still many car factories that, based on their own market positioning, R&D processes, and strength of independent R&D and investment, choose to cooperate through battery factories and Tier 1 in the form of adding. Therefore, in the end, it should be a state of diverse chains and coexistence for a long time.
Q: How does NXP adapt to the faster iteration cycle from 4 years to 2 years?
Li Xiaohu: Previously, it took approximately 4 years to develop a platform. If we use a “V-shaped” development cycle, the 2-year chip development time can be adapted to the platform development status. However, the problem is that if the traditional approach of waiting for customers to provide requirements is used and the platform is still developed in 2 years, it is certain that there will not be sufficient time. From a development process perspective, some time is necessary to ensure the quality of the car, such as the time for quality testing and reliability testing. Therefore, the most important point is to enhance interaction with the ultimate customers and not wait for them to directly provide requirements. This is why we need to communicate with the whole car factory and the battery factory because we need to understand where the requirements for electronics and electricity come from and ultimately solve the problems for customers. In summary, we need to enhance interaction with the car factory and battery manufacturers. The second point is to have a long-term road map. From Lars’ speech, we can see that we have a 10-year and 5-year plan internally, which is not only for product planning but also for system innovation planning. For example, we need to anticipate within 10 years what innovative changes will occur in this system. Under these innovations, we can step by step deduce system innovation changes, product innovation changes, and IP innovation, thus connecting the entire chain together. In this way, we do not need to wait for customers to provide requirements, but we can predict what kind of requirements customers will make in the coming years. By achieving these points mentioned above, we can adapt to the market changes with the semiconductor R&D cycle.
Zhou Xiang: Let me supplement a little. The MCUs and Arm cores we talked about actually have many reuses at the bottom, and the product is based on a basic core and iterated towards the rear. So this is also the angle we have to ensure acceleration and make our product iteration cycle faster. Therefore, we can respond quickly to some demands raised by car factories, even helping to define or customize some products.
Q: How does NXP help automakers reduce costs?
Li Xiaohu: I think in electric vehicles, the idea of cost reduction and the source are not focused on the chips themselves but on the system cost reduction. Regarding the potential for maximum cost reduction in electric vehicles, it is on the battery that accounts for 30% to 40% of the total car price. If we can improve the battery utilization rate and improve the efficiency of the electric drive, then the cost reduction potential is more valuable than a few percentage points of battery cost reduction. However, in the industry, everyone is digging into the cost of this part, whether it is on the hardware or software. If everyone wants the current battery to travel a little longer or be smaller for the same mileage, how can they do it? In fact, we need to consider the transition from IGBT to silicon carbide, which some say can increase efficiency by 8%, while others say it increases more. But overall, it is to use the process of transitioning from IGBT to silicon carbide to help improve our system efficiency and reduce the cost of batteries by increasing efficiency. In addition, the increase from 400V to 800V is the same. Although the cost of the device may be higher in the high-voltage part, typically speaking, the space saved in the wire and motor, as well as reducing energy consumption and improving the overall energy conversion rate. In fact, what we are doing is to help customers escort from a semiconductor technology perspective to realize the transition from IGBT to silicon carbide and the transition from 400V to 800V. During the transition, our gate driver itself needs to have better functional safety, more diagnostic functions, faster response speed, more protection, and higher voltage resistance level. Actually, through technical means, we can help customers achieve a smooth transition to silicon carbide and 800V, and customers can save energy costs far more than the cost of buying more advanced semiconductors in such a transitional process. This is the logic behind it. Basically, in the 800V system, the transition from IGBT to silicon carbide is cost-effective. In addition, whether to use a permanent magnet motor, an asynchronous motor, or a mixed-use method depends on solving the torque problem at low speeds and the energy efficiency problem at high speeds. In the development process of the next generation, the type of motor itself is not closely related to the gate driver, and NXP can support it. What we look at mainly in the future are the specific types of power modules, such as IGBT or silicon carbide, even next-generation technology, and whether it is 400V, 800V or even higher voltage, which will have an impact on our chip definition and design.# Q: BMS is now very important for electric vehicles, what do you think is the main pain point for customers? How does NXP Semiconductors solve it? How do you handle the relationship with battery manufacturers and automakers? Additionally, about wireless BMS, can you let us understand how NXP Semiconductors is doing in commercial realization? What are the advantages?
Li Xiaofei:
BMS has developed from more than a decade ago to now, and the overall technical route has become relatively mature. Industry has solved many of the most painful problems, and leading customers have relatively mature solutions whether from accuracy or safety perspective. Some customers have shared some pain points, which are the basis for the industry to further develop and optimize. In the future, everyone needs to continue to improve the efficiency and further enhance the existing mileage and charging speed. At the same time, from the platform of 100,000 to one million levels, the mass production quality and level need to be further improved. In the future, the industry will consider more how the carbon emissions and recycling capabilities of the overall battery are under the premise of a circular economy. In terms of battery management, revolutionary changes will take some time to appear, which may be related to the evolution of chemical substances. Under the current chemical system, more optimization processes are being done.
Another thing to discuss is that everyone always thinks that battery management is a hardware thing, but many bottlenecks are now in software capabilities. The accuracy of hardware performance can be done well in the hardware aspect, but the accuracy of SoC relies largely on the software modeling and model maintenance capabilities. Therefore, from the perspective of system thinking, the final problem to be solved is how to ensure the quality of data, how to use this data, which data needs to be locally processed, which data needs to be processed in the cloud, whether to use a fully machine learning method or a machine learning method based on physics model after sending it to the cloud, and how to use the learned back. Additionally, it is also considered how to increase the original variables, detect the original variables, and improve safety. Everyone needs to solve a problem on the next level.
NXP Semiconductors provides customers with a convenient set of architectures as much as possible. Users can exert their own advantages on our architecture. For example, we help customers optimize the system by coordinating all the hardware they need in the system (including microprocessors, SBC, networks, analog front-end, sensors, RTC, etc.). On this basis, we help customers to do drivers, functional safety manuals, etc., and write them in a software that supports mass production. Finally, customers can build their upper-level software on our system. In terms of cloud computing, we are preparing to help customers integrate our cloud platform, including using S32G’s cloud platform together, which is actually to build infrastructure for customers and help them develop their innovation ability in software on this infrastructure. This is what BMS needs to do next, which includes data, software, and further pushing the existing parameters to the extreme.## Evolution Path of Automotive Electrical Architecture and its Requirements on the Industry Chain
Talking about wireless BMS, the concept has been a bit abused. Everybody is asking what wireless BMS is, what problem it solves, and whether just adding a wireless interface would make it a wireless BMS. Is it to add the interface on the module or on the PACK? It is just a CAN with a wireless interface, but it is not BMS. If this communication interface is to be placed on each module when a car has 8 to over 40 modules, how can the synchronization, real-time and functional safety of the modules be ensured? Hence, there is a series of questions to be answered. Before discussing benefits, which includes saving costs, improving reusability, and enhancing ease of assembly, a few questions have to be answered. What is the pain point that benefits need to solve? Are there other alternative solutions for each link? There are many different technological routes that can be chosen. What can be seen on the market today is just one of the many options. Finally, engineering still has to solve engineering problems. Furthermore, optical communication can also be another way, but it is not high-frequency wireless. We believe that the future is a market with many choices, and the final decision on whether to use wired or optical communication or high frequency is likely to depend on the principles and methods of battery design. For instance, C2P and C2C, which are popular in China, have little flexibility with regard to wireless in the design system because it is not needed as the batteries are already designed well. However, when flexibility is needed in a large module, high-frequency wireless may have some application scenarios. Therefore, it is still too early to discuss which method to choose. From the technological route taken by NXP, we are trying to help different types of customers support different technologies. There are already several alternatives to wireless being considered, and some are conducting early-stage research, while others are working with customers to build a prototype. The final decision on which alternative to accept depends on the market and time.
Q: What are the evolution paths of automotive electrification architecture, and what are the requirements on the industry chain?
A: Actually this topic is quite broad. The automotive electronic and electrical architecture is affected by how the software of the car is written. We mention software-defined cars; in the future, software will define the car’s marginal value. Therefore, the logic is not that the electrical architecture has changed due to electrification, but because of the intelligent software of an autonomous driving vehicle. The software architecture has changed, leading to an increase in computing power or the emergence of centralized supercomputers, domain control, or complementary architectures to supercomputers. These will actually affect some of the technologies involved in our process of electrification implementation. For example, we initially mentioned a software-free battery. Previously, there was a microprocessor in the software, and many algorithms were implemented in the battery. The battery and domain controller were connected through a CAN interface. In the future, if there is a supercomputer and an integration operator, such as the edge computer, there may be fewer software programs in the battery for software maintenance convenience and kernel computing power sharing. We have also developed a CAN-to-BMS daisy chain gateway in different product lines to reduce microprocessor computing consumption. Moreover, we use an integrated hub and a more integrated analog front-end to implement centralized computing power under new software-defined cars and to improve the battery.Q: 800V high voltage platform may be a trend currently. What are the differences in semiconductor demand between this platform and traditional platforms? Could you talk about the specific requirements for the inverter?
李晓鹤: First of all, a digression. The demand for semiconductors will increase significantly with the 800V platform. The demand for battery management systems is directly related to the voltage, so the demand for semiconductors will be greater. Regarding the inverter, overall, the use of 800V and silicon carbide is strongly correlated. The choice between IGBT and silicon carbide in 400V may vary, but silicon carbide is more economical in the 800V system. Basically, we combine 800V with silicon carbide. From the perspective of 800V drive, there are corresponding higher requirements for the overall electrical system’s voltage resistance, response speed, and diagnosis. NXP’s products can also support the 800V system. We don’t think there will be a separate 400V system in the future because customers may have future 800V requirements when creating electrification platforms.
Q: Nowadays, different OEM manufacturers have different routes for the development of new electronic electrical architectures. When will there be several unified large development routes? Which one does NXP favor?
李晓鹤: According to the different starting points of each automaker and their different logic on software, there are now two dimensions. One is the evolution speed of physical configuration, and the other is the evolution speed of logic. The physical refers to where the boxes are placed, which boxes are combined together, and the logic refers to how the virtual machine runs and where the virtual box is placed. Now, it seems that most automakers will eventually reach the state of central supercomputing, but there will be different choices in which route to take first, either the physical route and then the logical route or vice versa. There may be several time points and ways to merge these two paths. We think there will definitely be a variety of choices, and not all automakers will take the same path. However, we believe that moving the software up and defining the software-defined car is the big direction that remains unchanged. NXP has no preference among them. There are actually new solutions for various routes in both microprocessors and network communications and the corresponding configurations. Therefore, we hope to support our customers’ needs, and NXP has the ability to support various customer needs.# Zhou Xiang:
We can see the differentiation between regions and domains currently. As Li mentioned earlier, in terms of integration of functions such as ADAS or electronic cabins, they still exist in domain control. It is expected to continue in the future. Region controllers will have a relatively fast evolution at the body end and in other communication integrations.
Q:
Regarding software-defined cars, is there also a requirement for software-defined electrification?
Li Xiahe:
We think that the so-called software-defined means that software always needs to run on hardware. In electrification, the most important thing is the triad of electric power. We have already talked about batteries. The possible result of a software-defined car is that the battery becomes pure hardware, and all or most of the software that needs to be updated will be done on the domain controller and supercomputer. Regarding inverters, because the efficiency of the drive module is increased, it will enhance the implementability and protection of the electric drive system, and many protective functions will be made into hardware protective functions. It is possible that with the software-defined car, we will see partial hardware intelligence in the electric power drive section of the triad, and put as much non-real-time software computing power as possible on the central computer. We may not be able to directly talk about software-defined electric drive, but say that there will be changes in the hardware of the electric drive system brought about by the software-defined car.
Q:
NXP has released an electrification cooperation with NIO and XPeng. Under this cooperation model, how will our products change in the future? If NXP cooperates with a specific automaker, will we push some more specialized chips?
Li Xiahe:
First of all, I am very optimistic about the cooperation with NIO and XPeng. Both car companies prioritize innovation and we have great respect for them. At the same time, as we mentioned in today’s meeting: among the top 20 automakers in the world, 10 have adopted NXP’s high-voltage battery management system solutions, and 9 have adopted our chip-designed inverter systems. NXP’s current products are still very universal. Different companies will have different upgrade needs. Some companies will use NXP’s new products faster, while others are relatively conservative. However, overall, NXP is a universal product. However, universal products do not mean that there are no differences in the system. The way customers use our products will vary greatly, so many of NXP’s collaborations with customers are about how to make our products more efficient and reliable in their usage scenarios. NXP has a large team in China dedicated to this, providing application support and helping customers optimize their system performance with our relatively universal products. There is actually a lot of work to be done here. Will we have some customized chips in the future? I think this kind of cooperation is not ruled out. Such discussions may be based on several big premises: one is that there is commonality in technology. We believe that this technology has prospects in the future. The second is that there is a basis for mutual benefit in business. The third is to seek reuse of an IP. Because the development cycle of semiconductors is very long, and the development cost is also very high, generally, few customers can support the entire R&D costs. We still need to find some reuse of underlying IPs and provide some customized services in this way.
Q: You mentioned that Mobility as a Service will first appear in major cities in China. During this process, may I ask what opportunities will belong to Chinese start-ups? Also, how will NXP Semiconductors practice in the new ecological construction process?
Li Xiahe: There are several prerequisites for the emergence of Mobility as a Service. Just now, we talked about autonomous driving, and the overall vehicle charging network needs to be very good. The advantage of this is that it can effectively improve the utilization rate of the vehicle ownership and support more travel services with less ownership. However, it also has problems, especially in big cities where commuting is one-way and the return trip is empty, which requires the concept of sharing, otherwise, if every person who goes to work has a car, it will bring significant problems to the city’s transportation. The solution is that this situation has led to the emergence of more light commercial vehicles. They are not traditional private cars but are shared by several people. The usage scenarios of light commercial vehicles are different from ordinary people, so the requirements for battery life, maintenance, fast charging, and electronic product life are also different. We believe that in the future, there will be some emerging operators. Whether it is a service start-up or a car start-up, there are opportunities. Service start-ups focus on how to bring added value through operations. In the end, everyone is not calculating the cost based on the price of the car but based on cost per km. Start-up car companies may need a vehicle that is different from the current car. More importantly, we need to do our homework at NXP Semiconductors: make products and systems reliable. When the car can run for one million km, at 1200V fast charging or 3MW fast charging and in normal fast charging scenarios, we must consider how to optimize battery life. These basic technical issues need to be solved by us, and the final business model can be expected.
Q: With the development of electric and intelligent cars, the use of chips is becoming more and more. I wonder which vehicle currently uses the most chips, and how many are there? Of course, the use of chips is not unrestricted, and the domain controller will gradually replace MCU, which will slightly decrease in quantity. I wonder how many chips will be needed in the end?
Li Xiahe: There is no clear data on how many chips are used on average because this depends on the vehicle model. We began to do car networking at the end of the 1990s. The first car network had only two CAN nodes. By 2008, there were probably more than a dozen nodes on average. When the domain controller appeared, the peak of the car network could reach more than 200 nodes. Looking ahead, the domain controller will gradually transform into a supercomputer. There doesn’t seem to be any significant change in the number of nodes because a supercomputer can solve computing problems, but it cannot solve real-time and power consumption problems. Originally, everyone thought that the supercomputer could replace many nodes. Perhaps in terms of computing power, this is the case, but in reality, Ethernet is mostly added in parallel to the original CAN network instead of replacing it. In such a case, we believe that there may not be a significant change in demand with a big up and down in the future. Some people say that there are many nodes now. Will there be a sudden change in the structure at a certain point in time and the demand will drop sharply? I think it won’t be like this. The transition will be smoother. Now it is stable growth, and it will be a smooth transition in the future. We believe that in the future, for the requirements of real-time, security, and power consumption, there will still be a large number of sub-nodes, and the high computing power platform may be concentrated. However, some sub-nodes will still exist to the last mile of landing, such as wipers, water pumps, and local control of computing power. The demand for semiconductors in the future may be relatively stable. Although the growth is fast now, it will be relatively stable at some point in the future.## Translation
周翔: Because the number of screens and cameras far exceeds the original calculation, we believe that in the future fusion, the number of MCUs will decrease due to the emergence of domains and regions.
Q: Currently, power domain integration is relatively slow. Will it be integrated in the future? Why? What are the needs and purposes of the cooperation between XPeng Motors, NIO and NXP?
李晓鹤: Technically, power domain integration can be achieved. However, the problem lies in why we should do this. In the current automotive supply chain, each supply chain has relatively mature players, each with its own solutions. Battery has its own solution player, electric drive has its own, and domain controller has its own supplier. These suppliers are not completely the same, and each has its own strengths. When purchasing, the entire vehicle factory usually purchases separately. Even if some manufacturers do it by themselves, it is also different teams that do this. Sometimes, the ability to integrate things may not be a technical issue, it may be a supply chain problem. If integration cannot bring obvious cost or technological advantages to the existing supply chain, this change may be relatively slow. We do see that some people are trying to put multiple technologies together. In the future, it may also be a situation of multiple approaches. Some are indeed able to integrate well and can be done in some programs. However, the existing separated industry chain may also exist for a long time, and sometimes some car manufacturers want to make batteries, but the electric drives have to be done by others, or vice versa they want to do the electric drive themselves, but still buy batteries from outside. Diversity will always exist.
In the early stage of our cooperation with car manufacturers, the most important point is that if this matter waits for the car manufacturer to ask us for it, and we then start development, the overall development time will be very long. When we can have relatively early communication with the car manufacturer, we can develop in advance and understand the car manufacturer’s requirements. Another advantage is that the car manufacturer can closely cooperate with our new products, and both sides can quickly verify and cooperate effectively to help the car manufacturer to introduce new technology. There are some customers who may attach more importance to the introduction of new technology, and we will cooperate with them accordingly. At the same time, we believe that some car manufacturers are more accustomed and have the ability to cooperate with semiconductor manufacturers, and some car manufacturers currently prefer to cooperate through Tier1. For car manufacturers who have both the ability and the willingness to directly collaborate with semiconductor manufacturers on innovative projects, we are very welcome. Therefore, we also hope to support these two car manufacturers as much as possible, one is doing battery BMS and the other is doing electric drive.Zhou Xiang: Both of the two car companies have had very good and industry-leading interactions and collaborations with us, which are developed through deep cooperation with the support of NXP’s new products. In addition, for NXP, we listen to the future demands of Chinese car factories, whether it is for the next generation of electric drive, IGBT, or MCU, NXP will try to integrate them into the product planning of the next generation. This kind of localization is happening more and more in China.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.