Chinese car companies flex their muscles at CES

Author | Zhou Zhiyu

Editor | Zhang Xiaoling

Taking advantage of the CES stage, Chinese car companies are redefining the future.

Compared to previous years' CES (Consumer Electronics Show), this year's participating Chinese car manufacturers and supply chain companies have shown less of the initial excitement and more pragmatism. If two years ago, the Chinese automotive industry was still trying to prove "cost-effectiveness" to the world through battery costs and supply chain efficiency, then at this year's exhibition, the core topic has completely shifted.

From the signals released on-site, the simple accumulation of hardware has reached the critical point of diminishing marginal returns.

When the number of screens and acceleration from 0 to 100 km/h can no longer stimulate the nerves of the capital market, Chinese car companies have begun to collectively charge towards the top of the industry chain—AI large models and advanced intelligent driving architecture. At this CES, Geely launched its "Full-Field AI 2.0" system, Great Wall Motors brought intelligence and power, and the performance breakthroughs of supply chain companies in chips and sensors together formed the Chinese narrative of this CES.

These core players from China are building new technological moats through advanced intelligent driving and AI large models. This will also be a thrilling leap for Chinese players from manufacturing output to standard output in the process of restructuring the automotive industry.

In Las Vegas, a silent battle for the distribution of value in the automotive industry has quietly begun. As hardware homogenization becomes an industry consensus, Chinese car companies are starting to make saturated investments in "soft power." This is not only an arms race at the product level but also a deep restructuring of supply chain integration capabilities and globalization strategies.

Technological Swordplay

At CES, it can be seen that Chinese car companies and supply chain enterprises are forming a "saturated attack" posture, with not only strategic pressure from leading giants but also breakthroughs from invisible champions in niche fields.

Geely Holding Group's actions at this CES reflect its intention for deep integration of R&D resources. Geely did not simply launch new cars but introduced a technological foundation of Full-Field AI 2.0, with its core being WAM (World Action Model).

Unlike end-to-end models, Geely's WAM logically goes a step further. Li Chuanhai, Vice President of Geely Automobile Group and Dean of the Research Institute, pointed out that the industry is currently largely unified in the direction of end-to-end large models, including VLA (Vision-Language-Action). However, WAM adopts a layered design, enabling a complete intelligent closed loop from "understanding-planning" to "rehearsing-judging-correcting." He emphasized that this gives automotive intelligence a "worldview" and "judgment" that can continuously evolve for the first time.

As the "One Geely" strategy continues to advance, technological innovation can be more efficiently applied to the product end. Geely Automobile Group CEO Gan Jiayue revealed that through this round of integration, the number of projects in product planning has actually decreased by 20%-30%, while quality has been enhanced.

This logic of "doing subtraction" is aimed at solving the internal friction issues under a multi-brand strategy. Gan Jiayue clearly stated that although brands like ZEEKR, Lynk & Co, and Geely Galaxy have different market positions, they have unified their intelligent driving software under "Qianli Haohan G-ASD." On the hardware level, while retaining different tier versions from H1 to H9, they have achieved unification at the software core This means that new models such as ZEEKR 9X and Galaxy M9 will share the same AI brain.

"Some good technologies can empower more products. For example, the Shield Gold Brick battery currently used by ZEEKR can also be applied to products from Lynk & Co, Galaxy, and other brands in the future," pointed out Gan Jiayue. This technology feedback mechanism allows models like Lynk & Co 900 to be the first to equip the NVIDIA Thor-U chip, achieving rapid penetration of high-end technology into the mass market within the group.

In addition to showcasing technology, Geely also brought the latest models such as ZEEKR 9X, Galaxy M9, and technological achievements like the Shield Gold Brick battery. Geely revealed that G-ASD will also push for high-speed L3 and low-speed L4 functions this year under regulatory conditions, and achieve Robotaxi operations.

Great Wall Motors chose a dual-line breakthrough path of "mechanics + intelligence." Great Wall released the Hi4-Z vertical dual-motor hybrid architecture, relying on its self-developed strength across the entire industry chain. This technology directly addresses the common concerns of global users regarding range anxiety and energy consumption.

Additionally, Great Wall showcased the Soul S2000 horizontally opposed eight-cylinder motorcycle, which is the world's first horizontally opposed eight-cylinder 2000CC engine, representing the pinnacle of global motorcycle technology. It is not only a display of mechanical aesthetics but also a silent declaration: even in the field of internal combustion engines, which has been dominated by the West for a century, Chinese car companies can possess top-notch craftsmanship.

On the other hand, Great Wall launched the ASL 2.0 (Agent of Space & Language) intelligent system. This system is based on Great Wall's self-developed Coffee EEA 4.0 architecture. This "central computing + regional control" super nervous system endows the entire vehicle with a clearly defined "brain" and "small brain," completely breaking down the information barriers between the cabin, intelligent driving, and power.

Driven by AI OS, ASL 2.0 has evolved two core intelligent agents: the "Smart Space Intelligent Agent" is no longer a simple executor of commands but a "new family member" that provides proactive services through the SOA architecture; the "Intelligent Driving Intelligent Agent" combines VLA and world models, possessing human-like causal reasoning abilities. This cognitive logic, akin to that of an "experienced driver," allows machines to truly understand the physical world with both intelligence and emotional intelligence.

From the initial concept proposed at CES 2025 to the realization of version 2.0 at CES 2026, Great Wall has demonstrated its commitment to investment in cutting-edge technology at a fast pace.

Beyond vehicle manufacturers, Chinese supply chains and new force companies showcased their specific mass production capabilities at this year's CES, directly competing with international giants.

Leapmotor showcased its aggressive strategy in electronic and electrical architecture. The newly released Leapmotor D19 directly equips the Qualcomm dual 8797 Supreme Edition chip, aiming to challenge the performance limits of central computing. Unlike Sony Honda's Afeela, which focuses on bringing the PS5 gaming experience into cars, Leapmotor emphasizes the hard power of "computing foundation" by directing computing power towards vehicle control and real-time reasoning.

Black Sesame Intelligence exhibited the A2000 autonomous driving chip, with a single computing power of 250 TOPS. Its aim is to directly enter the L2+/L3 level intelligent driving market Hesai Technology and RoboSense, two lidar manufacturers, showcased their next-generation products on-site. In addition to improvements in detection distance and accuracy, the core competitiveness lies in the cost control capability at the "thousand-yuan level."

Continuous Iteration

Through CES 2026, the evolution path of the automotive industry over the next three years can be clearly outlined. The rising technical thresholds and changes in business models are accelerating the survival of the fittest in the industry.

In the past five years, the threshold for car manufacturing was capital and qualifications; in the next five years, the threshold will shift to "model inclusion" (the proportion of AI models in vehicle control).

High-level intelligent driving has become an extremely expensive "data game." This is not just an algorithm issue, but also a problem of data scale and computing power costs.

Gan Jiayue believes that current assisted driving is not simply about rules or algorithms; future maintenance will require significant costs. The enhancement of AI capabilities also requires substantial investment in funds, computing power, models, data, and high-quality talent, including accident investigation capabilities, which all necessitate a large-scale team. It may not be very economical for a single company to achieve "full-stack self-research."

Geely is integrating internal R&D resources to build a unified G-ASD foundation, essentially spreading the high AI R&D costs through economies of scale. Gan Jiayue's statement of "reducing quantity, increasing quality" is an interpretation of this logic. Only companies with annual sales reaching millions and possessing massive data feedback capabilities can support the training costs of WAM-level large models.

This also confirms the current industry judgment on future patterns: the intelligent driving market will be dominated by a few giants.

For small and medium-sized car companies that cannot self-develop large models or cannot afford the high costs of computing chips, CES 2026 is a dangerous signal. In the transition from "software-defined vehicles" to "AI-defined vehicles," technological gaps will lead to rapid loss of market share.

From 2026 to 2027, the industry may usher in a new wave of mergers and acquisitions, and those companies that cannot achieve "cross-domain integration" like Geely will be forced to exit.

Moreover, the implementation of Great Wall's ASL 2.0 and Geely's Eva intelligent body signifies a fundamental shift in the commercial attributes of automobiles.

Gan Jiayue points out that future cars will be robots. Whether it is robots or assisted driving, it depends on underlying capabilities, namely computing power, model capabilities, and technical data capabilities. All three aspects are indispensable.

Cars are being redefined as "embodied intelligent terminals." When vehicles possess the ability to perceive the environment, understand intentions, and even make autonomous decisions, their value anchor will shift from "horsepower, genuine leather, space" to "computing power, services, experience."

This means that the valuation logic of car companies will change: hardware profit margins will gradually decline, while revenue from AI-based full lifecycle services (such as intelligent driving subscriptions and spatial services) will become a new growth pole.

This model is disruptive to the traditional supply chain. Automakers will reclaim more software control, while suppliers need to shift from merely selling parts to selling capabilities or ecosystems.

Two years ago, Chinese car companies relied on cost advantages to go global. However, at this year's CES, whether it is Geely daring to propose a WAM model "stronger than VLA" or Great Wall's definition of "spatial intelligence," it shows that Chinese car companies are attempting to export technical standards The technological confidence displayed by Chinese car companies at CES points to the next stage of the global automotive discourse power struggle.

Chinese car companies are no longer mere followers closely watching the movements of European and American giants, but challengers who dare to propose different technological routes and define the standards for "spatial intelligence."

In this game, Chinese enterprises no longer just need to prove "I can make good cars," but are beginning to confidently tell the world "what a good car of the future should look like."