In 2024, with fierce competition from giants such as Qualcomm, Nvidia, Intel, and MediaTek, as well as the rise of domestic chip manufacturers, the smart cockpit chip market ushered in an unprecedented "era of fierce competition".
Who will become the next "chip superpower"? How will this technological revolution reshape the "smart space" of future cars? Let's find out.
The concept of cockpit chips
The intelligent cockpit chip is a concept that evolved from the automotive E/E architecture. It is an important component of the domain control architecture (a method of dividing different control areas according to functions, also known as the Domain architecture), and the chip that controls the cockpit domain is called the cockpit chip.
The cockpit chip is like the brain of the cockpit, responsible for processing and controlling signals from various devices within the cockpit. External hardware devices in the smart cockpit domain also include connectivity subsystems, audio subsystems, camera subsystems, display subsystems, storage subsystems, and functional safety subsystems. The key word for the transformation of the traditional cockpit is "intelligence." The smart cockpit will integrate new technologies such as artificial intelligence, autonomous driving, and AR to achieve a multi-screen integrated interactive experience, including the central control unit, LCD instrument panel, head-up display (HUD), and rear-seat entertainment system.
Prior to 2015, cockpits were primarily controlled by MCUs or low-computing-power SoCs. As functionality has become more comprehensive, high-computing-power smart cockpit SoCs have become the mainstream.
Cockpit SoCs typically integrate multiple high-performance, multi-core processing units, including CPUs (Central Processing Units), GPUs (Graphics Processing Units), AI processing units, DSPs (Digital Signal Processors), and ISPs (Image Signal Processors). These processing units work together to provide powerful data processing capabilities and rich functional support for the automotive cockpit.
Specifically, the CPU is responsible for data calculation, control, and storage; the GPU focuses on image processing, providing high-quality graphics rendering for in-cabin display devices; the AI processing unit has intelligent and learning capabilities and can be applied to visual processing and functional expansion in the cockpit domain; the DSP is used for efficient processing of digital signals; and the ISP focuses on image signal processing and optimization.
Furthermore, while providing powerful functionality, cockpit chips must also prioritize information and data security. Therefore, SoCs typically employ a hybrid critical design, running both secure and non-secure workloads simultaneously to ensure that personal data and critical information within the vehicle are protected from unauthorized access and attacks.
Three major trends in cockpit chips
As automotive EEA (Electronic/Electrical Architecture) architecture continues to evolve towards integration, the industry is currently moving from domain architecture to zonal architecture. The next step in zonal architecture will involve even greater integration. In this scenario, the boundaries between the cockpit domain and the autonomous driving domain will gradually blur.
However, industry development cannot happen overnight, and the number of in-vehicle control chips will not immediately decrease from several to just one. Currently, the evolution of cockpit chips mainly includes three major trends:
First, the in-cabin display: one chip, multiple screens.
In traditional cockpits, systems such as the central control unit, instrument cluster, and head-up display (HUD) are controlled by independent ECUs. With the development of integration, these systems are being integrated into a single cockpit domain controller, forming a "one-chip-multiple-screen" solution. This means that a single high-performance SoC chip drives multiple screens (such as the central control screen, instrument cluster screen, and HUD).
The requirements for a SoC for "one chip, multiple screens" include: having multiple DP or DSI interfaces to support multi-screen display; having powerful CPU performance to ensure the smooth operation of multiple applications running simultaneously; having a high-performance GPU to support high-definition display and smooth animation; and having hardware support for hypervisor or hardware isolation to ensure stable operation of multiple systems.
Second, in-cabin interaction: multimodal interaction
The interaction methods of smart cockpits have expanded from traditional physical buttons to multimodal interactions such as voice, gestures, and vision (DMS/OMS), improving the user experience.
Front-end technologies for voice interaction include VAD, echo cancellation, and noise suppression, while back-end technologies include speech recognition, semantic understanding, and dialogue management. Currently, voice interaction does not require high NPU computing power and mainly relies on DSP and CPU.
Visual interaction, involving DMS (Driver Monitoring System), OMS (Occupant Monitoring System), and gesture control functions, is gradually being integrated into the cockpit domain controller. Furthermore, 3D TOF cameras support 3D gesture recognition and driver identification (Face ID), improving interaction accuracy and security.
Third, cross-domain integration: the cabin and driver are beginning to merge into one.
As cockpit functions are continuously integrated, they gradually merge with ADAS functions, forming a "cabin-parking integration", "cabin-driving integration", or even a "cabin-parking-driving" three-in-one system.
Integrated cockpit and parking systems connect surround-view cameras and ultrasonic radar to the cockpit domain controller, enabling 360-degree surround view and automatic parking (APA) functions. Advantages include reduced costs, optimized human-machine interaction, and full utilization of the cockpit's SoC computing power.
The cockpit-driver integration further integrates Level 2 ADAS functions, and even advanced autonomous driving functions. Currently, there are three implementation forms: One Box, One Board, and One Chip, with One Chip being the final form. Its advantages include reduced costs, improved system response speed, and easier iteration of new functions.
Key Players in Cockpit Chips
Currently, the market is gradually splitting into several schools of thought—traditional automotive electronics giants, cross-industry powerhouses in consumer electronics, and emerging domestic chip manufacturers. With the increasing integration of cabin, parking, and driving systems, some automakers are also entering this arena.
Specifically, looking at manufacturers, according to data from Gasgoo, in the domestic cockpit domain controller chip market from January to December 2024, Qualcomm firmly held the top position with 4,824,480 chips, accounting for 70% of the market share; AMD held 668,632 chips, accounting for 9.7% of the market share; Renesas held 380,610 chips, accounting for 5.5% of the market share; Chipone Technology and Huawei held 331,317 and 276,153 chips, respectively, accounting for 4.8% and 4.0% of the market share; Samsung Semiconductor, SemiDrive Technology, Intel, Nvidia and MediaTek saw a sharp increase in usage.
Currently, China has made significant progress in the field of cockpit SoCs and is gradually gaining a foothold in the market. Domestically produced chips have been in mass production since 2019. Unlike the traditional approach of replacing analog chips, power management chips, and MCUs with domestic alternatives, automotive systems, while aiming for smooth interactive operation, also have expanding computing power requirements. Emerging domestic companies are generally adopting new technologies and processes to achieve optimal performance.
International traditional automotive electronics giants
Renesas
Renesas has a strong foundation in automotive electronics, and its R-Car platform has a very wide product range, including autonomous driving or ADAS, connectivity, in-vehicle infotainment, cockpit and dashboard.
In particular, the R-Car X5H SoC, launched on November 13th last year, is the world's first automotive 3nm Chiplet chip and the world's second 3nm Chiplet chip. It's a converged chip, focusing not only on the development of individual systems like the cockpit, intelligent driving, and gateways, but on creating a versatile, cross-domain control and computing system. This product boasts a CPU computing power of up to 1000kDMIPS. In comparison, NIO's 5nm chip with 32 cores has a computing power of 615kDMIPS, using the Arm Cortex-A720AE core. The 32-core design is the first ARM V9.2 instruction set core specifically designed for the automotive field.
NXP
NXP places great emphasis on product portfolio solutions. Currently, many computing systems are distributed. To achieve software-defined vehicles, the current architecture is mostly divided into three layers: the top layer is the vehicle's computer, the middle layer is the local intelligent information system, and the bottom layer is the key terminal nodes. The computing systems in different areas are interconnected. NXP aims to provide appropriate system-level computing solutions to support the computing needs of these three layers, considering how different computing systems can be integrated within the vehicle, including the overall system performance, power consumption, and networking capabilities.
NXP's S32 CoreRide platform builds a robust and durable core foundation, serving as the central nervous system of a vehicle and encompassing key areas from propulsion and bodywork to connectivity, functional safety, information security, and energy management.
From the perspective of cockpit SoC alone, the i.MX 8 series and i.MX 8X series are both enduring products.
Texas Instruments (TI)
Texas Instruments' main divisions in automotive electronics applications are divided into five sub-sectors: hybrid, electric and powertrain systems, advanced driver assistance systems (ADAS), body electronics and lighting, infotainment systems and instrument clusters, and software-defined vehicles.
In the field of smart cockpit chips, TI dominated the automotive chip market in the early stages with its distributed architecture. As a key player in the automotive electronics sector, TI has also launched numerous products and solutions in the smart cockpit field, maintaining its important position in the industry thanks to its strong technological capabilities and innovation.
Consumer electronics cross-industry powerhouse
Qualcomm
Qualcomm's foray into the automotive market began in 2014, when the needs of automotive cockpit multimedia shared many similarities with those in consumer electronics, an opportunity Qualcomm keenly recognized. Leveraging its strong partnerships with leading global automakers, Qualcomm launched its first automotive cockpit chip—the Snapdragon 602A—followed by the second-generation Snapdragon 820A in 2016. The Snapdragon 820A's innovation lay in its ability to simultaneously control the instrument cluster display and the central infotainment screen. While this design was controversial in terms of cost savings, its powerful computing capabilities attracted the attention of emerging Chinese electric vehicle manufacturers. Li Auto, Desay SV, and Jike were among the first to adopt the Snapdragon 820A, and despite numerous challenges during development, they ultimately achieved a completely new cockpit experience.
As automotive cockpit functions become increasingly sophisticated, Qualcomm has gradually recognized the differences between automotive and mobile phone chips, and has specifically designed its third-generation cockpit chip, the Snapdragon 8155. The Snapdragon 8155 saw widespread adoption in vehicles in 2021, becoming the preferred choice for Chinese automakers. Last October, Qualcomm released its next-generation cockpit-driver integrated chips, the SA8797 and SA8799. Based on currently available information, the SA8797 features an 18-core design, a GPU computing power of 8.1 TFLOPS, and an AI computing power of 320 TOPS. These products are planned for sampling in 2025 and mass production in 2026.
However, Qualcomm is currently facing challenges, as all manufacturers are focusing on and working towards the cockpit-rider integration trend.
AMD
At last year's CES, AMD unveiled two new products designed to enhance the user experience in the automotive sector—the XA Versal AI Edge series and the Ryzen Embedded V2000A series processors—bringing new solutions to the digital cockpit. The XA Versal AI Edge, the world's first automotive-certified 7nm chip, features a new AI engine and vector processor array, significantly improving vehicle safety and optimizing the performance of key components such as LiDAR, radar, and cameras, injecting greater accuracy and responsiveness into vehicle systems. By processing massive amounts of data in real time, this chip enhances the vehicle's navigation and interaction with its surroundings, and is compatible with a variety of devices from LiDAR sensors to radar and cameras, demonstrating AMD's commitment to automotive innovation.
The Ryzen Embedded V2000A series processors, built on a 7nm process, feature Zen 2 cores and AMD Radeon Vega 7 graphics, delivering higher performance and a smoother user experience for digital cockpits. It supports up to four 4K displays, enhancing graphics quality and user input responsiveness, while meeting AEC-Q100 automotive standards to ensure reliability and stability. The series also offers 10 years of program availability, providing long-term support for automakers and partners.
Nvidia
Nvidia's primary focus is on autonomous driving; however, its single-chip Nvidia Thor solution, integrating cockpit, intelligent driving, and parking, is poised for release. Mass-produced vehicles equipped with Nvidia Thor are expected to launch in 2025. It has been confirmed that Zhiji and Jike will be the first to adopt it, while BYD, GAC, and Li Auto will replace their existing Orin chips with Thor by the end of 2025 or early 2026 at the latest.
The Orin series has four versions: the flagship Orin (12 cores, 275 TOPS), Orin-X (12 cores, 254 TOPS, the most commonly used in China), Orin NX (8 cores, 100 TOPS), and Orin Nano (6 cores, 70 TOPS). In contrast, the Thor series has five versions: Thor-Super (2000 TOPS), Thor X (1000 TOPS), Thor S (700 TOPS), Thor U (500 TOPS), and Thor Z (300 TOPS). Thor's biggest advantage lies in its "three-in-one" design, integrating cockpit, intelligent driving, and automatic parking functions, significantly reducing costs while maintaining high performance.
Samsung
Samsung's Exynos Auto V910, already in mass production, boasts approximately 1.9 TOPS of AI computing power, while the Exynos Auto V920 cockpit chip, planned for mass production around 2025, will significantly boost its NPU computing power to 30 TOPS. Meanwhile, Qualcomm's SA8155P chip, already in mass production, has an AI computing power of 8 TOPS, while its fourth-generation cockpit SoC chip achieves an NPU computing power of 30 TOPS, making it the cockpit SoC product with the highest AI computing power currently available on the market.
Intel
In early 2024, Intel re-entered the cockpit chip market with its first SDV SoC product. Employing Chiplet technology and the UCIe open interconnect standard, it broke away from the traditional monolithic SoC model, providing OEMs with customized computing platforms and diverse computing power combinations. Intel's open chiplet platform strategy aims to eliminate supplier lock-in risks for automakers and promote market competition and innovation, but its success depends on building a strong ecosystem.
Intel's SDV SoC, with its mature software ecosystem and powerful AI performance, can transform cars into living and working spaces, enhancing the in-cabin visual and voice AI experience and introducing generative large models. While Jike uses a TV box SoC in its MPV 009 to integrate a rich content ecosystem, adopting Intel's SoC not only achieves similar effects but also provides even stronger performance.
MediaTek
In March 2024, MediaTek and NVIDIA jointly launched four new Dimensity Automotive (DAuto) platform SoC chips: CX-1, CY-1, CM-1, and CV-1. These chips utilize an advanced 3nm process, making them top-tier products in the current DAuto SoC field. The new products' CPUs are based on the latest generation Armv9-A architecture and integrate NVIDIA's next-generation GPU-accelerated AI computing and NVIDIA RTX graphics processing technology, supporting the running of large language models on the in-vehicle edge. Furthermore, the chips highly integrate multi-camera HDR ISP and audio DSP functions, enabling innovative applications such as AR HUDs and electronic rearview mirrors.
The Dimensity cockpit SoC not only boasts high computing power and low power consumption, but also reduces BOM costs. It features a flexible AI architecture and high scalability, covering multiple automotive market segments from luxury to entry-level, and is planned for mass production and deployment in vehicles in 2025. This collaboration marks a new round of challenges from MediaTek and NVIDIA to Qualcomm. Leveraging MediaTek's market share and customer resources in the smart cockpit market, combined with NVIDIA's high-performance AI technology and brand influence, MediaTek aims to further seize market share in the cockpit segment.
Emerging forces in domestic chip manufacturing
SemiDrive
Currently, the ChipDrive X9 series of cockpit chips has become the mainstream choice for automotive-grade intelligent cockpit chips in China, covering more than 40 models and having dozens of key designated models. Models equipped with the X9 series chips from automakers such as SAIC, Chery, Changan, GAC, BAIC, and Dongfeng Nissan have all been put into mass production.
Last March, SemiDrive Technology released the X9H 2.0G, a new product in its X9 series of smart cockpit chips, dedicated to providing a more powerful and cost-effective cockpit infotainment system chip solution. The X9H 2.0G's CPU frequency has been increased from 1.6GHz to 2.0GHz, a 25% performance improvement, further enhancing the cockpit experience; it also features up to three pairs of dual-core lockstep Cortex-R5F cores, a built-in high-performance HSM engine and safety island, enabling it to be applied in scenarios with more stringent safety requirements.
In addition, SemiDrive's first-generation AI cockpit chip, X9SP, has been mass-produced and put into vehicles, supporting in-vehicle multimodal perception and cloud-based large-scale model interaction. The next-generation X10 has also entered the development stage. The flagship MCU product for regional controllers, E3650, has attracted industry attention. Samples were delivered to customers at the end of 2024, and it has been selected by several leading automakers.
In 2024, SemiDrive's global headquarters settled in the Beijing Economic-Technological Development Area and received a strategic investment of 1 billion yuan from the Development Area and the Beijing municipal and district governments.
Core Engine
Founded in 2018, ChipEngine Technology is a joint venture between ECARX and Arm China, focusing on the research and development of AI and high-performance chips for the automotive industry. Its flagship product, DragonEagle One, rivals the mainstream smart cockpit SoC Qualcomm Snapdragon 8155 in terms of specifications and has already entered mass production in models from major automakers such as Geely and FAW. Shipments are expected to reach millions in 2024.
The Longying-1, through its multi-core heterogeneous design, integrates a CPU with 100K DMIPS computing power, surpassing the Snapdragon 8155's 85K DMIPS. In terms of GPU, the Longying-1 is equipped with a GPU boasting approximately 900 GFLOPS of computing power and integrates an NPU unit with 8 TOPS @INT8 computing power. It also features an ASIL-D-level independent functional safety island and high-speed memory, demonstrating its strong competitiveness in the field of intelligent cockpit chips.
horizon
Horizon Robotics' Journey series chips, with their unique advantages, have achieved mass production in both the autonomous driving and smart cockpit domains. Through software and hardware co-optimization, they continue to expand their algorithmic leadership and accelerate product iteration.
In April of last year, Horizon Robotics released the Journey 6 series chips. The J6 series adopts the fourth-generation BPU architecture "Nash", which is optimized for large-scale parameter Transformer models and advanced intelligent driving. It has 6 configurations (B, L, E, M, H, P), with the J6P being the flagship product.
Cockpit chips: The battle reignites
In summary, the trends in cockpit chips in 2024 mainly focus on improving computing power, enabling generative AI, cockpit-vehicle integration, AR/VR and multi-screen immersive experiences, cost reduction and efficiency improvement of Chiplet, local AI multimodal perception, emotional intelligence, high-definition 3D interfaces, and architectural innovation, driving the transformation of the intelligent cockpit from an "information center" to a "smart space" and creating new value for automakers and users.
Since 2024, we can see that competition in the cockpit chip market has intensified: Intel returned to the automotive SoC field under performance pressure, attempting to break Qualcomm's first-mover advantage through its Sharp discrete graphics card and software-defined vehicle platform; MediaTek launched the 3nm CT-X1 with great fanfare, leading the cockpit chip process arms race.
Looking ahead to 2025, the integration of cockpit and driver has moved from proof-of-concept to mass production. Currently, the integrated cockpit-parking solution has been successfully implemented in models such as the XPeng M03 and Galaxy E5, with the market gradually picking up, and system costs decreasing by approximately 20%. Meanwhile, the integrated cockpit-driver solution is also rapidly advancing, with suppliers such as Bosch and Desay SV launching related solutions based on the Qualcomm 8775 platform, targeting the mainstream vehicle market priced between 100,000 and 200,000 yuan.
In 2025, the competition in cockpit chips will shift from multi-screen support to AI large-scale model support. Leading manufacturers such as NVIDIA, Qualcomm, and MediaTek will continue to maintain their advantages, while domestic chip manufacturers will need to address the challenges of process limitations by promoting the local foundry industry chain and ecosystem cooperation.
As the "cockpit-driver integration" trend becomes increasingly clear, Qualcomm is facing increasing pressure. It's worth noting that while the Qualcomm 8295 was successful, it failed to replicate the success of the 8155 due to changes in the market environment. Therefore, Qualcomm's near-monopoly position may be challenged in the coming years.
