NIO, XPENG, and Li Auto's intelligent driving departments undergo major personnel reshuffles: The technology roadmap shifts to world models, and they face pressure in the second half of the intelligentization battle.

When the inflection point arrives in 2025 with the electrification penetration rate breaking through 50%, the competition logic in the Chinese automotive market has quietly changed. "Electrification determines the lower limit of an automaker, while intelligence determines its upper limit." The assertion of industry analysts has become a consensus. As the "pearl on the crown," intelligent driving is becoming the core battlefield in the second half of the competition.

Just at this time, three leading new - energy vehicle startups, NIO, XPeng, and Li Auto, have carried out a series of major personnel changes in their autonomous driving departments. In just a few months, 17 senior executives have changed their positions. Behind this personnel upheaval lies a fundamental shift in the technical route and an urgent response to the accelerating catch - up of traditional automakers. These three companies, which have not yet achieved full profitability, are struggling to balance between costly R & D and maintaining their first - mover advantages. Every adjustment they make will affect the future market landscape.

The Collective Shift in Intelligent Driving under Route Reconstruction

This major personnel change in the intelligent driving departments is not accidental. It is an active change by NIO, XPeng, and Li Auto in response to the iteration of the technical paradigm. XPeng Motors' personnel adjustment has been the most dramatic. Previously, veterans such as Park Il - soo, the former head of XPeng's North American intelligent driving division, and Wang Tao, the former head of visual perception, have left the company one after another. Yuan Tingting, from Alibaba's DAMO Academy, has taken up the position of senior director of autonomous driving products, and Liu Xianming, a former senior engineer at Cruise, has recently become the head of intelligent driving AI. He Xiaopeng himself has frequently communicated deeply with the core team, which shows the company's emphasis on the transformation.

This "replacement of the old with the new" directly serves the shift in the technical route - from the previously supported VLA (Vision - Language - Action) route to the "world base model" with the ability to simulate the physical world. Liu Xianming previously demonstrated the vehicle - control ability of this model on complex roads at the CVPR 2025 conference, which marks the transformation has entered the practical stage. At the beginning of August this year, XPeng Motors held a mobilization meeting for its autonomous driving center, chaired by He Xiaopeng himself. At the meeting, He Xiaopeng proposed to allocate all AI resources to the base model team to support the implementation of the world base model in vehicles.

NIO's adjustment presents a complex situation of "architecture reconstruction + loss of core talents." Backbones such as Hu Chengchen, the former chief expert in technical planning, Fan Haoyang, the former core of autonomous driving algorithms, and Ma Ningning, the former head of the world model, have left the company one after another. The reason disclosed internally points to differences in technical paths. However, NIO quickly merged the perception and planning and control teams into a large - model team and reorganized the integration team into a delivery team, establishing a "4×100 relay" R & D model. The official said that this move is to "absorb general AI technology and strive for the World Model 2.0." Currently, its NWM (NIO World Model) can simulate 216 trajectories within 100 milliseconds, and functions such as emergency autonomous parking on the side of the road have been mass - produced. However, the lack of game - playing ability remains a shortcoming, and there is still a gap compared with the first - tier players.

Li Auto's adjustment has a sense of "correction." The size of its intelligent driving team has been reduced from its peak to 800 people. It has gone through multiple shifts, from relying on high - precision maps in the "rule - driven" approach, to learning from Tesla's "end - to - end" method, and then to betting on the "VLA visual - language model." Under the leadership of Vice President Lang Xianpeng, the team finally established the "VLA + world model integration" route. Its OTA 8.0.1 version has a passing rate of over 90% in roundabout scenarios. Although these frequent shifts have affected the team's stability, they have also helped the company find a technical balance that suits its users' needs.

Although the adjustments of these three companies seem different, they actually point to the same industry consensus: the traditional modular intelligent driving solutions have reached a bottleneck, and the world model has become the inevitable path to L3/L4. As the industry metaphor goes, the VLA is like a "student who memorizes by rote," while the world model can "truly understand knowledge." This technological development requires organizational structure changes, which may be the underlying logic behind the personnel changes.

How to Defend the First - Mover Advantage with Competitors Closing In

The strategic adjustments of NIO, XPeng, and Li Auto are also due to the "pressure from followers" brought by traditional automakers and technology companies. In 2025, the intelligent driving track has shown a pattern of multiple strong players. The Great Wall WEY Lanshan is equipped with the Coffee Pilot Ultra system, which can operate without relying on high - precision maps and achieve full - scenario NOA. GAC Aion's NDA 3.0 end - to - end system is available nationwide. BYD's BAS 3.0 plans to achieve "full - scenario navigation" by the end of the year, and the cumulative sales of its intelligent driving models have exceeded 710,000, with continuously increasing market recognition. Relying on their large sales volume and industrial chain resources, these traditional automakers have quickly made up for their shortcomings through the "self - R & D + cooperation" model. Some of their models have entered the second tier and are on the same starting line as NIO and Xiaomi.

What is more threatening is the entry of technology giants such as Huawei. Its ADS 4.0 system ranks among the first - tier players with an ultra - low takeover frequency of 0.6 times. The hardware solution of 4 lidars + 4D millimeter - wave radar and the WEWA architecture form a technological barrier. Huawei is opening up its technology through the "HIPLUS" model, attracting automakers such as Dongfeng and SAIC to cooperate, and is building an industry - wide technological foundation, further squeezing the living space of NIO, XPeng, and Li Auto.

Road - test data shows that the technological gap between traditional automakers and new - energy vehicle startups is rapidly narrowing. Tesla cannot fully exert its strength due to insufficient localization, and there are still scenario breakpoints for Zeekr and Audi. However, leading traditional automakers have closely caught up with the second - tier new - energy vehicle startups.

Funding constraints have made this defense war even more difficult. R & D investment in areas such as intelligent driving, AI, and chips is extremely costly. NIO's R & D expenses in 2024 reached 13.04 billion yuan, with a large amount invested in the NWM architecture and self - developed chips. XPeng's self - developed chips have been postponed several times, and the efficiency problems caused by departmental barriers have increased the cost pressure. Although Li Auto has achieved profitability, the implicit costs brought by the adjustment of its intelligent driving team and the shift in technology cannot be ignored.

Facing these two pressures, NIO, XPeng, and Li Auto are defending their first - mover advantages through differentiated strategies. XPeng is betting on the "cloud - based base model + vehicle - end distillation" model, achieving an iteration every five days with a ten - thousand - GPU intelligent computing cluster. A model with 72 billion parameters has achieved vehicle - end control. NIO is taking a two - pronged approach of "self - R & D + cooperation." While promoting the NWM 2.0, it is cooperating with Mobileye to quickly implement L2+ functions. Li Auto is integrating departments such as chips, OS, and models to reduce costs through "efficient collaboration."

The smoke of the first half of the electrification era has not yet cleared, and the competition in the second half of the intelligent era has reached a white - hot stage. The personnel changes and route transformations in the intelligent driving departments of NIO, XPeng, and Li Auto are not only an inevitable choice for technological evolution but also a necessity in market competition. Traditional automakers are catching up quickly with their scale advantages, technology giants such as Huawei are building ecological barriers, and the funding pressure is ever - present. This battle for the "pearl on the crown" is destined to be a long - term one.

The personnel upheaval is just the beginning. When intelligent driving becomes the core decision - making factor for users when buying a car, only those companies that can convert technological leadership into user experience and R & D investment into profitability can ultimately win in this intelligent competition. As the industry consensus goes, electrification only determines the entry qualification, and what really determines the final outcome is the evolution speed and commercialization depth of intelligent driving technology.

Image sources: NIO, Li Auto, XPeng, Hongmeng Zhixing

This article is from the WeChat official account "Automotive Market Insights", author: Yang Shuo. It is published by 36Kr with permission.