ZoZoGo and Shuiqingmuhua Research Center jointly released the "2026 Intelligent Vehicle Cockpit Domain Controller Research Report".
Driven by multiple factors such as the continuous evolution of the centralized automotive EEA and the rapid deployment of in-vehicle AI large models, intelligent cockpit domain controllers are evolving from single-SoC cockpit domains toward AI cockpit domain controllers and integrated cockpit-driving platforms. The essence is a clear mainline of "centralized computing power, functional integration, and software-hardware decoupling." Based on form and timeline, the development of cockpit domain controllers can be summarized into four major stages:
L1, Perceptual Cockpit Domain Controller: CPU < 100k, equipped with certain AI computing power, DRAM interface LPDDR3/4, memory bandwidth < 30Gb/s, capable of supporting < 3 screens with a single chip, supporting < 2 camera inputs, featuring networking and OTA capabilities, basic in-cabin perception capabilities (such as voice), and cloud-based voice functions.
L2, Partial Cognitive Cockpit Domain Controller: CPU > 100k, AI > 10 TOPS, DRAM LPDDR4/4X/5, memory bandwidth 30-70Gb/s, capable of supporting 3-5+ screens with a single chip, up to 2K resolution, supporting 2+ in-cabin camera inputs, supporting 100-200 million parameter edge-side large models, local voice command words + cloud semantic capabilities, with further enhanced cognitive capabilities inside and outside the cabin.
L3, Advanced Cognitive Cockpit Domain Controller: CPU > 200k, AI > 30 TOPS, GPU > 3 TFLOPS, DRAM LPDDR5/5X, memory bandwidth exceeding 70Gb/s, including cockpit AI SoC (System on Chip) and advanced integrated cockpit-driving AI SoC, capable of supporting 6-8 screens with a single chip, 2K / 4K resolution, supporting asynchronous display and touch, supporting 4+ in-cabin camera inputs, capable of supporting 5-15 billion parameter edge-side large models, large model-loaded voice functions, and full-scenario active perception.
L4, Comprehensive Cognitive Cockpit Domain Controller: CPU > 500k, AI > 200 TOPS, DRAM LPDDR6, GDDR7, HBM, etc., memory bandwidth exceeding 300Gb/s - 1000Gb/s, capable of supporting 8-12 screens with a single chip, 8+ camera inputs, and supporting 30+ billion parameter edge-side large models.
L3 AIDV Intelligent Cockpit is Rapidly Scaling Up, and High-Performance SoC Cockpit Domain Controllers Facilitate the Integration of Edge-Side AI Large Models into Vehicles
In 2025, the installation volume of intelligent cockpit domain controllers in Chinese passenger vehicles reached 12.974 million units, with the market penetration rate increasing to 56.6%. From January to April 2026, the penetration rate of intelligent cockpit domain controllers further climbed to 63.2%. As the E/E architecture accelerates its iteration and automotive AI intelligence rapidly advances, the penetration rate of intelligent cockpit domain controllers in Chinese passenger vehicles is expected to reach 98.3% by 2030, with the installation volume hitting 24.83 million units.
In 2025, L3 cockpit domain controllers officially entered the stage of scaled mass production, with the installation volume reaching 221,000 units. From January to April 2026, the installation volume rapidly grew to 226,000 units, and is expected to exceed one million units for the whole year. This also marks the official entry of the intelligent cockpit into the AIDV stage. As a critical node connecting the past and the future, the L3 cockpit domain controller will become the core battleground for industry competition and technological innovation in the coming years.
Under the wave of AI-defined vehicle technology, L3 cockpit domain solutions represented by SemiDrive X10, Qualcomm 8397, and MediaTek MT8676 have been deployed and landed. Relying on powerful edge-side AI computing power and breakthroughs in technical architecture, with the operation of edge-side AI large models at the core, they successfully support the efficient inference of physical world models and the real-time operation of in-cabin multimodal AI large models. This enables the cockpit to truly possess the closed-loop intelligent capabilities of "perception-understanding-decision-execution," driving the leapfrog evolution of L3 intelligent cockpits from "functional integration" to "AI-native."
In 2026, SemiDrive's latest upgraded X10, based on TSMC (Taiwan Semiconductor Manufacturing Company)'s 4nm automotive-grade process and designed on the ARMv9 architecture, achieves a CPU performance of 250K DMIPS and a GPU performance of 3000 GFLOPS. The VPU supports 8K resolution encoding and decoding, and can simultaneously drive 12 cameras and 8 displays, meeting the full-scenario needs of mainstream high-end cockpits. Meanwhile, the X10 also integrates four HiFi-5s Audio DSPs, supporting a 16-microphone and 32-speaker array without the need for an external audio chip, delivering a multi-zone, immersive, and ultimate audio experience.
In terms of software ecosystem, the X10 adopts a dual-system architecture of QNX 8.0 Hypervisor and Android 16, balancing functional safety and an open ecosystem. It also supports customers in conducting prototype development on the existing X9 platform based on the same software baseline, and then quickly and seamlessly migrating to the X10, which can significantly shorten the model deployment and performance tuning cycles.
SemiDrive X10 AI Cockpit Platform
Source: SemiDrive
The core breakthrough of the X10 lies not only in the improvement of parameters but also in its brand-new technical architecture specifically designed for the AI cockpit, which thoroughly reconstructs the computing power allocation and system logic of traditional cockpits. Through the single-chip AI cockpit solution, SemiDrive X10, with 80 TOPS NPU computing power paired with 154GB/s DDR bandwidth, can support the local deployment of large models up to 9B in size, flexibly adapting to mainstream edge-side large models such as Mianbi Intelligence / Qwen.
In addition, through the optimization of process and architecture, the X10 can achieve a 2x improvement in energy efficiency without the need for water cooling. Compared to the traditional "cockpit chip + AI Box" solution, the overall system cost can be saved by 1,500 to 3,000 RMB per vehicle, helping the AI cockpit accelerate its expansion from high-end flagship exclusivity to the mainstream consumer market, and promoting the democratization of AI technology.
SemiDrive X10 Architecture Design for Edge-Side Large Models
Source: SemiDrive
Driven by AI-defined vehicles and the growing demand for AI cockpit platforms, multiple mainstream suppliers such as Desay SV, AutoLink, PATEO, Hangsheng Electronics, Meijia Technology, Visteon, and Bosch are also actively deploying L3 cockpit domain controller products to meet the rapidly growing market demand.
In August 2026, the Freelander 8 was officially launched. This model is the first mass-produced vehicle under FREELANDER, an independent NEV (New Energy Vehicle) brand jointly created by Chery and Jaguar Land Rover. It is the world's first mass-produced model equipped with a cockpit domain controller powered by the Qualcomm Snapdragon 8397 chip. Not only is the main control SoC at the industry-leading level, but its auxiliary chips and core materials can also be described as a "luxurious configuration."
In terms of MCU (Microcontroller Unit), the Freelander 8 is equipped with the TC4D9XP chip from Infineon's TC4X series. The TriCore architecture of this chip has been upgraded from v1.6.2 to v1.8, with the clock frequency increased from 300MHz to 500MHz. It supports up to 6 pairs of lockstep cores running simultaneously, and its computing power has approached that of low-end SoCs, making it one of the most advanced MCU products in the industry today.
In terms of storage, the Qualcomm QAM8397P module is equipped with LPDDR5X memory, with a capacity between 24GB and 36GB, and a speed of up to 8533Mbps, which can efficiently support the real-time operation of edge-side AI models and achieve millisecond-level response. In terms of flash memory configuration, the Freelander 8 adopts the industry-leading UFS 4.0 standard from Kioxia. Compared to the Micron SH023/SH021 (mostly UFS 3.1 standard) commonly used by other manufacturers, the single-channel bandwidth of UFS 4.0 can reach up to 23.2Gbps (about 2.9GB/s). The actual measured continuous read speed is about 4200MB/s, and the continuous write speed is between 2800MB/s and 4000MB/s. The overall read and write performance is twice that of the previous generation UFS 3.1, and the power efficiency is improved by about 45%. The UFS model equipped in the Freelander 8 is THGJFJT0T25BAB8, with a capacity of 128GB and a maximum speed of up to 4640MB/s.
Under the Development Trend of AIDV, the Mass Production and Integration of Integrated Cockpit-Driving (Including Multi-Chip or Single-Chip Solutions) Accelerates
As one of the core forms in the evolution of cockpit domain controllers, the integrated cockpit-driving central computing platform has seen its installation rate in Chinese passenger vehicles rise from 3.1% in 2023 to 7.3% in 2025. Driven by the evolution of the EEA and the integration trend, the integrated cockpit-driving central computing platform will continue to grow rapidly, and its penetration rate is expected to reach over 30% by 2030.
In terms of installation volume, the multi-chip integrated cockpit-driving domain controller is the current mainstream form, accounting for 97.4% of the total installation volume of integrated cockpit-driving domain controllers from January to April 2026. Within the multi-chip integrated cockpit-driving, various solution combinations are presented, including dual-chip (1* AD SoC + 1* cockpit SoC), tri-chip (2* AD SoC + 1* cockpit SoC or 1* AD SoC + 2* cockpit SoC), and quad-chip (2* AD SoC + 2* cockpit SoC) combinations.
Among them, the dual-chip integrated cockpit-driving domain controller (1* AD SoC + 1* cockpit SoC) solution is the current mainstream, including forms such as 1* NVIDIA Thor + 1* Qualcomm Snapdragon 8 Gen 3, 1* AD chip (such as Qualcomm 8650, NVIDIA Orin, etc.) + 1* Qualcomm 8295 cockpit chip, and 1* NIO Shenji + 1* Qualcomm 8295 integrated cockpit-driving.
Meanwhile, the single-chip integrated cockpit-driving platform is in a period of rapid expansion. Currently, the market size is still small, but the growth rate is obvious, with its penetration rate increasing from 0.5% in 2025 to 2.0% from January to April 2026. In 2026, it is expected to become an accelerated year for the mass production and integration of single-chip cockpit-driving vehicles. The Qualcomm 8775 integrated cockpit-driving is mass-produced and integrated into multiple mainstream models such as BAIC Arcfox, with a rapid increase in installation volume. At the same time, the Qualcomm 8797 integrated cockpit-driving is also mass-produced and integrated into Leapmotor's high-end model D19. Additionally, new platforms such as GeoDrive Dragon Hawk II, Horizon Robotics Starry, ZTE Microelectronics Lanyue A1, and Renesas R-Car X5H have all announced that they will be mass-produced and integrated into vehicles in 2026-2027, which will further promote the rapid development of the single-chip integrated cockpit-driving market.
Driven by multiple new factors such as central integration and cost reduction, the single-chip integrated cockpit-driving is expected to develop rapidly.
In October 2025, models such as BAIC Arcfox Alpha T5/S5, equipped with an integrated cockpit-driving platform built on Qualcomm 8775, were mass-produced and launched. This product was jointly developed and supplied by AutoLink and Zhuoyu Technology. AutoLink undertook the development in the cockpit field. Through the embedded AI large model combined with the capabilities of the cloud large model, it can provide a series of AI applications for the whole vehicle. Zhuoyu Technology developed the intelligent driving functions with a full-stack approach. Relying on the Qualcomm Snapdragon SA8775P platform, the integrated cockpit-driving controller can not only provide high-level intelligent driving capabilities covering full scenarios but also support AutoLink's cockpit capabilities.
This integrated cockpit-driving domain controller is built on the Qualcomm Snapdragon SA8775P and a Renesas RH850U2A. It can connect to 4 screens, 12 cameras, 5 millimeter-wave radars, and 12 ultrasonic sensors, realizing a single-chip multi-system architecture, creating integrated cockpit-driving services and atomic middle-platform services for the cockpit. It builds an intelligent interactive product that integrates domain control + peripheral devices + perception algorithms + ecosystem services, understanding the car, the road, and the user even better.
This domain controller not only brings advantages such as powerful AI computing capabilities and ASIL-D level functional safety features, but also can meet ADAS driving assistance functions such as Lane Keeping Assist (LKA), Automatic Emergency Braking (AEB), and Automated Parking Assist (APA), as well as L2+ intelligent driving functions such as Highway NOA and memory parking supported by ADAS software stack suppliers, and NCAP and regulatory functions such as the integrated Camera Monitoring System (CMS).
"Contact Information" (Mobile number is the same as WeChat number)
Industry Research Department | Mr. Fu 15810027571
Mr. Zhao 18702148304
Data Service Department | Ms. Zhang 13716037793
Strategic Consulting Department | Ms. Han 15810133447
Promotion and Communication Department | Ms. Liao 13718845418
Mr. Du 13910162318