21st Century Business Herald reporter Wu Xiaoyu, trainee reporter Jiao Wenjuan reports
He Xiaopeng, CEO of Xiaopeng Auto, predicted in the letter of commencement in the the Year of the Loong this February that Xiaopeng planned to recruit 4000 people and invest 3.5 billion yuan in AI technology research and development.
Recently, Xiaopeng Motors' intelligent business has once again welcomed key talents. The reporter of 21st Century Business Herald learned from various independent sources that: Suo Kunlei, with a high German background, has joined Xiaopeng Auto to be responsible for the product experience of Xiaopeng smart cockpit. His position is senior director and he reports to Wei Bin, the head of the Internet Center.
The new head of the Data Intelligence Center (DIC) is Liang Jiankun, who reports to He Xiaopeng and has been with Xiaopeng Motors for over six months. This center is the technical center department of Xiaopeng, responsible for IT system operation and maintenance, data, information system management, etc.
Suo Kunlei has been deeply involved in the geographic information service industry for many years. He joined Gaode Map in September 2011 and served as a product manager. Maps provide key spatial information and navigation guidance for autonomous vehicle to ensure that vehicles can reach their destination safely and accurately. Therefore, map background talents are favored by car enterprises. Gaode Map is also an important source of talent for Xiaopeng Motors. Chen Yonghai, vice president of Xiaopeng's product planning, and Wei Bin, head of the Internet Center, were the former vice president and product director of Gaode Maps respectively.
Liang Jiankun, the current head of Xiaopeng DIC, has nearly 20 years of experience in the computer technology and communication industry. He has been engaged in a variety of jobs in IT related industries such as software development, hardware design, system integration, communication network, Internet operations, and calls himself "Internet entrepreneurial veteran".
Liang Jiankun is an alumnus of He Xiaopeng majoring in computer science at South China University of Technology, and the two have also worked together at AsiaInfo. AsiaInfo Technology, known as "Chinese Internet Architect", was mainly engaged in software development and system integration solutions for telecom operators at that time. After leaving CICA in 2002, Liang Jiankun successively worked for China Netcom Group and 21CN as a technical expert and CTO of Guangdong Branch. Since 2008, he has started his business in the Internet field, successively establishing Changying Network and serving as a 3N technology partner, which mainly provides technical services for some Internet enterprises. After joining Xiaopeng, Liang Jiankun took over the work of Huang Ronghai, the former head of the data intelligence center, who was transferred to the head of the human resources department in January this year.
Last August, after the departure of Wu Xinzhou, the "soul figure" of Xiaopeng Intelligent Driving, Xiaopeng's intelligent business entered a period of dynamic adjustment, with some elderly people leaving and new people joining.
According to incomplete statistics, in the past year after Wu Xinzhou's resignation, seven veteran members of the Xiaopeng Intelligent Driving team have left.
In November 2023 and January 2024, Parixit Aghera, the head of the software department at Xiaopeng Motors, and Liu Lan Gechuan, the project leader at Xpilot, resigned and joined Nvidia with Wu Xinzhou. Since January this year, Wang Tao, former head of visual perception in the US team, Liu Yilin, former senior director of autonomous driving products, Pengyue Zhang, head of the static perception module group, and Lin Yishu, head of the North American intelligent driving team, have all left Xiaopeng.
According to incomplete statistics from public reports, including Suo Kunlei and Liang Jiankun, Xiaopeng has hired 5 middle and senior executives in the field of intelligence this year.
Recently, Zhang Yu, former head of Qingzhou Zhihang Perception, joined Xiaopeng Motors. In March of this year, Xianming Liu, a former senior machine learning engineer at Cruise, a self driving company under General Motors, was appointed as the head of the Xiaopeng AI team, replacing Liu Lan Gechuan's position and responsible for the research and development of XNGP. In June of this year, Yuan Tingting, former head of Alibaba Damo Academy's autonomous driving business, took over the position of Senior Director of Xiaopeng Motors' autonomous driving products, succeeding Liu Yilin.
Overall, at the critical moment of exerting end-to-end efforts, Xiaopeng still needs fresh blood to fill the gap in the intelligent driving team.
End to end technology continues to heat up this year, with Tesla being the first leading car company. In May 2023, with the release of FSD Beta v12, Tesla became the first car company to introduce end-to-end architecture in intelligent driving. Afterwards, domestic car companies followed suit one after another. In May of this year, Xiaopeng Motors launched a full update of its AI Dimensity system, becoming the first domestic automaker to mass produce end-to-end large models for vehicles. In July, at the Xiaopeng Motors AI Intelligent Driving Technology Conference, the Xiaopeng XNGP Intelligent Driving XOS 5.2.0 public beta version was launched for release.
Xiaopeng's leading production of end-to-end large models does not mean that other car companies do not have the opportunity to catch up. A smart driving expert with over 10 years of experience told 21st Century Business Herald reporters that he believes that when entering a new technology cycle led by Tesla, car companies cannot estimate the time of new technology generation based on traditional time. Don't think that whoever spent as long, we have to spend as long
NIO and Ideal hope to gain opportunities to catch up in end-to-end technology direction through organizational adjustments. At the end of last year, Ideal established a separate team for end-to-end models, under the "Algorithm Development" team, led by Jia Peng. In June of this year, NIO's intelligent driving organizational structure was transformed from the original perception, regulation, and integration parts into a large model team and a delivery team.
Compared to Ideal and NIO, Xiaopeng, who first landed end-to-end, made the latest organizational adjustments. In August of this year, Xiaopeng adjusted the organizational structure of intelligent driving and established three new departments: AI model development, AI application development, and AI efficiency development, to split the original technology development department.
One viewpoint is that Xiaopeng has explored various solutions with and without graphics, accumulating a lot while also having a deep burden when adjusting.
But insiders close to Xiaopeng Motors expressed another viewpoint to 21st Century Business Herald reporters, that adjusting the organization is not the goal, but a means to rapidly develop end-to-end technology, and the ultimate goal is to compete with which car company has better end-to-end results after getting on board. He believed that other car enterprises adjusted their organizational structure before the implementation of the end-to-end model, "to speed up the implementation". And Xiaopeng chose to adjust after mass production is implemented, in pursuit of achieving "further" end-to-end vehicle effects. After getting on the car end-to-end, car companies compete for the system's fast iteration capability. If they cannot achieve a delivery iteration every 3-4 months and achieve exponential growth in effectiveness, it is a "pseudo end-to-end" competition.
Xiaopeng's goal is to become a global pioneer and popularizer of AI intelligent driving. Implementing end-to-end means that the transition time from autonomous assisted driving to fully autonomous driving is significantly reduced. According to He Xiaopeng's plan, Xiaopeng Motors is expected to achieve "door-to-door" intelligent driving in the fourth quarter of 2024. By 2025, Xiaopeng will achieve urban intelligent driving experience comparable to high-speed intelligent driving.
But the road from end-to-end to fully autonomous driving cannot be separated from the long-term real investment of car companies, including huge computing power and massive fleet data. In the first quarter of 2024, Xiaopeng's R&D investment is 1.35 billion yuan, and it is planned to invest 3.5 billion yuan in AI related technology research and development in 2024. In the first quarter of this year, the R&D expenses of Ideal Automobile reached 3 billion yuan, a year-on-year increase of 64.6%; NIO has invested 2.86 billion yuan in multidimensional research and development such as intelligent driving.
In terms of cash reserves, Xiaopeng currently does not have an advantage. According to the financial report, as of March 31, 2024, Xiaopeng Motors' cash reserves were RMB 41.4 billion, with an ideal of RMB 98.9 billion for the same period.
Selling cars is the core source of revenue for car companies. Based on the annual delivery target of 280000 vehicles, Xiaopeng's KPI for the first half of this year was only about one-fifth achieved. In order to achieve the ambition of AI intelligent driving, Xiaopeng's most urgent task at the moment is still selling cars.