PUBLISHER: ResearchInChina | PRODUCT CODE: 2007698
PUBLISHER: ResearchInChina | PRODUCT CODE: 2007698
Automotive Cockpit SoC Research: Passenger Cars in the Price Range of RMB100,000-200,000 Account for Nearly 50% of Total Sales, and New-Generation Cockpit SoC Products Largely Enter Mass Production
Passenger cars in the price range of RMB100,000-200,000 make up nearly 50% of total sales, and new cockpit SoC products will compete fiercely in this segment.
Qualcomm dominates the RMB100,000-200,000 passenger car market with its flagship chips Snapdragon 8155 and Snapdragon 8295. Meanwhile, Chinese chip vendors including SiEngine, Rockchip, and MediaTek are making rapid, large-scale breakthroughs in this segment.
SAIC GM Buick launched the ELECTRA brand in 2025. Its first model ELECTRA L7 has been mass-produced and launched on market with Qualcomm 8775, supporting flexible computing power allocation on a single platform across four scenarios: high-level ADAS, cockpit-driving balance, high-performance cockpit, and time division multiplexing.
BAIC Arcfox a5 YuanJing Intelligent Driving Edition is equipped with Qualcomm SA8775P, delivering 144 TOPS of computing power and realizing "single-brain collaboration" between the cockpit domain and intelligent driving domain. The system is deeply integrated with AI large models; memory configuration is 24GB + 256GB.
Geely Galaxy E5, powered by SiEngine Longying No.1 Pro, achieved a delivery of over 160,000 units in 2025, becoming a benchmark for Chinese chips in this price range.
Changan Deepal L06 and NEVO Q05, launched in November 2025, feature multiple high-performance chips such as MediaTek's flagship cockpit platforms Dimensity S1 Ultra and Dimensity P1 Ultra.
The 2026 Xpeng MONA M03, scheduled for launch in April 2026, upgrades to the MT8676 cockpit chip and Tianji AIOS 6.0, replacing the previous Qualcomm 8155P.
The mid-to-large battery electric SUV ID.UNYX 08 (expected to be launched in H1 2026) jointly developed by Volkswagen and Xpeng will adopt the China Electronic Architecture (CEA), offering two cockpit platform options: Qualcomm 8295 and MediaTek MT8676.
In SiEngine's case, its first flagship product Longying No.1 uses a 7nm advanced process, and integrates high-performance CPU, GPU, and independent NPU. It supports cockpit-parking integration and L2 ADAS functions, and also offers high-bandwidth, low-latency memory support and rich video signal input capabilities. Over 1 million pieces have been mass-produced since its launch, and widely used in Geely Galaxy E5, FAW Hongqi, and other mainstream models.
To meet the demand from the mid-to-high-end vehicle market, SiEngine has launched Longying No.1 Pro (Longying No.1 + NN ACC, 20% performance boost) and Longying No.2 (high-level cockpit-driving integration chip). The Longying No.2 Ultra and Longying No.2 Lite series are under development, and Longying No.3 for higher-level cockpit-driving integration will be rolled out in the future.
Longying No.1 Pro adopts a unique CPU hard isolation design for higher efficiency and greater memory bandwidth, supporting up to LPDDR5@6400MT/s. In terms of safety, it is an AEC-Q100 automotive-grade chip with stable reliability; it meets ASIL-B high-level functional safety for stable performance; it has obtained high-level information security certification to protect user privacy. In addition, it has built-in dual HiFi5 audio processing units to enhance sound quality, and 8TOPS@INT8 edge AI computing power for AI enablement; it supports fast system startup, and flexible and efficient power management to reduce power consumption during application operation.
Cockpit SoCs evolve toward advanced processes, SIP, cockpit-driving integration, and AI.
Today's high-performance intelligent vehicle cockpit SoCs can drive multiple displays, e.g., center console screen, cluster and HUD. Interaction modes of intelligent cockpits have expanded from physical buttons to multimodal interaction modes such as voice, gesture and vision (DMS/OMS). Additionally, cockpit-driving integration has developed from proof-of-concept to mass production. Cockpit-parking integration solutions have been successfully deployed in Xpeng M03, Galaxy E5, etc.; cockpit-driving integration solutions are advancing rapidly, with cockpit-driving integration chips such as Qualcomm SA8775/SA8797, and NVIDIA Thor coming in strong.
In 2025, the following changes have been taken place in intelligent cockpit SoCs inside and outside China:
Chip manufacturing processes become more advanced, and the installation rate of chips below 10nm process has increased. Wherein, the proportion of 4nm chips rose from 0.4% last year to 2.8%, and the proportion of 5nm chips rose from 7.0% last year to 10.7%. MediaTek's latest chip Dimensity S1 Ultra is the world's first mass-produced 3nm automotive-grade chip, with a much higher energy efficiency ratio.
System In Package (SIP) mainly integrates 5G modems or wireless communication modules, realizing the integration of high-speed connectivity and intelligent computing capabilities through a single chip, and also helping OEMs reduce costs. For example, the D9000 co-developed by BYD and MediaTek integrates MediaTek M80 baseband, supports Sub-6GHz bands with a downlink rate of 7Gbps, and is compatible with 2G~4G networks. It supports Wi-Fi 7 and Bluetooth 5.3 with a theoretical peak rate of 6.5Gbps, and dual-frequency concurrency. At present, BYD's flagship models such as Bao 8, Denza Z9, Denza N9, and Yangwang U7 are fully equipped with D9000.
Cockpit-driving integration SoCs have entered mass production. As the first mass-produced single-chip cockpit-driving integration solution, Qualcomm SA8775P has been launched with models like new Arcfox aT5/S5, SAIC-GM Buick ELECTRA L7/Encasa, and Dongfeng Nissan N6 since October 2025; Leapmotor D19 and Li L9 (2026 model) have also been the first to be equipped with SA8797, Qualcomm's next-generation cockpit-driving integration SoC, and scheduled to be officially launched in the second quarter of this year. The adoption by more models and the continuous launch of multi-domain fusion/cockpit-driving integration SoCs such as NVIDIA DRIVE Thor series and Renesas R-Car X5H mark the migration of cockpit SoCs to multi-domain fusion/cockpit-driving integration SoCs.
The intelligent cockpit is evolving from the traditional "function collection" to an "intelligent partner". AI is no longer presented as a single functional module in the intelligent cockpit, but gradually evolves into a central capability that organizes the entire human-computer interaction logic. As the core carrier of this transformation, AI Agent achieves the leap from "passive response" to "active service" by integrating perception, decision, execution and memory capabilities. To support edge AI large models, cockpit SoCs require higher NPU computing power, higher memory bandwidth, and more stable and sustained performance. At present, Qualcomm's new-generation cockpit SoC SA8397 has NPU computing power of up to 360 TOPS, three times faster than the previous generation, supports LPDDR5X storage with a memory bandwidth of 204.8GB/s, and allows for local operation of generative AI models.
OEMs accelerate the layout of AI large models, and AI BOX and Chiplet technologies support their development needs.
With the launch of the industry's first super agent - IM Ultra Agent, IM Motors builds the IM Fusion Nova, a cockpit-driving integration architecture which not only superimposes intelligent cockpit and high-level intelligent driving capabilities, but also is equipped with chassis-by-wire (Lingxi Digital Chassis) responsible for vehicle action execution. Wherein, the intelligent cockpit system is the world's first to connect the Qianwen large model which is responsible for understanding user needs and has the ability to perform complex tasks.
In addition, automakers such as Geely, Xpeng, NIO, Li Auto, Dongfeng, and GAC are accelerating the layout of vehicle AI large models. Emerging carmakers mostly adopt an independent R&D model, with examples including NIO NOMI GPT, Li Auto Mind GPT, and Xpeng Lingxi large model; some OEMs choose to cooperate with technology companies on R&D, e.g., iFlytek Spark large model and Huawei Qianwu engine large model.
The deployment of AI large models on vehicles marks a leap-forward evolution of automotive intelligent cockpits from the basic intelligence stage to the cognitive intelligence stage of "in-depth interaction and self-evolution", while driving cockpit SoCs to evolve toward higher computing power and larger storage:
At the chip level, cockpit SoCs can improve chip performance with Chiplet technology, while greatly reducing design thresholds to shorten R&D time. For example, Renesas R-Car X5H, in addition to the native NPU and GPU processing engines, is expanded based on Chiplet technology, and the on-chip NPU + external NPU can improve AI processing performance by at least 3-4 times or more. In the post-Moore era, Chiplet technology is the most promising and realizable disruptive technology path for high-compute platform chips, and can cope with the trend of future intelligent vehicle E/E architecture moving toward cross-domain integration and central computing platforms.
At the system level, OEMs can superimpose AI computing power through an external AI Box, deploying multimodal large models without reconstructing the entire vehicle architecture to achieve flexible expansion of AI capabilities. AI Box is an independent vehicle AI computing unit that does not rely on the closed architecture of traditional automotive-grade chips in terms of computing power, and supports direct connection to corresponding systems in the intelligent driving domain/cockpit domain/Internet of Vehicles domain in terms of functions to achieve plug-and-play. AI Box is generally equipped with high-performance, high-compute SoCs, such as Turing AI chip and NVIDIA Thor, and its flexibility is favored by automakers such as Xpeng, Geely, and SAIC.
In the cockpit, AI Box is mostly pre-installed in vehicle models. Pre-installed AI Box cooperates with domain controllers, aiming to achieve rapid mass production and deployment of vehicle large models through the technology route of superimposing high-performance AI Box on a medium-compute cockpit platform.
Preface (1)
Preface (2)
Cockpit SoC Development Trends: Functions
Cockpit SoC Development Trends: Technologies
Intelligent Cockpit Technology Development Trends (1)
Intelligent Cockpit Technology Development Trends (2)