editor’s note:tencent auto's editorial department calls the past decade of electrification the "stormy era" of china's auto industry. now, standing at the historical node of 2024, which is known as the "first year of intelligent driving", we can't help but ask: what technical route will the major players in the industry adhere to? how will they build their own competitive barriers? tencent auto has specially launched a series of intelligent driving planning. through interviews, actual tests, cross-comparisons, reviews and other methods, it strives to stand at the origin of history and further understand the huge changes that may occur in the auto industry in the next ten years, so as to provide readers and the industry with a more comprehensive content guide, which can leave a valuable historical footnote for the industry.

tencent news "high beam"

author: aodun

editor｜liu peng

over the past decade, a leading company that manufactures smart driving "boxes" has been hidden in huizhou city, guangzhou.

desay sv has undergone many changes in the past four decades. its predecessor was sino-european electronics, which was established in 1986. at that time, it was a joint venture that mainly engaged in car audio systems. in 2001, after receiving investment from siemens, it was renamed siemens vdo, and its core business was still concentrated in traditional instrument panels, audio navigation systems, etc. in 2010, desay group acquired 70% of the foreign party's shares after more than two years of negotiations. desay sv was officially born, and its business direction has gradually shifted to key components of smart cars - domain controllers, which are called "boxes" in the industry.

before 2018, desay sv was positioned as an automotive electronics supplier, and intelligent driving domain control was not an area that desay sv excelled in. the domain controller is the core of autonomous driving technology, and behind it is a high-performance computer that can control and manage the entire vehicle system.

however, in the eyes of industry insiders, desay sv has always been an "invisible" rich man. since it bravely entered the intelligent driving track in 2014, it has won the top new forces of xiaopeng and ideal, as well as great wall, gac,chang'an、nissan、toyotacompared with many traditional car companies, the installed capacity of intelligent driving domain control has ranked first in the industry.

but now, as more and more new car companies expand the scope of their own research and development and gradually reduce their dependence on external suppliers; at the same time, a number of emerging intelligent driving supply chain companies have begun to get involved in the field of intelligent driving domain control. desay sv still faces many challenges, and its market share will also face the risk of being compressed.

regarding the current market competition, li lele, executive vice president of desay sv and general manager of the intelligent driving business unit, told tencent news "high beam" that the next-generation products of the new forces are already in the cooperation process, but the cooperation model will change. "if customers want to do more, we will do less, and we will continue to cooperate with these manufacturers."

li lele said that he still remembers a company strategy meeting in september 2014. on that day, desay sv officially began to discuss whether to enter the intelligent driving track. at that time, he did not know the term acc (adaptive cruise control system) in his colleague’s ppt, and “i even searched it on baidu on the spot.”

"at that time, i had no idea about intelligent driving. i only knew that it could assist in some safety warnings. however, the interesting point is that we are very interested in new technologies and new development directions, and we have certain experience in making products, so we have to think about how to explore and how to slowly build it into a product line that can make certain business contributions to the company." li lele said that this is a very exciting thing.

the following is the transcript of the conversation between tencent news "high beam" and li lele, edited:

the seven-year rule for new business

"high beam": what plans does desay sv have in the intelligent driving race?

li lele:we currently have three major business lines. one is the sensor business line, which is responsible for cameras, millimeter-wave radars, including products such as t-box; one is the controller business line, which includes not only hardware but also underlying software, middleware, and upper-level application algorithms; the third is the smart transportation business line, which is mainly incubated in guangzhou and is responsible for the cloud and road sides of the vehicle. we call it "building smart roads."

the current business structure is relatively large, sensors and controllers. in 2023, sensor revenue was 1.13 billion, controller revenue was 3.35 billion, and overall, the intelligent driving business revenue reached 4.48 billion. in the first half of this year, our overall intelligent driving performance development remained at a high level, with significant growth. compared with last year, it increased by 45% year-on-year. the cumulative total sales in the first half of 2024 reached 2.67 billion.

high beam: when launching the smart driving business, what important discussions and decision-making processes did you go through internally? you didn't do smart driving business at first, but now you have a relatively large share.

li lele:from 2010 to 2014, we continued to strengthen our presence in the cockpit sector. in 2014, our market position and technology development in the cockpit sector were already very stable, and we were ranked first in the domestic industry. we found that if we wanted to continue to grow, it would be difficult to do so in a single product area, because our market share was already high enough, which limited our growth space.

in the following period, we began to focus on horizontal development. we saw that adas (advanced driver assistance system) was quite popular in the smart car track. foreign countries had some experience in this field, but china was just starting out at that time. when we formulated the strategy and decided to invest heavily in adas in 2014, many issues were still unclear.

what exactly is adas? what products does it include? what is the direction of future efforts? today we see autonomous driving, which is already very clear. but at that time, china had just started using driving assistance.

based on the market conditions at the time, we decided to invest in preliminary research in 2014. the investment was not very large, and there were only a dozen people in charge of this matter in the early days.

at that time, we began to gradually understand the industry, but we still faced many problems, including which competitors, which oems had demand, product demand planning, the development of technology itself, and what was the level of industry development? in 2016, we expanded and entered new product lines within the scope of existing product technology capabilities, and successively developed cameras, 360-degree surround view systems, algorithms and other products, starting with simple adas products. in 2016, we had reached 20 million in sales, and the technical center officially separated and formed a new intelligent driving assistance bu. from september 2016, we gradually began to really get involved in this field, and the number of bus also increased to nearly 40 people.

in 2017 and 2018, we made a lot of hiring and personnel investment based on business considerations. by 2019, before the pandemic, we had nearly 300 people. today, the scale of the intelligent driving business line continues to expand, and the number of r&d personnel has reached about 1,100.

high beam: in 2014, intelligent driving was a new direction for you to explore. what was your mentality when you received this task? were you worried about not being able to do it well or messing up?

li lele:i never worried about messing up. i joined desay sv in 2011, and my job at the time was project management of the early r&d projects of the technical center, specializing in innovative projects. in actual work, our team has always maintained a high curiosity about new things, and we are keen to develop products with an exploratory and curious mindset.

i remember that the company held a very important strategic meeting in september 2014. the ppt shared by the marketing department talked about some functions of adas, and mentioned the term acc (adaptive cruise control system). i didn’t know what it meant at the time and even searched it on baidu on the spot.

at that time, i really had no idea about intelligent driving. i only knew that it could assist in some safety warnings. however, the interesting point is that we are very interested in new technologies and new development directions. we have certain experience in making products, so we have to think about how to explore and how to slowly build it into a product line that can contribute to the company's business. this is a very exciting thing.

"high beam": what goals did you set for yourself when you first entered the intelligent driving circle?

li lele:this goal has not been announced to the public. many years ago, we proposed a goal internally. in china, we hope to achieve the largest scale of intelligent driving product line; in the world, we hope to be in the first echelon and become one of the best partners in the world.

high beam: from 2014 to now, is there any project that is the most important and milestone project in this decade?

li lele:2019 and 2020 are two very important nodes. we have mass-produced two major products. in april 2019, we mass-produced fully automatic parking that integrates surround view cameras and ultrasonic radars. desay sv is the first company in the world to develop this product, which has strongly promoted the development of automatic parking technology. the industry used to use surround view, camera, and radar alarm solutions, which did not involve vehicle control. for the first time, we have achieved vehicle control and integrated parking, which has enabled people to interact and control vehicles for the first time. this product is a great improvement for our own technology and has also promoted our business leap.

in 2020, we officially mass-produced the world's first autonomous driving domain controller product based on nvidia's xavier chip, the ipu03, which is the world's first high-computing platform for autonomous driving. it took us about two years from the decision to start working with nvidia in early 2018 to mass production.

on april 27, xiaopeng p7 was officially launched. this product helped xiaopeng p7 achieve high-speed ngp. at that time, especially in the domestic market, we should be the first and only one that could compete withteslait is a product that competes with fsd, and it can help you achieve high-performance product functions such as high-speed lane change and automatic lane change. this is a milestone and very important product for the entire smart driving industry in 2020.

based on the industry and our own business development at the time, we began to increase our investment in 2019. many companies laid off employees during the epidemic, but we continued to increase our investment. from 2021 to 2022, we recruited about 300 people each year. according to statistics in 2022, 60% of our engineers have been working for less than one year.

according to our estimates, the rapid growth in performance over the past two to three years is mainly due to the large-scale application of these high-computing platforms. of course, the market size of the sensor business is also constantly expanding, especially in the field of cameras, from hundreds of thousands of units in 2019 to nearly 10 million units last year. the radar field is also expanding rapidly at the same time, gradually growing to the million level in recent years.

high beam: it was your first time to develop intelligent driving domain control products. why did nvidia and xiaopeng choose you? at that time, you were actually newcomers. how did you build trust?

li lele:in 2014, desay sv began looking for potential partners in the field of intelligent driving chips. it first discussed cooperation with mobileye, but found that they were technically exclusive. desay sv had its own technology path planning and hoped to cooperate in a more open ecosystem, so it did not proceed. in 2017, desay sv successively investigated intel, qualcomm, amd and nvidia.

finally, we decided to cooperate with nvidia on the intelligent driving solution. during the communication process, we found a very good advantage of nvidia - high execution. once everyone decides on a thing, they will definitely try their best to make it happen. this is in line with our philosophy and can ensure that the research and development of new products can be smoothly promoted and implemented.

"high beam": what do you think is the core factor that enables desay sv to gradually expand and strengthen its uncertain and unclear business segments?

li lele:first of all, it is very important to have insight into the market. we have organizations related to strategic insight to seek important directions for products and markets, such as insights into the next five-year plan and market changes. we revise and update it every year, which has become an important mechanism for desay sv. in addition, the pace of different business segments is also different, and new products and new technologies are bound to have risks.

another point is that we have a firm sense of conviction. when we formulate strategies, if we identify a product as an important direction, we will definitely go ahead with it. our strategic determination and perseverance are very strong, and we will not doubt or lose confidence due to market fluctuations. i think this is a corporate culture, and the bosses are also very patient with this.

we believe that in the automotive industry, it takes about seven years for any new business line or product line to break even from incubation to the end, based on past experience. now that the domestic automotive industry is developing at a faster pace, we are also looking at whether some new businesses can shorten this cycle. therefore, we have our own expectations for the development of new businesses. we do not pursue too fast, have higher expectations, or give up if we cannot achieve the goal. desay sv's planning is very rhythmically controlled.

in addition, our entire team must reach a consensus on the goal. for example, you want to do this thing, the boss has made a decision, but do your colleagues and participants know what the significance of doing this is? the content that each person is responsible for in the specific work may be relatively specific. everyone needs to know the significance of his work and what our common goals are in the future, so that everyone can truly devote themselves to it.

end-to-end is a direction, but it still needs to evolve

high beam: looking at the entire smart driving industry chain, do you have any business that is absolutely "forbidden"? will you expand your business scope in the future? for example, huawei will not do chemical-related business.

li lele:at the beginning, we did not have absolute limits on which business boundaries could not be broken. for businesses that we had not done before, we could only say that we were still exploring. for some businesses that you did not have the ability to do at the time, we realized them through some cooperation. in this process, we would also make corresponding plans.

our algorithm business is like this. we started investing very early, but the investment was relatively limited. when your business develops very fast, especially when the algorithm technology itself develops and changes too fast, the algorithm personnel you invest in today may have an "outdated" knowledge structure in two or three years. you need some new personnel to work on new technologies, even though some high-end talents and engineers have very strong self-learning abilities. in an era like today where algorithms are changing very fast and ai intelligent technology is changing very fast, we need to carefully control the investment process.

in 2017, we recruited a top-notch algorithm team in singapore, who came from the driverless car team of the agency for science, technology and research (a*star), but we will remain cautious about related investments.

until last year, our entire technical direction became increasingly clear. for example, originally it was bev+transformer, and now it is end-to-end. we call this development process "a big change in algorithm every year, a revolution every two years". now looking back, the direction towards the final outcome is becoming increasingly clear and definite.

therefore, we have been increasing our investment in algorithms since last year. currently, we have a pure algorithm team of more than 100 people in shanghai, and this organization is still expanding. therefore, we will have a certain rhythm in investment and make adjustments and decisions based on actual business development.

however, our intelligent driving business does not have the so-called boundaries of "what to do and what not to do". it is different in different periods because we need to balance our investment, sales profits, etc.

"high beam": now that you are increasing investment in the algorithm team, what are your plans for the future?

li lele:we are currently researching new end-to-end large model algorithms, as well as developing traditional bev+transformer+occupancy networks products. because end-to-end refers more to higher-level intelligent driving products, and there are also relatively low-end products, such as high-speed noa products, city memory driving and other products for the fuel vehicle market. intelligent driving products are constantly penetrating downwards. these products and these markets still need corresponding solutions to solve them.

high beam: with the algorithm and data services, you have another mode of cooperation with car companies? you can provide intelligent driving solutions.

li lele:desay sv's intelligent driving business aims to build full-stack capabilities. in the past, it may have been relatively weak in algorithms. today, we need to make up for this shortcoming and possess full-stack capabilities so that we can provide flexible and diverse solutions based on different customer needs.

for example, some customers do not want to develop their own smart driving at all, and we can provide full-stack delivery; for example, if they have a designated algorithm provider, we provide hardware, basic software, middleware and sensors, etc.; there is another type, the oem (original equipment manufacturer) wants to do some self-research, but has only done part of it, such as regulatory control, but perception may require a relatively large investment and have certain bottlenecks, so we can cooperate on the perception part.

high beam: how many modes do you have for cooperation with oems? actually, i want to ask how many payment methods there are, and which mode do you prefer?

li lele:our cooperation model is varied. for example,changan qiyuantheir sda platform uses our solution. it is not a single car, but a series of models. the main business model is controller cooperation. desay sv will provide software development and services, which involves product sales + nre model (one-time engineering fee).

the model we prefer is the nre+product sales model, which includes providing domain control products in addition to software services.

high beam: compared with 2016 and 2019, will there be any adjustments to the organizational structure of desay sv's intelligent driving business this year? recently, many oems have made corresponding adjustments to their organizational structures in response to the end-to-end model.

li lele:yes, we considered the need for organizational changes for the new end-to-end algorithm from the very beginning. looking at the entire industry, end-to-end investment only started on a large scale last year, so we took this factor into consideration from the very beginning.

in addition, in addition to our key domestic business, we have also increased our international investment since the beginning of last year and attached importance to the expansion of overseas customer business. therefore, the process system requirements of domestic customers and overseas customers are completely different. internally, although there is no internal split within the bu, we will also form a dedicated team for overseas within the bu, for example. the software is basically shared, the algorithm is shared, and we will set up a software center-level organization, rather than saying that all the software is within the bu. we have formed a plan for a software center shared by the bu. this organization is also making corresponding adjustments.

"high beam": we recently talked with professor zhu xican of tongji university, and he said, "i believe that car companies with algorithms will sooner or later make chips. if they don't have their own chips in the end, they won't run well on other people's chips." will you also make chips in the future?

li lele:this is a very good question. how to make this chip? how to build the architecture? how much computing power is used? in fact, the chip definition requirements come from the algorithm requirements, but the algorithm development and changes are still relatively fast. although we see that the end-to-end big model is a general direction and everyone is relatively certain, in fact, two or three years ago, everyone may think that bev+transformer is a relatively certain direction. so i think that the end-to-end big model is a certain direction today, but it still needs to evolve.

how to make this model? for example, the criterion model may need to consider how to join, how to improve safety, how to raise the lower limit, etc. in fact, there are still many cutting-edge technologies that need to be studied, especially based on driving safety and bottom-line safety considerations.

if the algorithm is relatively fixed, i think the chips are relatively uniform. because intelligent driving is completely different from cockpit business or other businesses, from the final experience, the functions and performance are basically in the direction of homogenization, and the final result must ensure high safety and reliability.

therefore, until the algorithm is finalized, the chip development is still in progress, and the investment will be very large. so i think that the chip must be tape-out after the algorithm is relatively mature, and its economic efficiency will be better, and it may be easier to win in the end.

high beam: when will the end of this algorithm come? what is your judgment?

li lele:this question can be solved by observing the recent situation. tesla proposed end-to-end, and implemented the fsd 12.5 version on hardware 3.0. in fact, the computing power of this single chip is 72tops. the human-like driving experience of fsd 12.5 is indeed very good, but it still has many problems. it cannot achieve unmanned driving. it can be used as domestic l2++ or urban noa, and it is still an assistant.

tesla has actually also launched hardware 4.0, but we think it will be abandoned. it will only serve as a transition and will not be launched on a large scale like hardware 3.0. tesla is also accelerating the hardware development of hardware 5.0.

why is this so? it is precisely because of the end-to-end large model algorithm in 3.0 that it sees the possibility of achieving l4 or even higher autonomous driving technology. therefore, its 5.0 plan is probably to plan for 5000tops or higher computing power, and directly develop to a higher level.

if tesla can achieve a higher level of 5000tops, everyone may think that this is the right direction. so i think the next stage of chips will be aimed at this level. now, nvidia's thor single chip has 1000tops and qualcomm 8797 single chip has 750tops. this level is actually about the same. i think it will take a few years to reach 5000tops or even higher. let's wait and see.

traditional car companies' self-developed intelligent driving is "reinventing the wheel"

"high beam": xiaopeng is your initial partner. it has played a very good role in promoting the development of your entire intelligent driving business. have you encountered similar partners later?

li lele:at present, it seems that there may not be. xiaopeng first used our xavier ipu03 platform in the chinese market. we have made nvidia and xiaopeng successful, and xiaopeng has also made nvidia and desay sv successful, because the car is a carrier. whether our domain control product is good or not, we need a car to realize this function to prove it.

at that time, nvidia chose to cooperate with many global tier 1 (first-tier suppliers) abroad, but there was no actual mass production. judging from the actual results, it was indeed desay sv that helped it achieve mass production on the xiaopeng p7.

this project is indeed a process of mutual achievement among the three parties. it has actually proved that high computing power is the future, so many domestic customers are using nvidia's orin platform. in addition to new forces, such as great wallwei brandnew blue mountain,lynk & co z10、zeekr007 and others all use desay sv's domain control.changan qiyuan a07we are the ones who are doing the advanced intelligent driving, and some other partners are developing it. of course, we also have joint venture brands, such as saic-gm and cadillac, which actually shows that everyone believes that this direction is right.

so, from this perspective, our cooperation with xiaopeng has set a new direction and has a huge impact on the entire industry. therefore, similar cooperation may be difficult to happen again in the future. it has made traditional oems realize that high-level autonomous driving can be achieved through high computing power + advanced ai algorithms. everyone has reached a consensus and is motivated to actively deploy.

"high beam": nvidia xavier chose to cooperate with the global tier 1 but later did not go into mass production. what was the reason?

li lele:first of all, we must realize that at that time, domain controllers were a new product with high technical barriers, and we did not have a very definite solution to guarantee success. we also did not understand the actual situation of our competitors in product development.

from our own perspective, desay sv made huge investments because we believed that this product could be successful due to the rise of the intelligent wave at that time. this has to do with our strategic focus. on the other hand, it has to do with the great efforts made by our product team. we will try our best to promote the mass production of the product.

high beam: how do you think traditional manufacturers can advance faster in the intelligent driving race? is the path of self-development feasible?

li lele:we believe that car manufacturers’ self-development is “reinventing the wheel”. the algorithm of autonomous driving itself is still evolving rapidly. we are constantly investing in it, and we will continue to invest. the final goal of intelligent driving is driverless driving, and this process requires continuous investment, so we believe that reinventing the wheel has no social and economic value.

we believe that traditional car companies should cooperate with multiple parties to speed up the installation of intelligent driving functions in their products and allow consumers to use them. this model may be the best option for traditional car manufacturers.

in addition, if the end game is driverless driving, there will be no differentiation and everything will be homogenized. it will actually be meaningless for new forces to increase investment and actively conduct independent research and development.

why do so many people do this now? because everyone thinks it still takes a process, and if i can do it quickly, it may be valuable to my company. so in the long run, if homogeneous algorithms converge more and more, and functional performance converges more and more, i think it will be meaningless for everyone to do it.

"high beam": you think that the high-investment model will gradually stabilize, is that right?

li lele:high investment depends on who makes the investment. we believe that it is not very meaningful or necessary for oems to make high investments.

in the past few years, when the general direction was uncertain, we could do it. especially for xpeng, intelligent driving is its corporate attribute and label. everyone thinks it is a technology company, not just a car manufacturer. it is leading in autonomous driving and can compete with tesla. being the best in china has brought great value to its financing and valuation as a whole.

however, just like when we race, the finish line is the same, it’s just a matter of who runs faster and who runs slower. for car companies, facing the end of autonomous driving, the economic benefits of self-development may decrease when the functional experience is homogenized. we believe that this time point will come faster.

the 150,000 yuan market will be released quickly

high beam: from the perspective of market and competition strategy, what is your strategy? increase market share first?

li lele:our role is different from that of car manufacturers. desay sv’s strategy is to gain a high market share because we need to increase and continue to invest.

same as the above point, we believe that there is no need for car companies to invest heavily. reinventing the wheel, collecting data repeatedly, and building computing data centers for training repeatedly will not bring better social and economic benefits. as a supplier, desay sv will do a good job and everyone will use it. this model has higher economic benefits. therefore, we will insist on increasing investment in this area.

"high beam": next, in the next few years, what are your target customer groups?

li lele:in the domestic market, the first category is the earliest batch of new force customers. we are already working on their next-generation products, but the cooperation model will change. if customers want to do more, we will do less. we will continue to cooperate with these manufacturers.

the second category is traditional car manufacturers, which are also following up quickly. it can be seen that their market share is rising rapidly, and they are also building high-end brands at the same time. this category is becoming the second wave of customer groups.

the third category is joint ventures and overseas brands. in overseas markets such as europe, japan, and the united states, except for tesla, the advanced intelligent driving progress or mass production of brands such as gm and ford is not fast. therefore, we are now vigorously expanding the european and japanese markets. i think this should be our main customer group for the next wave of customer acquisition.

high beam: how are european customers doing? are they difficult to deal with?

li lele:we are in communication with several european brands. compared with the chinese market, the european market is more traditional and conservative. if we want to achieve high-level intelligent driving, we will mainly mass-produce high-value vehicles. after scale, the cost will continue to be shared, and then we may slowly penetrate into low-end brands.

we have been communicating for a year and the progress is pretty good. i think the next wave of customers will come from european car manufacturers. we have already made the japanese market public. we officially signed a contract with a leading japanese car company around the spring festival this year and are currently in the development process.

"high beam": for these overseas brands, you not only sell products in the chinese market, but also in overseas markets?

li lele:yes, different oems have different ideas. for example, we have started to cooperate with another japanese customer, whose products are developed and produced in the chinese market, and then exported as complete vehicles.

because china's vehicle industry chain has a huge cost advantage, with fast r&d speed and low cost. in other words, the performance of chinese vehicles is the same as that of overseas vehicles, but the cost is much lower. if they are at the same price, our performance may be much higher. after their vehicle products are exported, our products will go overseas with them.

it is not just chinese car companies that are going overseas, but also joint venture brands and overseas brands, which are going overseas through chinese supply chains and chinese technology manufacturing. i have seen many oems have such strategic discussions and plans.

"high beam": in terms of growth, which type of customers will have more growth in the future?

li lele:first of all, the chinese market will still have a relatively large growth. in fact, in the past few years, there have been many new forces in mass production of smart cars, such asnio, ideal, xiaopeng, zeekr,ask the worldcars of brands such as are indeed very attractive, but in fact these cars are targeting the high-end market and are priced relatively high. their total shipments actually account for a relatively low proportion in the broad chinese passenger car market.

in addition, there is a huge market, such as phev models, and even other brands or market segments with lower prices, which are very large in scale.

"high beam": using price ranges, which price range will have the largest increase in products?

li lele:i think that in the price range of around 150,000 yuan, between 150,000 and 200,000 yuan, smart driving products have not yet achieved a wide penetration, so i believe that this incremental market will be further released and develop rapidly.

"high beam": you are currently a bu with a population of about 1,000 people. in the future, you will have more and more customers. how do you ensure delivery and follow-up guarantees?

li lele:this is a very good question. first of all, i think it is very important to have strong organizational capabilities. the platform we developed with xiaopeng in the past is called ipu03. from the start of development to mass production and delivery to xiaopeng, it took us two years to develop this p7, which concentrated a lot of company resources at that time.

however, when we were working on the nvidia orin platform, we had already provided multiple solutions to four or five customers in about ten months. for example, some customers wanted a single orinx, some wanted dual orinx, and the operating system required linux, so it was a very complex combination of solutions that could be developed in parallel for different customers.

by comparison, we can see that our overall capabilities have been greatly improved. of course, we have also increased investment in resources to ensure product delivery. in the subsequent actual project process, the fastest speed from helping customers to develop to mass production is half a year to eight months, such as gac, great wall and other customers.

our capabilities have improved qualitatively over the past five years, and we are far ahead of the industry in terms of development speed and quality.

secondly, more of our customers will come from overseas in the future, and their process system requirements are different. we will continue to increase investment, and even set up a r&d center in europe to serve european customers. these tasks are being deployed, planned, and carried out in an orderly manner.

"high beam": how is your cost competitiveness in this market?

li lele:from the perspective of cost, the solutions are now diversified. for example, the cost performance of qualcomm's solution is higher. we also have a first-mover advantage in china. this product can be installed on fuel vehicles without water cooling, which will greatly increase the installation rate. this product is priced at less than 3,000 yuan.

"high beam": is it so low? what functions can it achieve?

li lele:yes. high-speed noa+ city memory driving can be achieved within 3,000 yuan.

"high beam": therefore, we should start popularizing it from highways.

li lele:yes. now we see that l1 and l2 may become standard configurations. if the cost of high-speed noa can be reduced to more than 2,000 yuan, it may replace the l1 and l2 product markets in the future. its penetration rate will increase significantly and it will gradually become standard configurations. we believe that high-speed noa will gradually become standard configurations.

“what we want to overcome most is european luxury brands”

"high beam": i noticed that your intelligent driving domain control installed capacity ranked second in 2023, with tesla's oem factory ahead and many competitors closing in behind. what do you think of the current competitive environment for this business?

li lele:indeed, the industry has been following up on intelligent driving domain control very quickly in recent years. competition means that this product has a good prospect, and more people will make more money. many companies claim that they can do it, but in fact, large-scale mass production is very difficult. desay sv's principle is to do things well, laying the foundation layer by layer from the underlying software to the middleware, and has also done a lot of safety certifications, and continues to do pre-research and investment in technology evolution.

high beam: unlike previous years, the competition in the intelligent driving field is very fierce now. what core competitiveness do you have now to ensure that you can get customers? now the new forces also do domain control and have their own chip algorithms. basically, they also have the full stack.

li lele:desay sv is positioned as a supplier. in the ecosystem, we hope to become the preferred partner for mobility transformation. most of our customers are car manufacturers. if you want to provide value to customers, you must have your own value. therefore, we clearly stated in our strategic deployment that we need to build full-stack capabilities.

at the beginning, we also mentioned that we need to have full-stack capabilities, but full-stack capabilities do not mean that we have to provide full-stack products. of course, providing full-stack products must be a business model. if our customers do not have such development capabilities and need intelligent driving products, then we can be responsible for the production of this part.

in addition, it is fine for new forces to do self-research, but they do not produce. we have a strong supply chain manufacturing system and a good cost control system, so we can cooperate in development and production. second, car companies may have relatively little experience in the development of domain control, and there may be certain difficulties in development. we can empower customers in some software modules and improve our cooperation stickiness in hardware products.

as mentioned earlier, ideal also wants to develop its own second-generation products, but we can still supply it with hardware, which is also a cooperation model for our business.

"high beam": ideal used your first-generation orin platform. how did you negotiate the deal?

li lele:we have been cooperating since ideal’s first car. later, due to regulatory reasons, the relationship between the two parties has always been very close. our first smart cockpit was installed on the ideal one, and we have experience in mass production of ipu03 xavier, so the subsequent cooperation was a natural outcome.

"high beam": to what extent can car companies achieve domain control? why do they do it themselves?

li lele:first of all, car companies have no way to produce. doing domain control by themselves is mainly for autonomous control, and this is the core purpose. of course, the r&d department of the car manufacturer has to prove its value. currently, it is mainly reflected in new forces such as wei, xiao, and li. most car manufacturers do not adopt this model.

high beam: have you been involved in the integration of cabin and driver? what do you think of the future trend?

li lele:in the field of cockpit-driver integration, desay sv has a clear advantage because our cockpit-related capabilities are leading in china, and our intelligent driving, controller hardware, and middleware software are also leading in china. cockpit-driver integration requires strong capabilities in both areas, and we happen to have both of these capabilities.

however, the development of cabin driving is currently mainly dependent on the development of chip technology. currently, only qualcomm has just launched the 8775 cabin driving chip, and the domestic company black sesame is also developing low-end chips. we are currently developing a project with qualcomm and expect to form a fixed point with customers in october.

"high beam": after listening to you as a whole, i feel that you are still full of confidence. although the market is very competitive, desay sv can always find its own living space.

li lele:i think we need to have a very full understanding of the competitive environment and future development trends. we cannot say what direction we feel we are heading in, as these are rather theoretical. on a practical level, each of us has its own difficulties in development.

so i think that first of all, you must know very clearly who your customers are in this industry ecosystem, what their needs are, and how to plan your own capabilities to provide different values ​​to different customers. i think these are very important.

"high beam": when the intelligent driving team talks about customers, what type of customers are the most difficult to overcome?

li lele:the most difficult thing is the one i haven't gotten yet. what i want to get most now is the european luxury brands. if i get them, and look back after many years to see if it was difficult, it may not seem difficult, but it just seems difficult at the moment.

high beam: what is the difficulty for luxury brands? is it because their assessment system has too many dimensions, or is there some other reason?

li lele:for european companies, building trust is very important. for example, even if i show you my products, cars and various capabilities, the opportunity to be selected may not necessarily belong to you. it has a set of process systems, especially for european car companies, which have very high requirements for the methodology system. however, they also recognize the technical capabilities of chinese companies, the leading position of china's intelligent driving, our current industry status, etc. it's just that it may take a supporting process to really decide.

“there are fewer and fewer real players in the field of intelligent driving”

"high beam": what do you think about tesla's fsd landing in china?

li lele:we believe that the introduction of fsd into china can accelerate the further development of china's intelligent driving, and it can play a "catfish" effect. we do not think it is a "shark" as if it will kill all chinese companies. in fact, it is not. i think it is a "catfish" effect that can stimulate the vitality of chinese companies.

during the epidemic in 2020, the intelligent driving industry in the chinese market was under great pressure. in the first half of the year, everyone was restricted in their innovation in intelligent driving, but in the second half of the year, they entered a stage of rapid development. i think this has a lot to do with tesla's progress.

if tesla fsd enters china, it means that the world's largest automaker with the highest level of autonomous driving, the strongest innovation capabilities, and the most capital will enter china, and everyone will take it as a benchmark and work hard for it.

for example, in the first two months of this year, many senior executives of domestic car companies went to the united states to test drive waymo and the unproduced experimental version of fsd. after the test drive, they had different feelings. a while ago, the latest version of fsd was released, and he xiaopeng and other senior executives of car companies went to the united states to test drive their cars. after returning, everyone was more supportive and spoke on weibo, thinking that this was a direction. after entering china, you don’t have to go abroad to try it. you can see its benchmark effect in china. if everyone wants to win, they must be as strong as tesla, or even better than it. so i think it can play a catfish effect.

high beam: how big do you think the gap is between china's intelligent driving level and tesla's? in terms of time, how long will it take to catch up with tesla?

li lele:i feel that china's overall intelligent driving capabilities are not much different from tesla's. many domestic car manufacturers are developing their own algorithms, and third-party algorithms in the industry are also working hard on them. the gap should not be large, and may take one to one and a half years.

however, we are uncertain about the future. for example, in the longer term, tesla's goal is to release a 5000 tops computing power solution by the end of next year. at the same time, it has spent billions of dollars on data center computing cluster servers, so tesla's investment in ai capabilities, training capabilities, and data processing capabilities is huge.

in addition, it has a huge data pool, and its data volume may be more than 2 billion miles. huge data is also a kind of ability and wealth. so from this dimension, we are not sure about the gap with it. however, i think chinese manufacturers are really very powerful. no matter what, i have to catch up and beat it. this is a very tenacious characteristic of the chinese people, and it is also an excellent quality. based on the above views, i think we will not lose to tesla in the future.

high beam: in terms of market competition, will there be significant changes before and after the epidemic? from our current experience, more and more cars have intelligent driving capabilities, and everyone is competing fiercely. but from the supply chain perspective, will the competition during the epidemic be more intense?

li lele:we also developed very quickly during the pandemic, but the past two years will be different. on the one hand, we have developed domain control in 2021 and 2022, and today we are further exploring its value. we originally provided fixed-point support to new forces, but now more traditional car manufacturers may have seen the direction and are also catching up, including joint venture brands. therefore, for us, we need to continue to explore the value of our products.

second, chip technology is also developing. we must emphasize the rapid development of chip technology. algorithms are also developing rapidly, which will bring lower-cost and higher-performance solutions. low-cost solutions may make your market scale larger, so we see that competition is slowly becoming diversified. however, competitors are converging. in 2022 and 2023, desay sv will mainly focus on high-computing platforms and high-end markets. many startups will have multi-chip solutions and combined chip solutions developed in a heterogeneous way. at that time, there will be more players in this solution, and everyone is doing it.

now the competitors have converged, but there may be more and more solutions. from what we can see, there are fewer and fewer real players in the field of intelligent driving.

"high beam": the elimination round has begun? why did they restrain themselves? is it because their products are not good or because of funding problems?

li lele:you mentioned at the beginning that intelligent driving is quite competitive. in the chinese market, most of the intelligent driving companies are start-ups. how do start-ups make profits? now they are investing a lot. the algorithm changes every year and is overturned every two years. they continue to develop a new generation of products and research another generation at the same time. this kind of investment is actually very large.

second, most startups rely on financing, but they can only get financing if they have customers and projects. regardless of whether the project makes money or not, i must first have a project. so in terms of competition, many startups use very low prices, or even lose money to do projects regardless of cost, which leads to further deterioration of their operations.

so we see that most startups have basically exited the market in the past two years, and projects and competition are increasingly concentrated at the top. i think this is a very obvious change in the past two years.