--- title: "Qualcomm Anshuman: Flex is here, the popularization of cabin-drive integration and the introduction of large models on the device side are worth paying attention to." type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/31384639.md" description: "Produced by Zhineng Auto In 2025, the intelligent competition in China's automotive industry has entered the deep-water zone. L2+ level advanced driver-assistance systems (ADAS) are rapidly becoming popular, driving "practicality" and "scale" as the key words of the new phase. The competition, which previously focused on single-point technological capabilities, is gradually shifting towards the overall architectural design of the cockpit and driving assistance, especially the implementation speed of centralized electronic and electrical architectures, which is becoming a benchmark for measuring a carmaker's intelligent capabilities. At the 2025 Qualcomm Automotive Technology and Collaboration Summit..." datetime: "2025-07-02T05:45:19.000Z" locales: - [en](https://longbridge.com/en/topics/31384639.md) - [zh-CN](https://longbridge.com/zh-CN/topics/31384639.md) - [zh-HK](https://longbridge.com/zh-HK/topics/31384639.md) author: "[芝能-烟烟](https://longbridge.com/en/profiles/11273666.md)" --- # Qualcomm Anshuman: Flex is here, the popularization of cabin-drive integration and the introduction of large models on the device side are worth paying attention to. ​​ Produced by Zhineng Auto In 2025, the intelligent competition in China's automotive industry has entered a deep-water zone. L2+ level advanced driver-assistance systems (ADAS) are rapidly becoming popular, driving "practicality" and "scalability" to become the buzzwords of the new phase. The competition, which previously focused on single-point technological capabilities, is gradually shifting toward holistic architecture design centered around the cockpit and driving assistance. In particular, the implementation speed of centralized electronic and electrical architectures is becoming a benchmark for measuring a carmaker's level of intelligence.   At the 2025 Qualcomm Automotive Technology and Collaboration Summit, we saw Qualcomm's next-generation platforms under development: Snapdragon Ride and Snapdragon Ride Flex. They aim to address three core issues:   ◎ First, achieving the optimal engineering balance between computing power, energy efficiency, and cost.   ◎ Second, the scalability of the platform architecture.   ◎ Third, achieving true cockpit-driving integration and building a unified hardware-software system synergy on this foundation. During the summit, we also noted that Qualcomm is accelerating the "localization" of its automotive business. Not only is it promoting the adoption of its latest premium platforms, such as the Snapdragon 8397 and Snapdragon 8797, among Chinese automakers, but it is also making more thorough local adaptations at the ecosystem and development toolchain levels. These two chips are likely to become key variables in the new wave of cockpit-driving integration.   ![Image](https://imageproxy.pbkrs.com/https://wx2.sinaimg.cn/mw1024/64f0c940ly4i2y86d9beej201q00w0ni.jpg) **Pa****rt 1** **The Core Proposition of ADAS: "Implementation"** The core issue in the popularization of ADAS is not whether the technology can be developed, but whether a truly mass-producible system architecture can be built within the constraints of limited computing power, cost, and development resources.   This has been a common challenge for almost all automakers and chip manufacturers over the past few years. Qualcomm's recent practice in the Chinese market is based on its two product lines, Snapdragon Ride and Snapdragon Ride Flex, which focus on the three elements of "low cost, high efficiency, and strong scalability" to build a comprehensive ADAS platform for the mainstream market.   Qualcomm's solution is a "system package" rather than a mere accumulation of single-point capabilities. It provides most automakers with a more controllable engineering path.   This systematic approach is now entering a phase of concentrated harvest. According to Qualcomm's figures, 2025-2026 will be the window for large-scale mass production of its ADAS projects: globally, over 20 ADAS/AD projects are planned for implementation during these two years, with collaborating automakers covering Chinese and international brands such as Volkswagen, Geely, BAIC, Chery, GM, Honda, Mercedes-Benz, and BMW. Taking the Snapdragon 8650, the flagship SoC in the Snapdragon Ride platform, as an example, Qualcomm is attempting to do more within a given cost framework. The 8650 adopts a heterogeneous multi-core architecture, integrating high-performance CPU, GPU, NPU, and ISP. It shows significant improvements in inference efficiency, energy efficiency, and bandwidth utilization compared to previous generations and competing products. According to Qualcomm's data, its IPS (inferences per second) is 30% higher than competing SoC products, DDR bandwidth is as low as 1/7 of competing products, and energy efficiency is up to 2x that of competing products. These parameter optimizations can find specific applications in the highly cost-sensitive Chinese market. For example, the Leapmotor C10 and B10, with starting prices of 119,800 yuan, have achieved highway and urban NOA capabilities on the Snapdragon 8650 platform. Another chip in these models, the Snapdragon 8295, serves as the main controller for the AI cockpit. The entire system covers the full functional set of driving assistance and digital cockpits on a low-cost hardware foundation, driving high-level ADAS capabilities down to the 100,000-yuan vehicle segment. The Snapdragon Ride platform is not limited to chips but is a multi-level solution covering everything from basic ADAS to urban NOA.   From the Snapdragon 8620, mainly used for highway NOA scenarios, to the Snapdragon 8650 for urban NOA, the Snapdragon 8775 for cockpit-driving integration, and the high-performance premium Snapdragon 8797, combined with universal accelerators and software stacks, automakers can flexibly choose deployment paths based on vehicle cost structures and target markets. The core SoC of the Snapdragon Ride Flex platform, the Snapdragon 8775, is designed to perform parallel computing for ADAS and cockpit functions on a single chip, addressing the demand for "cockpit-driving integration" while supporting the highest ASIL-D safety mechanisms. It supports air and passive cooling, with significant optimizations in vehicle thermal management and wiring complexity, making it particularly suitable for mainstream vehicles priced between 100,000 and 200,000 yuan. Anshuman Saxena, Vice President of Product Management for Qualcomm's ADAS business, stated at the summit's main forum: "The multi-ECU domain architecture is a thing of the past. Integrated multi-computing module centralized zonal controllers have become an established trend. Central computing centered around a single general-purpose SoC is becoming a new phenomenon, and the Snapdragon Ride Flex SoC is precisely the SoC that meets this new trend of central computing. Flex is here!"   Another notable change is that Qualcomm's technological deployments in cockpit chips have been aligning with those in the mobile sector in recent years.   Four years ago, the manufacturing process of Snapdragon cockpit chips caught up with that of flagship mobile chips for the first time. The latest generation, the premium Snapdragon cockpit platform, the Snapdragon 8397, is the first to adopt Qualcomm's self-developed Oryon CPU, which was just applied in flagship mobile devices, and is customized for automotive needs, achieving generational leaps in performance. CPU and GPU performance are 3x higher than the previous generation, and NPU computing power is 12x that of the previous generation. For smart cockpits increasingly reliant on AI rendering and voice/image interactions, this means they now have the underlying hardware capability to support "large model deployment in vehicles." At the software level:   ◎ Qualcomm has deeply integrated with local Chinese algorithm teams—leading solution providers such as Momenta and Yuanrong Qixing have already completed integration on the Snapdragon Ride platform. ◎ At the same time, Qualcomm has built a complete toolchain for developers, including virtualized development environments and version consistency migration mechanisms, ensuring algorithm compatibility across multiple generations of products, reducing testing costs, and significantly shortening vehicle development cycles. According to incomplete statistics, over 30 Chinese automakers and Tier-1 suppliers have deployed Snapdragon Ride as their core ADAS platform. From entry-level vehicles priced at tens of thousands of yuan to high-end products over 300,000 yuan, the adaptability and performance distribution of the Snapdragon Ride platform basically cover the needs of mainstream vehicle segments, offering solutions that balance architectural hierarchy, computing power configuration, and cost control.   The Snapdragon Ride Flex platform has further pushed the boundaries of "software universality." Leveraging virtualization architectures and task isolation mechanisms, developers can complete algorithm migration and testing on existing Snapdragon cockpit platforms or Snapdragon Ride platforms and then directly deploy them on Snapdragon Ride Flex.   This not only reduces the risk of stack fragmentation but also avoids the cycle and budget waste caused by repeated validation. Most importantly, it adopts a virtualization architecture and isolation mechanisms, allowing multiple operating systems and functional domains to be deployed on the same chip, ensuring that ADAS task security does not interfere with the user experience in the cockpit domain, achieving "both safety and experience." Regarding this, Anshuman stated: "The Snapdragon cockpit platform and Snapdragon Ride platform adopt a heterogeneous architecture, where different technical modules (such as NPU, ISP, CPU, etc.) on a single SoC collaboratively run different application workloads, rather than relying solely on a single core engine. At the same time, this does not compromise performance or safety."   He also emphasized that safety is the most important cornerstone of Snapdragon Ride Flex, noting that these solutions have been tested and validated globally, including standards from North America, Europe, Germany, and China.   The Snapdragon Ride platform is characterized by system efficiency and architectural flexibility, while Snapdragon Ride Flex highlights Qualcomm's deployment capabilities in cockpit-driving integration and central computing. As China's ADAS enters the "mass delivery" phase, this approach has a certain degree of foresight. In summary, Qualcomm's advantages in the ADAS market are mainly reflected in three aspects:   ◎ Performance-power efficiency balance: Providing the computational support required for high-level functions while controlling power consumption and cost.   ◎ High architectural scalability: A unified platform covering entry-level to high-end vehicles, reducing resource waste.   ◎ Ecosystem collaboration capabilities: Forming a closed-loop support system from SoCs and algorithms to vehicle development toolchains.   The combination of these three capabilities is determining whether an intelligent technology can truly enter the mass production cycle of mainstream vehicles.   ![Image](https://imageproxy.pbkrs.com/https://wx1.sinaimg.cn/mw1024/64f0c940ly4i2y86dd9jqj201d00u0m3.jpg) **Part 2** **Qualcomm's New Variable: On-Device Large Models**   China's new energy vehicle market remains the fastest-evolving battlefield for intelligence globally. Unburdened by traditional powertrains, Chinese automakers have greater freedom in the leapfrogging of electronic and electrical architectures and interaction layers. Qualcomm is betting on this window of opportunity.   According to Anshuman Saxena, at least four new Chinese automaker partners have been added after the 2025 Shanghai Auto Show, all planning to develop vehicles based on the Snapdragon Ride platform.   Regarding large model deployment in vehicles, he proposed a direction for intelligent evolution: "Everyone is focusing on unified AI experiences, large language models (LLMs), and visual language models (VLMs). Why not integrate some user interaction input data from the cockpit with the ADAS system to avoid the burden of running multiple different instances of the same AI, LLM, or VLM on the system? This is precisely where Snapdragon Ride Flex excels—we are realizing these experiences based on this platform."   According to Qualcomm's official data, over 10 Chinese automakers and Tier-1 partners are adopting Qualcomm's premium automotive platforms (including the Snapdragon 8797 and Snapdragon 8397) to develop ADAS and smart cockpit solutions. For example, Leapmotor's upcoming D-series flagship model, set to launch in Q1 2026, will be the first to feature dual Snapdragon 8797 chips: one 8797 for ADAS tasks and another for the AI cockpit.   Qualcomm's path to supporting the automotive industry in deploying on-device large models is also becoming clearer: Based on the premium Snapdragon automotive platform, through model lightweighting and scheduling optimization, Qualcomm has partnered with Tier-1 manufacturers to achieve smooth on-device operation of 14 billion-parameter (14B) large models locally, as well as integrated 7 billion-parameter (7B) on-device model solutions—including scenarios requiring privacy protection (such as in-car conversations) and low-latency responses (such as real-time vehicle control). From a performance metrics perspective:   ◎ 14B model inference achieves over 40 FPS, optimized to 50-60 FPS.   ◎ 7B models achieve 60–72 FPS, meeting real-time interaction needs, such as understanding user voice and semantics and building user profiles. The significance of on-device deployment lies not only in improved response speeds (reduced from 1–2 seconds in the cloud to 0.2 seconds on-device) but also in enhanced privacy protection and connection stability.   For example, sentry mode can now use large models to identify abnormal behavior (such as door handle pulling or vehicle scratches), automatically generating event summaries and highlight clips, replacing the traditional 3-minute video segments and significantly reducing redundant data. A notable viewpoint from Qualcomm is worth attention—"AI is the new UI." This change is reshaping the underlying logic of smart cockpits: shifting from "humans adapting to machines" to "machines understanding humans." Human-machine interaction no longer centers on menu clicks but instead completes more natural execution of commands through semantic understanding and multimodal perception.   **Summary** From chips to systems, platformization is becoming the core variable determining the mass production pace of intelligent technologies Returning to the fundamental question: It's not about who has stronger computing power or larger model parameters, but about who can use the least development cost to stably, quickly, and scalably deploy "AI capabilities" in mass-produced vehicles. Qualcomm's answer is the "platform thinking" represented by Snapdragon Ride and Snapdragon Ride Flex: Using scalable architectures, universal software stacks, and strong ecosystem adaptability to lower the barriers for automakers in each leap from L2+ to cockpit-driving integration and then to AI interaction. Chips are just the starting point; the real value lies in the engineering path. The deployment of AI in vehicles is transitioning from "whether it can be done" to "how to do it stably," and this is precisely the home field of platform players.​​​​ ### Related Stocks - [QCOM.US](https://longbridge.com/en/quote/QCOM.US.md)