Competition in the stock phase of China's automotive market can no longer solely rely on "involution" to break through; the development strategy for the second half of the industry has become the core factor for whether car companies can stay at the table. At the 2026 World Intelligent Connected Vehicle Congress, the performance of various car companies not only echoed current trends but also conveyed the future direction of competition.
Different from many past conferences, more people dare to speak this time, and the topics discussed are more practical. Six industry leaders argued heatedly over intelligent driving routes; He Xiaopeng and Yang Dongsheng even clashed publicly on "whether to skip L3"; Li Shufu emphasized "repeated construction cannot be done", while Yin Tongyue spoke of "going overseas is rooting", FAW and Dongfeng leaders discussed agglomerating intelligence, and Dongfeng's Feng Changjun also mentioned mutual recognition of standards.
Even though there are viewpoint collisions among many "leaders", almost no one is confined to conventional "involution" anymore. When tags like "banding together", "rooting", "safety", and "trust" appear simultaneously at a World Intelligent Connected Vehicle Congress, it indicates that the competitive strategy of China's automotive industry finally shows signs of moving away from the past "scale involution" driven by prices.

Divergences in technology routes, reconstruction of industrial logic, and elevation of globalization have released signals that the automotive industry is beginning to move towards high-quality competition. And the answers to questions such as "What will allow China's automotive industry to win the second half of the competition?" and "Whether Chinese car companies can undertake a new round of tactics?" have also been reflected in this congress.
He Xiaopeng and Yang Dongsheng "Heated Debate", Fighting for "Who Takes Responsibility"
Discussions about intelligent driving in the industry have always been intense, but few people have taken the contradiction between the rapid development of intelligent driving technology and responsibility bearing to the public stage. One point of this congress worth noting is that at the summit dialogue named "How Global Autonomous Driving Enters the Second Half", XPeng Chairman He Xiaopeng and BYD Senior Vice President Yang Dongsheng, through viewpoint collisions, brought the problems previously existing in "gray areas" directly onto the table.
He Xiaopeng's stance is very clear, he said: "I consider L2 and L4 together, I believe L3 should be skipped." In his view, L3 to some extent transfers responsibility, but user experience improvement is limited, safety levels have not improved, and the best L3 is just an enhanced form of L2. He judges that L4 technology will likely迎来 the scaled landing window in the second half of 2028, and the policy window will open in batches by country, region, and scenario between 2028 and 2032.

Yang Dongsheng's rebuttal left no room, he defined the jump from L2 to L3 as "a huge threshold crossing". In his view, the requirements for safety at L3 are not multiplied, but possibly hundreds of times. At the L2 stage, car companies mainly solve special scenarios like "sudden pedestrian emergence", while L3 must directly face long-tail scenarios such as severe weather, tire blowouts, and uneven/wet roads. In these extreme scenarios, the system must take over. He even admitted that the 200 to 300 million data volume generated by BYD's 4 million vehicles annually "still feels insufficient".
One says "should skip L3", one says "L3 is a threshold that cannot be bypassed". He Xiaopeng believes L3 that "transfers responsibility but not experience" has no value, while Yang Dongsheng believes the responsibility transfer itself is more important. What looks like an intense collision between new forces and traditional car companies on routes, the actual true divergence lies in responsibility definition. Huawei Senior Vice President and Xinwang Company CEO Jin Yuzhi said "When L2 has an accident, the responsible subject is the driver; entering L3, the responsible subject transfers to the system". That sentence is actually more straightforward, directly pointing out this core contradiction.

Other car company leaders also have different views. NIO founder Li Bin separates technology and compliance, emphasizing that mass production of L3 "depends on the pace of legal and regulatory opening". In other words, L3 can exist, but industry regulation needs to provide a "ruler" first. Ultimately, regardless of whether L3 will ultimately be bypassed, "who takes responsibility" has become a problem that cannot be avoided on the path of intelligent driving development.
Even so, "who takes responsibility" is not that easy to answer. After all, this is not something that can be supported by simple sentences. Before this, what can the system use to prove it can bear the responsibility? Why should users dare to give up safety to the system? These two problems need answers first. Momenta CEO Cao Xudong is relatively optimistic about this, he feels the "Intelligent Driving Moore's Law" is accelerating, changing from 10 times improvement every two years to 10 times improvement every year. 2027 to 2028 may become an important inflection point for L3 and L4.

Horizon Robotics founder Yu Kai is more cautious, he believes that the safety takeover interval of urban intelligent driving is currently only at the hundred-kilometer level. To truly let the system bear responsibility and make users dare to use it with confidence, we must move towards the thousand-kilometer even hundred-thousand-kilometer level. Every 10-fold jump is a platform leap. That is to say, for every 10 times more road conditions the system holds, there must be support from technology, data, and verification capabilities behind it. Otherwise, everything is "empty talk". Some see the acceleration of technology, some see the gap within. The two-sidedness of the intelligent driving development path and different attitudes of car companies towards technology development inevitably lead to such divergence.

Of course, after clarifying "who takes responsibility", there are even longer-term things to consider. For car companies, survival is a more realistic issue in front of them. So, business models must be discussed. The value of autonomous driving will ultimately be realized by whether users "dare to use" and "are willing to pay". He Xiaopeng proposed three core indicators: safe collision mileage, full-scenario 10,000 km takeover rate, and user payment rate; Li Bin believes that if users are unwilling to pay, even if the product is done well, it cannot truly create user value.
For this matter, everyone's understanding is actually the same in different words. Ultimately, behind assisted driving is the continuous cost of fulfillment, training, communication, and upgrading. If an enterprise has no revenue source, related services become unsustainable. It is very obvious that the question of "who takes responsibility" runs through the entire business loop of intelligent driving.

Although the future competition of intelligent driving has not yet settled down, some things are already developing in a clearer direction. At this congress, everyone's views on intelligent driving were very clear. More valuable is that it brought the problem of "who bears the responsibility", which was previously covered by the technological halo, onto the surface for discussion. Behind the "heated debate" of many car company executives also refracted the new logic of future competition in this field: Whoever can first establish verifiable safety trust under a responsibility framework can truly stay at the table.
Li Shufu says "No Repeated Construction", Actually Meaning Collaboration
In previous various conferences, everyone preferred to "show muscles" on the surface. If they did not talk about some "black technology", it seemed difficult to support their own decorum. In fact, if we jump out of the separate technology sector and look at the statements of various car companies, we may see more different things. Regarding industrial development, Geely Holding Chairman Li Shufu said: "Cannot do repeated construction, Geely does a set, NIO does a set."

China's automotive industry has developed quite well, but everyone indeed likes to fight separately. In the past, such strategies were not problematic. The beginning of industry development inherently required a process of conquering land for each. But when industry development reaches a certain level, the disadvantages of fighting separately become apparent, which is not conducive to industry growth. Just as Li Shufu said, small and weak will inevitably lead to chaos and poor, only large and strong can lead the industry to better development. This is actually one of the reasons why he advocates activating existing capacity cooperation and intelligent technology collaborative research.
It is not just Li Shufu who thinks this way. FAW General Manager Liu Yigong believes the industry is experiencing a three-fold leap in technology, routes, and boundaries. Agglomerating intelligence "belongs to a systemic engineering of cross-boundary integration, cannot rely on single-handed struggle or closed development." He advocates collaborative research of government, industry, academia, research, and use; cross-boundary collaboration of complete vehicle and parts; and sticking to the safety bottom line. Dongfeng General Manager Feng Changjun's view is that the industry is moving from chain-based division of labor to integrated co-creation. Joint research of key technologies, collaborative sharing of computing power data, and open mutual integration of innovation ecology will become new trends.

Including many high-level executives of state-owned enterprises, all conveying the same signal: the competitive logic of the automotive industry is moving from "single-handed struggle" to "collaborative construction". However, the premise of "co-construction" is indispensable. Xiaomi founder Lei Jun believes that without safety, everything else is zero in the automotive industry. Dongfeng previously jointly released the "Global Safety White Paper for the Era of Embodied Intelligence" with units like CATARC, elevating safety from enterprise responsibility to a common standard language for the industry. Car companies are talking about collaboration, and the regulatory end is tightening synchronously. Everyone is promoting the industry's transition from "fighting separately" to "collaborative construction". It seems that in the second half of China's automotive industry, it may be very difficult to have "lone heroes".
Yin Tongyue's "Overseas Rooting", Actually a Way to Survive
"Going overseas is not harvesting, it is going to take root."
"Resolutely cannot start a price war, otherwise it easily becomes filling the sea."
"Chinese car companies going overseas must be humble and low-key, do not become 'sea blowers' overseas; actually, many aspects of Chinese OEMs overseas still have huge gaps with international traditional car companies."
Chery Chairman Yin Tongyue's words are releasing another signal: China's automotive globalization is moving from "trade output" elevation to "ecological rooting". Li Shufu gave even more specific ideas. He is optimistic about battery swap system going overseas, believing that in the European market with uneven energy supply, poor reliability, and high usage costs, battery swapping is a comprehensive solution combining battery assets + energy management + infrastructure + user services.

From an international dimension, this "rooting" is already in practice. Shen Zhiqin, Deputy Minister of Malaysia Investment, Trade and Industry Department, actively proposed that ASEAN is willing to become an important pole of the global supply chain, welcoming Chinese intelligent connected vehicle enterprises to radiate ASEAN markets with Malaysia as the regional center; German Association of the Automotive Industry Chairman Muxiya emphasized that Chinese-German industrial complementarity is strong, should deepen cooperation through dialogue; International Organization of Motor Vehicle Manufacturers called for maintaining technology fusion under global regulatory frameworks.
It is visible that China's automotive going overseas is no longer one-sided "selling out". Integrating into the local industrial system, supply chain, technology chain, and value chain for two-way co-construction is true long-termism. Just as Yin Tongyue understands, internationalization is a complex systemic engineering. Selling cars is far from enough; one must also integrate into the local industrial system, supply chain, technology chain, and value chain.

At this congress, the "overseas rooting" recognized by Yin Tongyue and other car company executives is not only an overseas path, but also provides a direction for the long-term survival of China's automotive industry overseas.
Conclusion
At the 2026 World Intelligent Connected Vehicle Congress, the development strategies conveyed by various car companies have obviously made very great progress. Some car companies presented the long-term game of intelligent driving responsibility definition "boldly" onto the table. Some car companies used ideas like "no repeated construction", "collaborative co-construction", and the proposal of "going overseas is rooting", all providing references for the survival of car companies in the stock era.
In short, technology, industry, and global these three dimensions remain the core direction of China's automotive true second-half competition. In the stock era, everyone needs to find their own incremental path. The reality is telling participants in the automotive industry: In this era, there are no lone heroes. As to who can stay at the table safely, it depends on who can truly achieve open boundaries, co-build systems, and deeply integrate globally.