A simple phrase like "The child has fallen asleep" allows the system to combine this command with sensor data to coordinate adjustments for air conditioning, volume, lighting, seats, and suspension, eliminating the need for item-by-item operation. This is a cockpit AI agent demonstration frequently seen at new car launch events in 2026.
Automakers are willing to devote significant space to such demonstrations. However, competition in the real market has already moved to another stage.
At the 2026 China Automotive Supply Chain Conference, Yu Zhiyao, Senior Data Analyst at Gasgoo Research Institute, presented a set of data: In August 2026, the penetration rate of smart cockpits in domestic passenger vehicles reached 87.5%, approaching 90%. Basic configurations such as central control screens and voice interaction are nearing saturation in new models.
When nearly 90% of new cars are equipped with large screens and voice interaction, an increasingly long configuration list can no longer easily distinguish superiority. The focus of cockpit competition is shifting from "configuration lists" to "experience reconstruction."
After "Standard Equipment," What Is Being Competed On? Has the Configuration Race Reached Its Limit?
The statistical criteria of Gasgoo Research Institute are specific: A vehicle is considered to have a smart cockpit only if it simultaneously possesses four features: an 8-inch or larger central control screen, voice interaction, Vehicle-to-Everything (V2X) connectivity, and OTA (Over-the-Air) updates. Based on this criterion, the penetration rate of smart cockpits in domestic passenger vehicles has approached 90%, while NEVs (New Energy Vehicles) have largely achieved standard inclusion, becoming the main driver of cockpit intelligence penetration.
As the penetration rate rises, configurations are also rapidly trickling down to lower price segments. Data from Gasgoo Research Institute shows that the cockpit configuration rate for models priced under CNY 100,000 has increased from 52% to 70%. Yu Zhiyao stated that for economy models, adding basic cockpit configurations has a more pronounced marginal effect on final sales volume.
Breaking down the structure, basic configurations are becoming standard, while advanced configurations are still spreading rapidly. In the first eight months of 2026, the penetration rate of central control screens approached 100%, and voice recognition reached 90%. High-level configurations like DMS (Driver Monitoring System) are ramping up quickly; the penetration rate of cockpit domain controllers exceeds 50%, and differentiated configurations such as HUDs (Head-Up Displays) are being installed at a fast pace.
Screens are the most intense area of the configuration race. Larger and multiple screens remain the main trend. The proportion of central control screens over 15 inches has reached 43% and is moving toward 50%, while the share of passenger entertainment screens and multi-screen front-row layouts continues to climb. In the first eight months, rear-seat entertainment screens were installed in 739,000 vehicles, accounting for 6.5% of installations, nearly doubling year-on-year. High-specification screens such as Mini LED and OLED are steadily being adopted and beginning to penetrate into the sub-CNY 200,000 segment.
Yu Zhiyao does not deny the value of large screens. He mentioned in his speech that equipping consumers with a large central control screen provides an immediate improvement in technological feel and user experience, which is what the industry often calls "perceived value." However, the marginal surprise brought by large screens is diminishing. As they shift from niche selling points to mainstream standard equipment, the room to differentiate based on screen count and size is narrowing.
Users' willingness to pay for cockpits remains unchanged. Surveys show that nearly 80% of users consider smart cockpit configurations when purchasing a car, and over 60% are willing to pay for cockpit functions. It is projected that the domestic smart cockpit market size will reach CNY 330 billion by 2030.
What users are willing to pay for is no longer just "whether it exists," but "whether it is worth it."
According to industry classifications, smart cockpits are divided into five levels from CL0 to CL4. Currently, the overall stage is transitioning from CL2 to CL3, meaning a shift from partial cognitive intelligence to higher-level cognitive intelligence, and from passive interaction to active perception. The "Technology Roadmap for Energy-Saving and New Energy Vehicles 3.0" sets a timeline to achieve CL3 by 2027 and move towards CL4 after 2032.
Yu Zhiyao categorizes cockpit experience scenarios into four types: entertainment, home relaxation, office work, and health care. From entertainment-focused scenarios like rear-seat viewing and in-car karaoke, to relaxation-focused ones like zero-gravity seats and in-car refrigerators, to office-focused ones like in-car meetings and screen casting collaboration, and finally to health-care-focused ones like fatigue monitoring and emotion recognition, the cockpit is being redefined as the "Third Living Space."
"In the past, competition emphasized how many configurations and functions existed. Now, the focus is more on what role these configurations and functions play in actual driving scenarios and what experience they bring to users, rather than staying on a cold parameter specification sheet," said Yu Zhiyao.
In his view, what matters more than a longer list is the scenarios where these configurations are applied.
From "Stacking Hardware" to "Prioritizing Scenarios"
The value of configurations must ultimately be realized in specific scenarios, and the evolution of HUDs illustrates this well.
Early HUD competitions focused on image size and brightness. Installation data from Gasgoo Research Institute shows that in the first eight months of 2026, W-HUD (Windshield HUD) still accounted for 56% of standard HUD installations, AR-HUD (Augmented Reality HUD) rose to 40%, and P-HUD (Phantom/Combiner HUD) held 4%. In the NEV market, the share of AR-HUD has reached 46%, just one step away from W-HUD's 48%. At this pace, AR-HUD may surpass W-HUD to become mainstream in the next one or two years.
More critical than structural changes is the migration of competitive focus. Yu Zhiyao analyzed representative models including the AITO M9, NIO ES8, XPeng G7, and Li Auto L9, finding that the core improvements in HUDs are no longer just the parameters emphasized at launch events.
Different manufacturers are pursuing different experience paths. Some focus on fitment; the AITO M9's dual-focal-plane AR-HUD deeply integrates navigation information with the real road. Others prioritize visibility in strong light; the NIO ES8 uses a peak brightness of 20,000 nits to ensure clarity in complex environments. Some compete on low latency and dynamic alignment; the XPeng G7 controls distortion within 1% and reduces latency to approximately 100 milliseconds.
The Li Auto L9 takes a contrarian approach, not pursuing ultra-large fields of view but emphasizing streamlined core information and reduced visual interference. Meanwhile, models like the Leapmotor D19 and Buick Zhijing E7 are pushing large-size AR-HUDs into more mainstream price brackets, with the Buick Zhijing E7 already penetrating below CNY 200,000.
Yu Zhiyao's judgment is that beyond providing basic navigation, how HUDs integrate with assisted driving and respond to users' real needs determines their differentiation. Clear, stable, low-interference presentation of driving information, the alignment of AR graphics with the road surface, and coordination between multiple screens are the watershed moments for experience.
Another aspect evolving from "existence" to "quality" is interaction. Voice remains the core entry point for in-car interaction, and advanced functions like continuous voice commands and "say what you see" have high penetration rates. When "being understood" is no longer rare, human-vehicle interaction is entering a 3.0 phase characterized by multimodal, proactive, emotional, and personalized interactions.
The boundaries of interaction are extending outside the vehicle. In narrow parking spots where boarding and alighting are inconvenient, a single voice command from outside the car can guide the vehicle in and out. External light interactions can project safety warnings to pedestrians and display personalized content. The scope of cockpit interaction is expanding from inside the car to full-scenario coverage including boarding/alighting, parking, and external communication.
Cockpit AI agents push this reconstruction deeper. Yu Zhiyao found after reviewing multiple automakers that deploying cockpit AI is approaching standard status. The true direction of iteration is enabling a simple command to coordinate multiple domains and get things done, rather than focusing on how many rounds of conversation are possible.
The opening scenario of "The child has fallen asleep" is an example: The system combines this statement with sensor data to coordinate adjustments for air conditioning, volume, lighting, seats, and suspension without requiring item-by-item operations. Cockpit AI is moving from Q&A interaction to intent understanding, task planning, and multi-domain collaborative execution.
Experience continuity also requires the cockpit to break the boundaries of a single vehicle. "Human-Vehicle-Home" integration has become a long-term strategic direction for all parties. One path uses smartphones as super entry points; Huawei HarmonyOS and Xiaomi HyperOS leverage mature device ecosystems to connect phones, cars, and home scenarios. Another path involves automakers directly partnering with appliance companies; collaborations between Changan and Haier, NIO and Midea, and BYD and Midea have successively landed, achieving "controlling the home from the car" and "controlling the car from home."
"People's daily lives revolve around smartphones, cars, and homes," said Yu Zhiyao. How cars link with home appliances and smartphones, allowing services to seamlessly continue across different scenarios, is itself a key focus of experience competition.
HUDs, interaction, AI agents, and ecosystems all point in the same direction. The number of functions is no longer the metric; the completeness and continuity of functions in real-world scenarios are what matter.
Who Will Deliver the Experience?
The richer the scenarios and the more proactive the agents, the higher the requirements for the electronic/electrical architecture. Experience reconstruction ultimately falls on the reconstruction of architecture and supply chain roles.
Data from Gasgoo Research Institute shows that installations of central computing platforms are growing exponentially, with penetration exceeding 10%, while cockpit domain controllers and ADAS domain controllers maintain rapid growth. Cockpit chips are shifting from the previous "one chip, multiple screens" model to "one chip, multiple uses." The competitive focus has moved from simply increasing computing power to task isolation, computing power scheduling, software reuse, and system cost reduction.
Currently, there are four main paths for cross-domain fusion in the cockpit domain. Cabin-parking fusion has the highest degree of industrialization and has already been implemented in mainstream price ranges. Cabin-connectivity fusion links communication modules with the cockpit, representing a key layout for communication module manufacturers. Cabin-driving fusion has numerous participating companies, and single-chip cabin-driving integration has already been deployed in vehicles. Looking further ahead is multi-domain central computing integration, which is currently still in the platform development and early introduction phase. "Central Computing Platform plus Zone Controllers" is becoming the common direction for OEMs' electronic/electrical architectures.
The core of the cross-domain platform battle lies in the underlying computing foundation, requiring consideration of both computing power and capabilities in security isolation, software ecosystem, and system integration. Qualcomm's Ride Flex platform was mass-produced and installed last year, and NVIDIA's DRIVE Thor platform is also advancing. Among local players, Horizon Robotics' "Starry Sky" adopts a 5nm process, supporting cockpit AI and high-level intelligent driving on a single chip. It has secured cooperation intentions from over ten brands and plans mass production on models like Chery iCAR. Suppliers such as Desay SV and CheLian TianXia have also implemented single-chip cabin-driving integrated solutions in vehicles.
However, in the most critical cockpit domain controller chips, the landscape has not fundamentally changed. Qualcomm still holds over 70% market share, while local enterprises represented by Huawei, SiEngine, and SemiDrive are rising rapidly, with their share trending upward.
Architectural fusion is also changing supplier roles. Yu Zhiyao judges that under the trend of AI-defined vehicles, Tier 1 suppliers are transforming from traditional vendors to technology partners.
In the past, suppliers primarily delivered hardware and software products according to requirement specifications. Now, cooperation is deeper, with suppliers involved in the product definition stage to jointly assess scenario requirements, computing power configurations, and technical feasibility. The deliverables have also changed, expanding from domain controllers and display modules to computing platforms, foundational software, AI deployment tools, and system integration capabilities.
The logic of cooperation has shifted accordingly. Development methods are moving from model-specific customization to platform-based and modular approaches, using standard interfaces to support reuse across multiple models. Metrics for evaluating suppliers have expanded from unit price, performance, and delivery time to task completion rate, software reuse rate, iteration efficiency, and total system cost.
The boundaries of division of labor are becoming clearer. Automakers lead brand experience, function definition, and whole-vehicle acceptance, holding data permissions and vehicle control boundaries. Tier 1 suppliers oversee platform adaptation, AI engineering, system validation, and mass production maintenance, connecting chips, models, and service ecosystems. Chip and AI companies provide computing platforms, models, and tools.
Responsibility cycles and revenue structures are also changing. AI functions are continuously updated via OTA, so work continues after mass production delivery. Suppliers need to persistently participate in model updates, problem diagnosis, and experience optimization. Revenue structures are expanding from one-time hardware supply to combinations of hardware supply, software licensing or platform adaptation, and continuous development and maintenance.
"For suppliers, profitability largely depends on whether the platform can be reused and whether reliable mass production delivery capabilities exist," said Yu Zhiyao. Platformization capabilities can dilute R&D costs, while the ability to continuously improve experience and optimize costs after mass production will determine how far the partnership can go.
Returning to the initial question: What is competed on after "standard equipment"?
Whether a HUD aligns with the road, whether a single command truly gets the job done, and whether services continue seamlessly between car and home—these are the experiences users genuinely care about. Supporting these experiences are cross-domain architectures, platform reuse capabilities, and continuous operations throughout the vehicle's lifecycle.
Configurations can be leveled out in a short time, but capabilities in platform reuse, mass production delivery, and continuous operations cannot be rushed. When configuration lists lose their distinguishing power, these "slow variables" are what truly matter to compete on after "standard equipment."