Baidu high-profile car manufacturing, its car manufacturing company Jidu frequently discloses progress to the outside world. The latest development is that Jidu has obtained another US$400 million financing from Baidu and Geely, raising US$700 million (about 4.5 billion yuan) in less than a year, once again verifying that car manufacturing does burn money.
Baidu plans to mass-produce the first car next year, which is expected to shorten the time by 1-2 years compared with Wei Xiaoli and so on. This reflects that the window of electric vehicles is shortening, and also reflects the urgency of Baidu to promote the commercialization of autonomous driving technology.
Eight years of research and development layout, Baidu autonomous driving this road is not smooth, frequent loss of talent, business model wavering, into the car is still spit out late set, the next five years or so or need 74 billion yuan of capital investment, but profits still face many challenges, Baidu's new test is coming.
Talent is frequently lost, and the strategy is swinging back and forth
Baidu's layout in autonomous driving has been around since 2013, when it successively established a deep learning research institute, a big data lab, and an artificial intelligence lab. In 2017, driven by the strategy of then-president Lu Qi All in AI, Baidu established the Intelligent Driving Business Group (IDG) and launched the Apollo autonomous driving platform covering hardware and software systems.
In this process, Baidu pulled many talents to join, known as the "father of Google brain" Wu Enda was personally invited by Robin Li, Baidu then gradually formed a top-of-the-line automatic driving team represented by Wang Jin, Wu Enda, Yu Kai, Ni Kai, Han Xu, Peng Jun, Lou Tiancheng, Huang Chang, Wu Xuebin, etc.
However, with the frequent changes in the business organizational structure, these talents have left successively, especially in 2016 and 2017, the most prominent, and have founded autonomous driving companies to go to the opposite side of Baidu, such as Wang Jin and Han Xu founded Wenyuan Zhixing (formerly known as Jingchi Technology) after leaving, Wu Enda founded Drive.ai (which has collapsed), Ni Kai founded Heduo Technology, Yu Kai and Huang Chang founded Horizon, and Peng Jun and Lou Tiancheng established Xiaoma Zhixing.
The most recent time to attract greater attention from the outside world is Gu Weihao, the former general manager of the intelligent automobile division who left at the end of 2019, who joined Great Wall Motor as CEO in February last year, and many engineers left without receiving attention from the outside world. It is difficult to say that the departure of these early participating technology celebrities has not affected the direction and development of Baidu's intelligent driving business.
According to the latest data, as of the third quarter of last year, Apollo has accumulated more than 16 million kilometers of test mileage at the L4 level, an increase of 189% year-on-year, obtained 411 automatic driving test licenses, and has more than 210 global ecological partners, bringing together 65,000 global developers and 700,000 open source codes, becoming the world's most active open platform for autonomous driving.
But for a long time in the past, Baidu Apollo has not explored an effective business model, and even has wavered back and forth. It was not until the release of the first quarter of last year that Robin Li first clarified the three business models of Baidu Intelligent Driving (Apollo): providing autonomous driving technology solutions for OEMs, car manufacturing, and sharing unmanned vehicles with Robotaxi as the core.
For these three ways of playing, Zhang Xiang, an analyst in the automotive industry, told Sohu Technology that the most unsatisfactory thing is to share unmanned vehicles, and the car is relatively optimistic, and the prospect of selling solutions is not very good. He believes that car manufacturing and selling solutions cannot be carried out at the same time, and huawei, can not be both a supplier and a car company, car manufacturing and customer competition, car companies will be more taboo. A domestic head OEM automatic driving people also hold a similar view, the most feasible is to sell cars, the solution will not be particularly good, L4 level Robotaxi is not feasible.
"If L4 can work now, L2 or L2.5 is not necessary to do it, the main reason is still technical reasons, and Baidu's focus is not enough, the layout is too scattered, and many projects have been killed." The person said that the L4 accident is the responsibility of the company, and now it is impossible to achieve real "mass production", that is, there is no one in the driver's seat, which is possible in a very limited range, but it is difficult to expand the scope, especially to open roads.
It can also be seen from this that Baidu covers the mainstream business model of automatic driving technology, in the field of sharing unmanned vehicles, Baidu is more multi-scene coverage, in Robotaxi, unmanned trucks, unmanned buses, mines and parks, etc. have layouts, greed for perfection or to a certain extent to distract energy, and Baidu announced last year to join the tide of car-making, there is no lack of external doubts.
This is also Baidu's own helplessness to punch in the face. As early as three years ago, Robin Li said that Baidu will not build cars, only do the best automatic driving technology research and development, and open up, and all the enterprises that are interested in unmanned cars, have ideas and capabilities to do it together. Companies such as Alibaba and Huawei that want to intervene in the automotive field are similar ideas, but it is clear that Baidu lacks determination.
Forced to build a car and shout out L4 is a gimmick?
Most in the industry believe that behind Robin Li's self-punching face is a forced choice under the wall of Baidu's autonomous driving solutions. As the new competitive core of the car after the electrification, especially in the era of software-defined cars, whoever masters the core technology will have a higher right to speak.
Most car companies are reluctant to give this right to speak to people, and last year, the "soul theory" of Chen Hong, chairman of SAIC Motor, expressed the voice of some car companies. Previously, it was also mentioned that Ideal and Xiaopeng had negotiated investment matters with Baidu, but Baidu hoped that they would only use Apollo as an autonomous driving solution, but Ideal and Xiaopeng did not want to be subject to others, and eventually dispersed.
The above-mentioned OEMs also revealed that there have been more contacts with Baidu before, there have been cooperation intentions, the feeling is that "it is better to hurt ten fingers than to break a finger, but Baidu often prefers to hurt ten fingers, rather than cut off", and finally his main engine factory did not reach cooperation with Baidu and chose to build its own automatic driving system. He believes that basically first-class and second-rate or stronger machine factories generally do not cooperate with Baidu, and the whole machine factories that do not have the ability to develop their own may have more willingness to cooperate.
Therefore, Baidu, which has been wary of car companies, has embarked on the road of car manufacturing, which does not surprise the outside world. In January last year, Baidu officially announced that it would enter the automotive industry as a vehicle manufacturer, and reached a cooperation with Geely Holdings, and established a joint venture in March. Baidu leads the overall design, software and hardware research and development with 55% of the shares, and currently has no plans to build its own factory, while Geely Holdings, which provides a pure electric SEA Haohan architecture, has been largely reduced to Baidu foundry, and the early participation is not deep.
Baidu car to play the concept of automotive robot, less than 10 months to complete the development of software integration simulation prototype CAR SIMUCar, in February and March this year will use SIMUCar to carry out mass production function development. Jidu said that at this stage, the automotive robot will have the L4 level automatic driving capability of urban domain and high-speed domain integration, which can achieve point-to-point automatic driving to meet the needs of mass production.
It has been mentioned many times in the external caliber of Jidu that the automotive robot will have L4 level automatic driving capabilities. Robin Li said at the Create 2021 conference at the end of last year that Jidu's upcoming auto robot has L4 level automatic driving capabilities and can move freely.
This statement has been questioned by the outside world. "Now car companies generally do not take the initiative to publicize what level of its automatic driving technology is, and the landing of automatic driving is actually more difficult than we expected, and the current L4-level mass production car is completely out of hand in a city or an area, which is not realistic." Zhang Xiang pointed out that Baidu's first car cannot reach this level within one or two years, which is more out of public relations or preparation for financing.
It is understood that the intelligent driving system of Jidu comes from the intelligent driving products launched by Baidu Apollo, mainly including AVP (autonomous parking system) and ANP (pilot assisted driving), of which AVP only cooperates with WM W6, GAC Aean V Plus, Great Wall WEY Mocha, ANP is currently only mass-produced on W6 WM, and will also be applied in Jidu automobile.
According to Baidu, ANP is based on the L4 level of autonomous driving pure visual city road closed-loop solution Apollo Lite, through L4 technology dimensionality reduction, high-precision map, and light sensor light calculation power production program, to create a parking / high-speed / urban three-domain integration product, can provide users with the starting point to the end of the whole process of nearly zero intervention auxiliary driving capabilities, is the only L2+ technology application that continues the Robotaxi lineage.
That is to say, AVP and ANP are actually dimensionality reduction applications of high-level automatic driving solutions, and are the application of L4 level automatic driving technology in L2+ functions. So, how to define the level of autonomous driving of Baidu's first car? Some industry insiders interviewed by Sohu Technology believe that it should be defined from the level of automatic driving to which the function belongs, and the WM W6 equipped with Baidu ANP + AVP function is defined as a model with L2 + automatic driving capabilities, and it is easy to mislead if the L4 is hyped.
From this point of view, Baidu's first car may be difficult to surpass in intelligent driving. At present, models that claim to be able to achieve such capabilities include Tesla Model 3, Xiaopeng P7 and Xiaopeng P5, WEILAI ET7 and ET5, Polar Fox Alpha S Huawei HI Edition, etc., which are both based on pure vision-led models and lidar-led models.
Zhang Xiang believes that the advantage of Jidu car manufacturing is that it has accumulated more experience in the field of automatic driving, and has strong strength in integration, software and hardware, etc., but if the first car wants to be outstanding, it must make automatic driving different, which is a higher level than other car companies. If homogenization is more serious, consumers may still be less receptive.
In addition, for Jidu, a reality that must not be faced is that Baidu lacks C-end market experience, the brand image has been damaged before, and the entry time is late, as the market competition becomes more and more fierce, whether it can get user recognition still needs to be questioned.
Or it will need to burn 74 billion, and the profit will still be hopeless
After more than eight years of continuous investment, Baidu's commercial demand has become more and more urgent, whether it is rapidly advancing car manufacturing or actively promoting robotaxi's commercial operation, we can see Baidu's desire to monetize.
Previously, it was reported that Baidu IDG's employees and assets have been transferred from Baidu Group's relevant main companies to several Baidu wholly-owned subsidiaries, and some commercialization departments need to formulate and complete annual revenue targets, and Chu Ruisong, vice president with To B business experience, has been transferred as the general manager of IDG's intelligent automobile division, which is regarded as a signal to accelerate commercialization.
However, technology development, product iteration, etc. all need continuous investment, especially in addition to solutions, the other two paths need to be fed by capital, and losses will remain the norm for a long time to come. According to public data, Baidu's R& D investment has accounted for 15% in recent years, and since 2015, Baidu has burned tens of billions of yuan on automatic driving every year, and may have invested hundreds of billions of yuan so far. Robin Li revealed last year that Baidu invested more than 20 billion yuan in research and development on autonomous driving in 2020 alone, and believed that autonomous driving would not really mature in the next 10 or even 20 years, requiring continuous investment.
Zhang Xiang said that in the field of Robotaxi, Baidu technology is leading, but there are difficulties in large-scale commercialization, and it also involves issues such as policies and regulations, and the market prospects of this piece are the least optimistic. He believes that Baidu's commercialization ability in travel services is very weak, and it is currently more of a demonstration operation, and it is difficult to obtain more users, and Didi will still have advantages in the future travel market.
Robin Li does not seem to be pessimistic, he previously said at the earnings report that in 2025, autonomous driving technology will enter the stage of large-scale commercialization, robotaxi comprehensive cost will be lower than manual driving, when the single-day service orders exceed 50 million, the cost of Robotaxi will be one-fifth of the current, lower than the average price of ordinary express services, there will be opportunities to achieve normalized profitability.
This will be a huge challenge. According to Didi, the head platform of online ride-hailing established in 2012, it was not until the day of the Qixi Festival in 2020 that the global daily order exceeded 50 million for the first time, and the average daily order volume of domestic travel in the third quarter of last year was about 26.17 million, and the average single income was about 16.55 yuan.
Baidu launched the radish fast run in the third quarter of the year daily order volume of less than 1300, at the end of November last year only began a small range of fixed-point commercialization pilot, charging a starting price of 18 yuan (mileage fee of 4 yuan / km), much higher than the current price of online car (such as about 5 kilometers away radish fast run is 40 yuan, Didi Express is less than 14 yuan), and at the same time to obtain the commercial pilot of the Ma Zhi behavior a price of 4.9 yuan, relatively strong price will also become Baidu Robotaxi's later operation obstacles.
According to Baidu's plan, Baidu Robotaxi will expand to 30 cities by 2023, while building a large-scale L4-level autonomous driving service fleet of 3,000 vehicles to serve 3 million users; it hopes to expand to 65 cities in 2025 and 100 cities by 2030.
But Robotaxi is a high-cost, asset-heavy operating model, and continued expansion in the future means continued huge investment. According to the fifth generation of Robotaxi product Apollo Moon 480,000 yuan of bicycle costs, the daily order volume is equal to the number of users and the aforementioned 1 vehicle corresponding to 1,000 people Preliminary estimates, the daily service order reached 50 million about 50,000 vehicles, the required vehicle cost is about 24 billion yuan, that is, by 2025, the average annual investment of 6 billion yuan, plus operation and maintenance costs, this investment will be greater.
Relatively optimistic car building is more expensive. Since its establishment, WEILAI has raised more than 60 billion yuan, Xiaopeng has more than 50 billion yuan, and the ideal has exceeded 20 billion yuan, Xiaomi has taken out 10 billion US dollars to build a car, and Tesla has raised more than 20 billion US dollars since its establishment. Jidu CEO Xia Yiping previously said that it plans to invest 50 billion yuan in the next five years to produce smart cars.
Zhang Xiang mentioned that the investment in car manufacturing funds is very large, the threshold is very high, but the profit of selling cars is actually not high, and now the profits of the automotive industry are declining every year, and the profits of Mercedes-Benz and BMW are less than 10%. This is also a reason for Baidu's cooperation with Geely, after all, the funds are limited, and baidu is responsible for doing what it is good at.
At the same time, the profitability of car-making companies often requires a long cycle, such as Wei Xiaoli in the establishment of four or five years after the establishment of the gross margin in 2020 to achieve positive, although last year sold nearly 100,000 vehicles, still has not yet ushered in the profit inflection point, and even the loss intensified, last year sold nearly a million cars Tesla until the establishment of 17 years to achieve the first annual profit.
According to the above data, in the next four or five years, Baidu will need to burn at least 74 billion yuan, which will put forward higher requirements for Baidu's continuous cash flow. At present, Baidu's core advertising business has reached the ceiling of growth, and last year it also caused huge losses due to investment in Kuaishou, and Baidu's cash flow has become more and more tight. As of the end of the third quarter of last year, Baidu's free cash flow, excluding iQiyi, was only 2.9 billion yuan, and the demand for open source throttling was urgent. Previously, Baidu games, live broadcasting, intelligent driving and other departments broke out of layoffs, and postponed the issuance of year-end awards.
Baidu autonomous driving, there are still many difficulties to pass, after eight years of Baidu still have this patience?