Passenger Car Electronic and Electrical Architecture (EEA) and Technology Supply Chain Construction Strategy Research Report, 2026-2027
Passenger Car EEA: The Next-Generation AI EEA Is Gaining Momentum, Enabling Vehicles to Achieve "Autonomous Perception - Autonomous Decision-Making - Autonomous Evolution"
Vehicle Architecture Is Evolving Toward the AIDV Framework, and the Next-Generation AI EEA Is Entering Mass Production
Currently, vehicle architectures are accelerating toward the new AI-Defined Vehicle (AIDV) framework. As a new generation of intelligent vehicles based on electrification, supported by connectivity and centered on AI, AIEV allows for "autonomous perception - autonomous decision-making - autonomous evolution" by vehicles, through the deep integration of technologies such as intelligent driving, intelligent cockpit and intelligent chassis, and hardware-software coordination.
Against this trend, the new-generation vehicle framework relies on high-compute fusion chips and central computing architecture, and leverages application-layer software, AI basic software and vehicle AI framework to integrate and upgrade cockpit, intelligent driving and chassis (cockpit-driving-control) functions. Meanwhile, the data closed loop drives continuous model evolution. Through gradual integration, a full-scenario intelligent vehicle with multi-agent collaboration, external vehicle connectivity and emotional interaction capabilities is being built.
The next-generation AI EEA is no longer partitioned around functional modules. Instead, it uses AI as the scheduling hub so that vehicle perception, decision-making and execution collaborate within a unified framework, reconstructing vehicles into a system-level foundation of "central computing platform + zone controllers + high-speed communication backbone + SOA/AI OS + safety redundancy + vehicle-cloud collaboration". Its key characteristics are as follows:
Computing: high computing power, heterogeneous SoCs, cockpit-driving-control integrated computing center, AI supercomputing architecture. Leading OEMs have begun to build an EEA system of "one central computing platform + 2-3 zone controllers," and are gradually embedding AI applications on the basis of the central brain and neural network. The AI supercomputing architecture acquires real-world information via sensors such as cameras and lidars, combines AI large models to understand physical laws and make decisions, outputs actions through actuators, and continuously corrects errors and evolves on its own in the feedback loop;
Communication: automotive TSN/AVB Ethernet, 10G/25G, fiber/optical PON, PCIe optical communication, 10M automotive Ethernet/CAN-XL. By 2030, the backbone is expected to be 10G automotive Ethernet paired with CAN XL and 10M automotive Ethernet; by 2035, 25G automotive Ethernet will serve as the backbone to support vehicle-infrastructure-cloud-network-map collaboration;
Software: SOA/MSOA, Hypervisor, Adaptive AUTOSAR, containerization. The software architecture is layered under SOA into system software, functional software, application software and cloud platform. The Hypervisor handles multi-system isolation, while middleware extends from Classic AUTOSAR and Adaptive AUTOSAR to ROS2, CyberRT and Iceoryx;
Safety: fail-operational, ASIL-D, SOTIF, cybersecurity and redundancy. L4 scenarios require the system to do more than merely achieve "safe stop after failure" - it must remain available after partial failures. Functional safety needs to move from fail-safe to fail-operational, introducing communication redundancy, redundant controllers and end-to-end (E2E) protection;
Vehicle-cloud collaboration: from vehicle intelligence toward vehicle-infrastructure-cloud integration. The edge cloud sinks to the 5G access network side to undertake intersection-level real-time control; the regional cloud hosts weakly real-time road-network-level applications, while the central cloud is responsible for global data management and optimization.
In 2026, passenger car EEA in China has generally entered the central integration stage. Domestic OEMs such as Geely Auto, XPeng, Leapmotor and SAIC Motor have begun to lay out the next-generation AI EEA, whose core features include central integration, AI native, cockpit-driving integration, high-compute redundancy, vehicle-cloud collaboration and software definition.
Geely GEEA 4.0: a universal architecture for both ICE and electric vehicles, with reserved interfaces for the central computing platform and the 800V high-voltage system. Its underlying integration level is higher than that of GEEA 3.0. GEEA 3.0 already integrates 14 domain controllers into two central computing units, and gigabit Ethernet increases signal transmission speed by 200 times. GEEA 4.0 evolves further on this basis.
Compute and chips: it features the world's first "quantum-level AI EEA - EEA 4.0," integrating three flagship chips - NVIDIA SuperChip, Thor U and Qualcomm Snapdragon 8797 - with total computing power exceeding 3,000 TOPS, reserving ample compute for L4 autonomous driving and significantly improving model inference efficiency;
AI models and perception: World Action Model (WAM) + G-ASD 4.0 algorithm model + 2,160-channel lidar. Based on the ultra-high-compute platform of GEEA 4.0, the parameter count of vehicle WAM increases 7 times and the model inference frame rate increases 3 times, thereby strengthening environmental perception, behavior prediction and decision-making and reasoning capabilities;
Safety and vehicle-cloud: quantum encryption, SOVD and full-lifecycle protection. Through quantum encryption, GEEA 4.0 achieves full-link, quantum-level security protection on both the vehicle and cloud ends, and features the industry's first SOVD vehicle-cloud integrated diagnosis technology;
AI all-terrain: the AI all-terrain digital chassis is based on GEEA 4.0 and adopts GVMC 3.0, supporting 9+6+X multi-terrain modes. It can identify upcoming road conditions, automatically switch modes, and adjust torque output, suspension status and body posture in real time.
Leapmotor LEAP 5.0: under the central domain control architecture, other controllers serve only as execution units; the vehicle harness is shortened to less than 500 meters, vehicle-wide Ethernet reaches up to 10 Gigabit, and the full-domain 48V power supply supports more high-power functions.
Vehicle-cloud collaborative AgentOS: core agents and safety skills are deployed on the vehicle end for fast response and privacy protection; core vehicle control, navigation and entertainment remain available under weak network or even offline conditions. The cloud handles complex reasoning, long-term memory and skill orchestration, and iterates continuously via lightweight updates.
Vehicle-wide Ethernet: the central brain and zone controllers form a backbone of "big loops enclosing small loops," and edge devices connect locally through low-speed Ethernet, eliminating protocol conversion gateways. The backbone adopts an asymmetric design - 10 Gigabit downlink and 100 Megabit uplink - with a single cable transmitting both video and control data, and supports hardware-level security authentication.
Full-domain 48V power distribution: the traction battery supplies 48V directly to each zone controller, which distributes power locally and uses eFuse for protection; current is reduced by 75% and heat loss by 93%, and the D19 power harness is expected to be 50% lighter. Combined, these three measures further cut vehicle harness to less than 500 meters.
AI intelligent chassis: undertakes the core conversion task of translating environmental perception information into vehicle dynamic control commands;
MM-i multi-mode hybrid electric drive: continuously optimizes power output characteristics, transmission structure and energy conversion efficiency;
Physical space and integration: the LEAP 5.0 architecture expands available cabin space along the X, Y and Z axes, solving the problem of mechanical layouts encroaching on the passenger cabin in the traditional ICE era. Leapmotor borrowed a concept from real estate - a usable space ratio of over 100%. At its core are two integrations: the highly integrated all-in-one air conditioning unit and the rear-mounted powertrain. Leapmotor highly integrates and front-mounts the previously dispersed and complex air conditioning system in the front compartment, reducing component count by 18%, longitudinal space occupation by 34% and piping by 30%, and lowering operating noise by 3 dB.
Intelligent Chassis: Move Toward Drive-Brake-Suspension-Steer Integration, and the Chip-Level Drive-Brake Integration Solution Has Entered Mass Production for GAC Aion Ray7
Based on the central integrated architecture, high-level intelligent driving functions are increasingly sensitive to response time. The latency of traditional CAN communication can no longer support high-frequency, high-precision integrated posture control, so vehicle manufacturing has posed a range of requirements for intelligent chassis, such as redundancy and by-wire control. Meanwhile, vehicle motion is not independent movement of a single axle but the result of multi-axle coordination, requiring a unified control strategy and efficient compute support.
Currently, drive-brake-suspension-steer integration has become the industry consensus: the four traditionally independent chassis systems - drive, brake, suspension and steering - are converted from mechanical connections to electrical signal control (by-wire), and coordinated and scheduled by a unified chassis domain controller (DCU/ICC/VMC) to achieve cross-system fusion control across the X axis (longitudinal), Y axis (lateral) and Z axis (vertical). A number of OEMs have already launched related products.
Chery's Feiyu Intelligent Driving Unit packs four components - electro-mechanical brake (EMB), in-wheel motor, four-wheel independent kingpin steering and active suspension - into a single wheel, coordinated by the chassis domain controller. It enables complete decoupling of the motion of four wheels, each outputting a different torque, steering angle and braking force.
Maximum wheel-end steering angle reaches 90°, supporting lateral driving, tire-scuff-free U-turn (center deviation ≤ 0.1 m) and 5m lateral movement, allowing side parking to be completed in one maneuver;
Zero body roll during high-speed oblique driving at 120 km/h; moose-test entry speed ≥ 100 km/h; steady-state cornering roll gradient ≤3.0%;
In terms of safety, brake, steer and drive each feature independent three-channel redundancy; combined braking deceleration is ≥ 1.2 g, and safe driving is maintained after a single-point failure.
Chery's four-wheel independent integration solution thoroughly decouples the motion of the four wheels: the chassis has evolved from a "mechanical system" into a fully programmable "electronic system," which can in principle achieve any driving posture, while placing greater demands on sealing, heat dissipation and reliability. In addition, the current cost of the complete system is far higher than that of a traditional chassis, and scaling up to reduce costs will take time. Chery’s Feiyu Intelligent Driving Unit made its first public outdoor appearance at the 2026 Chery Global Innovation Conference, completing dynamic demonstrations in four scenarios: V-shaped road, concentric-circle driving, extreme-speed parking and single-wheel escape. It will later be applied to production models of the Sterra series.
The chip-level drive-brake integration architecture of the GAC Aion Ray7 goes one level deeper: drive and brake control are merged into a single-chip domain for coordinated decision-making. Together with the HUAWEI DriveONE EMB, it can brake safely to a stop even if all four wheels fail. Official test results show a stopping distance of 33m from 100 km/h.
In the traditional distributed chassis control architecture, the drive motor is managed by the motor control unit (MCU), while braking is managed by the ESP/ABS electro-hydraulic pump. The two control domains operate independently and exchange commands via the vehicle network CAN/LIN. Later drive-brake integration solutions began to use one domain controller to coordinate both electric drive torque and hydraulic brake torque, which reduced coordination latency between the two sets of actuators, yet communication between the drive motor and the brake master cylinder still had to traverse multiple controller nodes. The chip-level drive-brake integration architecture achieves three-domain fusion of "electric drive + braking + chassis control," but it also raises development difficulty and safety requirements. The real performance of the GAC Aion Ray7 remains to be seen.
T-BOX: It May Exist as a Plug-in Card in the Future
Currently, within the vehicle EEA, the three software functions of the T-BOX - OTA management, remote diagnosis and data upload to the cloud, and gateway protocol conversion - have been taken over by the cockpit-driving integrated platform. However, RF hardware, eCall and safety backup MCU, HSM and key security are difficult to fuse and must remain in place long term.
At present, the standalone T-BOX remains the mainstream form, while integrated solutions are gradually becoming dominant, for example, integration into smart antennas, cockpit domain controller or communication domain controller. The Xiaomi YU7 integrates the T-Box communication module, vehicle domain controller (VCCD), cockpit domain controller (DCD) and assisted-driving domain controller into a central computing platform through four-in-one integration technology, packaged onto a 3.6 kg motherboard: volume reduced by 57%, components reduced by 75%, OTA completed in 15 minutes, sentry-mode power consumption reduced by 40%. The vehicle can therefore travel an additional 16km.
In the future, the T-BOX may be split apart, but its functions will remain: it will evolve from a standalone ECU into a communication-security plug-in card of the central platform. Bosch, Continental, Desay SV, Huawei and Quectel among others have laid out early, shifting from selling T-BOX to selling "communication platform software + hardware modules" - the box form will disappear, but the commercial value of communication capability, security capability and cloud-platform interfacing will not, and it will simply be encapsulated in thinner modules and a deeper software stack.
Passenger Car Electronic and Electrical Architecture (EEA) and Technology Supply Chain Construction Strategy Research Report, 2026-2027
Passenger Car EEA: The Next-Generation AI EEA Is Gaining Momentum, Enabling Vehicles to Achieve "Autonomous Perception - Autonomous Decision-Making - Autonomous Evolution"
Vehicle Architecture Is Ev...
Next-generation Central and Zonal Communication Network Topology and Chip Industry Research Report, 2026-2027
Research on Zonal Communication Topology: New Architectures Including 10 BASE-T1S+RCP, AVB/TSN Ethernet Are Under Rapid Development
Vehicle communication networks are mainly divided into in-vehicle n...
Software-Defined Vehicles in 2026: Software Supply Chain Panorama and Business Models Research
Research on Automotive Software Business Models: Exploration of Token Economy, AI Service Revenue Sharing and Other Businesses
Under the Trend of AI Agents: Exploration of Business Models Based on T...
New Energy Commercial Vehicle 800-1000V High-Voltage Platform & Supply Chain Panorama Research Report, 2026
Research on 800V Architecture of Commercial Vehicles: New Energy Penetration Rate Exceeds 35%, and 800V Full-Domain High-Voltage Architecture Rapidly Penetrates Heavy-Duty Trucks
Explosive Growth in ...
ADAS and Autonomous Driving Tier 1 Suppliers Research Report, 2026 – Chinese Companies
Intelligent Driving Industry Leap: From Widespread L2.9 Adoption toward Physical AI and New Full?Domain Mobility Ecosystem
As vehicle intelligence advances, the installations and penetration rate of...
Vehicle-Road-Cloud Integration and C-V2X Industry Report, 2026
Vehicle-Road-Cloud and C-V2X Research: with the three-tier promotion system of "top-level planning + mandatory pilot + evaluation binding", the industry turning point is approachingThis report focuses...
China OEM Data Capability Analysis Report, 2026
OEM Data Capability Research: 24 OEMs × 9 Capability Domains, Who Takes the Lead?
An L4 autonomous driving R&D and test vehicle can generate approximately 10TB of raw data per single day — data v...
Autonomous Driving Domain Controller and Central Computing Unit (CCU) Industry Report, 2026
Autonomous Driving Domain Controller (ADCU) Research: One-Chip CCU solutions are quickly available in vehicles
In China's passenger car market, autonomous driving systems involved in mass production ...
Automotive EMS and ECU Industry Research Report, 2026
Automotive EMS Research: The boundaries between EMS providers and Tier1 suppliers are becoming blurred, and the supply chain model is evolving from a linear chain to a networked structure
Mainstream ...
Automotive Microcontroller Unit (MCU) and Application Scenario Research Report, 2026
Automotive MCU Research: Increased demand for redundant security drives ASIL D-compliant MCU shipments to exceed 100 million units
In a vehicle EEA, MCUs are widely distributed in various functional ...
China Intelligent Cockpit Patent Analysis Report, 2026
Intelligent Cockpit Patent Research: Cockpit AI, In?vehicle Health, and Infotainment Are Upgrade to Scenario?based and Proactive Functions
From a patent perspective, this report explores the cutting?...
Research Report on Application of VLA Large Model in Automobiles and Robots, 2026
Research on Automotive and Robot VLA: Hybrid Architectures Become Mainstream, VLA Integrates with General World Models, and Reinforcement Learning Serves as Core Engine
Vision?Language?Action (VLA) m...
China Charging Infrastructure (Supercharging, Battery Swapping, V2G, etc.) and High-Performance Supercharging Battery Research Report, 2026
Charging and battery swapping research: as 5C+ supercharging vehicle models go into mass production, the pace of OEMs self-building supercharging networks quickens
800-1000V high-voltage platforms ar...
China Passenger Car Electronic Controlled Suspension Industry Report, 2026
Electronic Controlled Suspension Industry Research: As the First Year of Full Active Suspension Unfolds, Three Technical Routes Race Forward
1. Penetration Rate Surges Three Times, and Electronic Con...
Intelligent Driving SoC Research Report, 2026
Research on Intelligent Driving SoCs: Competition Focus Shifts to L3/L4 Embedded Hardware, and Computing Power of Flagship Vehicle Models Exceeds 1,000 TOPS
In terms of intelligent driving system in...
Automotive AI Large Model Technology Research Report, 2026
Automotive AI Large Model Research: Competition Shifts from "Who Has the Stronger Model" to "Who Boasts Higher Link Efficiency"
ResearchInChina released the Automotive AI Large Model Technology Resea...
Intelligent Vehicle Cockpit Domain Controller Research Report, 2026
Cockpit domain controller research: L3 AIDV intelligent cockpit domain controllers are entering a boom period
Driven by multiple factors such as the continuous evolution of the automotive central int...
Automotive Acoustic System (Audio, Multi-Channel) Industry Report, 2026
Automotive Acoustics Research: Multiple Channels, AI Tuning, and Self-Developed Algorithms Drive the Transformation of High-End Cockpit Sound Fields
I. Automotive Audio Hardware Solutions with 6–9 S...