Bonus Depth | Mi Liangchuan Resigns: Unveiling the Departure of the Core "Iron Triangle" Team of XPENG Robotics' Previous Generation
36Kr's "Workplace Bonus" (ID: ZhiChangHongLi)
The robot business has become the most pressing matter for He Xiaopeng in the second half of the year.
On June 5, 2026, "Workplace Bonus" exclusively reported that Shi Xiaoxin, the person in charge of the core products of XPeng Robotics, had left the company. Just five days later, He Xiaopeng issued an internal letter to all employees, announcing that he would personally serve as the CEO of XPeng Robotics' business, and personally lead the crucial battle on the eve of its "mass production and commercialization."
Part of the pressure comes from the goals that have already been made public. Just last month, He Xiaopeng, along with Vice President Gu Jie and the head of the Robotics Center Mi Liangchuan (LC Mi), held a mobilization meeting for the mass production of robots with thousands of participants. In addition to clearly stating that "humanoid robots will be mass-produced by the end of 2026," he also emphasized XPeng's positioning as "the only company in China with full-stack self-developed capabilities" in areas such as chips, operating systems, joints, and dexterous hands.
For the team, this also means much higher difficulty and pressure compared to peers.
He Xiaopeng made adjustments quickly. Last Friday, "Yijian Auto" reported the new organizational structure of XPeng's Robotics Center, which includes nine secondary departments (see the figure below). Gu Jie, Liu Xianming, and Yu Tao serve as the heads of the three departments. The vacancy in the Product Department left by Shi Xiaoxin's departure was filled by He Xiaopeng himself, who has a background as a product manager.
●
However, the name of Mi Liangchuan disappeared from this internal letter and the information about the organizational structure adjustment. An internal source told "Workplace Bonus" that Mi Liangchuan also had the idea of leaving the company after Shi Xiaoxin proposed to resign. "Although his Feishu status doesn't show that he has left, he is actually close to leaving without a new job lined up." "Workplace Bonus" contacted XPeng's public relations department to verify this information, but did not receive a response by the time of publication. Two hours before the publication of this article, "Yijian Auto" released an exclusive flash news about Mi Liangchuan's departure.
So far, the signal of "changing of the guard" at XPeng's Robotics Center has become more obvious. Multiple headhunters told "Workplace Bonus" that as early as the end of May last year, Zheng Cunyuan (Tyler), another person in charge of the whole machine body of XPeng Robotics who was as important as Shi Xiaoxin, had quietly left the company to start his own business. Before the debut of the IRON robot, which was suspected of having a person inside due to its high degree of realism in November last year, Mi Liangchuan, Shi Xiaoxin, and Zheng Cunyuan together formed the "iron triangle" leading the XPeng Robotics team.
So, who will be the new person in charge?
"He Xiaopeng may not have figured it out yet," an informed source revealed to "Workplace Bonus." After Shi Xiaoxin proposed to resign, He Xiaopeng contacted a well-known expert in intelligent driving. Although they didn't talk about recruitment, He Xiaopeng's motivation for the meeting may include "wanting to verify whether he can recruit a person in charge across fields again."
"We don't want to see another cross - field leader who didn't do robotics before," a front - line employee commented. Looking back at Mi Liangchuan's previous position and contributions in the team, the employee believes this should be regarded as a trial - and - error in XPeng's employment.
The ambiguity and vacillation in the high - level employment decisions have been affecting XPeng Robotics' recruitment, and this time is no exception. "At the beginning, they could consider candidates from cross - fields for high - end positions. For a while, they preferred those who had directly worked on robots. Recently, they can consider cross - field candidates again," a headhunter said in confusion. "XPeng is not a company that can attract talents with high salaries. Its biggest selling point is that its embodied scenarios are closer to data feedback."
Multiple headhunters' experience when recruiting for XPeng Robotics is that the relevant HR contacts leave or are replaced every few months. "It's possible that they are replaced just when they have just established knowledge of the industry."
To understand the pressure behind the "year of mass production" of XPeng Robotics, the technical aspect is not the top priority.
Former cross - field leader ╱ 01
The "iron triangle" ╱ 02
Mass production ╱ 03
Vacillation ╱ 04
Former cross - field leader
"Did Mi Liangchuan have any actual output?" A person in the industry questioned Mi Liangchuan's value in XPeng Robotics' business.
In previous public exposures and some media articles, Mi Liangchuan (LC Mi) was "undoubtedly the person in charge" of XPeng Robotics [1]. A former executive at NVIDIA, an alumnus of the University of Science and Technology of China, and a researcher at Carnegie Mellon University... Under these halos, he joined XPeng Motors in 2021 as a senior director of autonomous driving. In September 2023, he took over the Robotics Division. Subsequently, he further served as the chairman of XPeng Motors' AI Technology Committee and was promoted to the company's vice president. Judging from his resume, he straddles XPeng Motors' business and the robotics business, and is a representative of the "XPeng system" in a real sense.
"Mi Liangchuan is a typical 'report - type talent' and hardly participates in substantial business," a former employee commented. An HR who worked with Mi Liangchuan believes that Mi Liangchuan is good at upward management, knows He Xiaopeng very well, and knows how to communicate with him. "He almost has the user manual for He Xiaopeng." In 2023, during the special period when XPeng's robotics business was transformed from an independent company to a group division, Mi Liangchuan helped the organization re - establish its culture, and his cross - business resume also played a certain role in the internal resource coordination between the robotics and automotive businesses.
The problem is that although "upward management" can help a new business obtain resources and time within the organization, it can also create potential risks during the mass - production period - a vacuum in middle - level execution. "The most feared thing is that the boss thinks everything is fine, but the front - line staff know that nothing has been implemented."
"Leading a robotics business with an automotive background is destined to have a gap," an HR from another robotics company analyzed her recruitment experience in the past six months to "Workplace Bonus." She believes that the biggest problem with Mi Liangchuan, as the person in charge of the robotics business of a well - known company in the industry, is that "he doesn't understand robotics well enough." "It's not that there is a problem with his basic ability, but he may not be in the right position."
In his public resume, after graduating from undergraduate, Mi Liangchuan pursued a master's degree in robotics at Iowa State University in the United States and served as an electrical engineer at the Robotics Institute of Carnegie Mellon University (CMU), participating in the development of the obstacle - avoidance system for medical service robots and the design of the laser navigation scheme for drones.
"Fairly speaking, Mi Liangchuan is not a completely 'nominal' leader. His educational background seems to be close to robotics, and there is a possibility of leading across fields. However, his hands - on experience in his subsequent formal work is more in line with the AI - Native (NVIDIA, autonomous driving, deep learning) profile, rather than the Embodied - Native (systematic understanding of the hardware supply chain, mechanical structure, motion control, and manufacturing yield) that investors currently favor the most."
Previously, an article introducing Mi Liangchuan said that he "innovatively translated the 'automotive - grade' standard and data closed - loop system in the automotive industry to the robotics track." The problem with this is that the automotive - grade thinking emphasizes process, standards, and zero tolerance for errors, while the current mass production of robots requires rapid iteration, tolerance for trial - and - error, and flexible adjustment, and it is not yet at the stage where automotive - grade standards can be applied.
A more vivid way to put it is that the person in charge of the robotics business on the eve of mass production is preferably someone who can lead the team from the laboratory to the production line, who can monitor the yield in the factory and stand up to suppliers. Looking back at XPeng's Technology Day at the end of 2025, when Mi Liangchuan explained the implementation path of the IRON's anthropomorphic gait on - site, he said: "This is an inflection point of our generative controller. But I can't really say which optimization brought about the change. I can only say that when the data and computing power reached a certain level, a step - change occurred."
"The person in charge is preferably someone who directly understands the industry. The people below can come from different industries. If you want to hold a high - level position in a team with a science and engineering background, at least your professional ability should be recognized by the team, and you should have deep accumulation in vertical professional knowledge and have invested a lot of energy and time in understanding this technology," an analysis suggests. Perhaps because He Xiaopeng has experience in crossing from the Internet to the automotive industry, he also places more hope in cross - field employment.
[1] "Who is Leading XPeng Robotics: The Four Key Figures Behind IRON" by QbitAI
The "iron triangle"
However, under the coordinated leadership of Shi Xiaoxin and Zheng Cunyuan - two "1.5 - position" figures who were almost invisible to the media before - and Mi Liangchuan, XPeng Robotics still presented the debut of the new - generation IRON robot at the end of last year as scheduled.
"In terms of administrative relations, Shi Xiaoxin has a higher rank than Zheng Cunyuan, and they both report to Mi Liangchuan." In this "iron triangle" system, Mi Liangchuan is responsible for reporting upwards and dealing with the outside world, Shi Xiaoxin focuses on product management, and Zheng Cunyuan is in charge of the whole - machine technology.
As a veteran who has been with the company since the Pengxing Intelligence era, Shi Xiaoxin, the former senior director of product planning, is one of the few people in the team who has been responsible for both technology and productization. He and Zheng Cunyuan experienced the transformation from Pengxing Intelligence to XPeng's Robotics Center and went through the lay - off period together (only a few dozen people from the old Pengxing team were retained at that time). "Only they two remained in the key leadership positions."
According to "Workplace Bonus," among the three, Shi Xiaoxin has a low - key and delicate personality and pays more attention to balance. "Mr. He values him very much. If he hadn't chosen to leave, he would have had the opportunity to retire at XPeng," a team member said with regret. "He is in charge of things like selecting materials and steel gratings in the factory. The drawback is that he doesn't know how to report upwards."
Zheng Cunyuan left the company as quietly as Shi Xiaoxin. According to public information, Zheng Cunyuan graduated from Columbia University with a master's degree in mechanical engineering. He founded Harmony SHR (which develops exoskeletons) during his undergraduate years, interned at SpaceX, and served as a researcher at General Electric's GE Research in the United States during his master's studies. Before joining XPeng's Robotics team, he served as the R & D director and product operation officer of industrial robots at Tuosida Technology in China.
On Maimai, Zheng Cunyuan's work experience has not been updated and still shows XPeng. "The background wallpaper of Tyler (Zheng Cunyuan's English name)'s WeChat Moments is still IRON. He even posted a message in his Moments to congratulate his former employer when IRON made its debut last year." Before leaving the team, he completed the general contracting and integration work of the latest - generation IRON at that time. "Tyler has a more distinct personality and speaks directly. Sometimes he may seem to be bluffing, but he is down - to - earth in his work."
Around May this year, a co - founder named "Zheng Tianle" appeared in the press release and announcements of Chaowei Dynamics. After cross - verification by "Workplace Bonus," "Zheng Tianle (Tyler)" is Zheng Cunyuan.
How did the three people with different personalities cooperate during their work?
"They made the most of their respective strengths and complemented each other's weaknesses. They supported XPeng Robotics for two years," a former employee believes. The three people's thinking about the "first - principles" of robotics and the future business direction is highly similar, and there are no major differences in key decisions regarding the business direction. "They highly emphasize the first - principles, starting with hardware, incubating algorithms at the same time, and accumulating intelligence." In terms of radicalness, He Xiaopeng once publicly said that Mi Liangchuan was more "radical" than him in building robots, and he himself wanted to "build humans" rather than just "build robots."
● Due to the compression limit of WeChat images, reply "XPJQR1103" in the background of the official account to get the high - definition version.
Mass production
The mass - production goal has caused differences.
According to an informed source, in the early stage when the iron triangle worked in harmony, "they could set their own goals, and He Xiaopeng would be more patient." Before the new - generation IRON was launched, after the mass - production goal was proposed internally, He Xiaopeng began to give more macro - strategic guidance, and all parties could no longer reach a high - level consensus.
He Xiaopeng's stance is easy to understand: First, the development of software and hardware cannot always be in the laboratory mode. The maturity of the product ultimately comes from delivery. Second, without large - scale production, the cultivation of the industrial chain will always remain at the level of hundreds or tens of units, and it will be difficult to reach the automotive - grade or consumer - grade quantity. Third, as a listed company, XPeng needs to align with the external consensus and meet the expectations of investors.
The problem is not mass production itself, but "how to mass - produce."
At present, the robots of any company are in a stage of rapid iteration. If the initial mass - production quantity is set too high, once the product's performance fails to meet the market expectations, neither the company itself nor the entire industry can guarantee to bear the consequences. How to control the scale and quantity is the core of the problem.
An internal view once held that "the focus of assessment should be to define what capabilities are achieved behind the quantity, whether a mass - production delivery standard and a good long - term relationship have been established with high - quality suppliers, and whether the stability of the software - hardware integration is infinitely close to the product stage - rather than focusing on the figure of ten - thousand - unit shipments."
The essence of the difference is the conflict between two completely different "product genes." XPeng Motors' revenue reached 76.7 billion yuan in 2025. In the automotive - grade and consumer - grade thinking, revenue is meaningless if it does not reach a certain scale. However,