28 December 2025
Chinese automotive chipmaker SiEngine has officially begun mass production of its 7-nanometer advanced driver-assistance system (ADAS) chips, marking a major milestone in the automotive semiconductor industry and accelerating partnerships with global automakers. The new chips — named AD1000 and AD800 — bring industry-leading computing power, safety standards and scalability to support a wide range of autonomous driving applications from Level 2 driver assistance to higher-level automated driving.
According to the company’s announcement, the AD1000 is built on a 7nm automotive-grade process and features a heterogeneous multi-core architecture that integrates high-performance CPUs, GPUs and neural processing units (NPUs) optimized for AI workloads. Each AD1000 chip delivers up to 512 trillion operations per second (TOPS) of AI compute — a critical capability for real-time perception, decision-making and sensor fusion in ADAS and autonomous driving. What’s more, the architecture allows multiple chips to be used together, scaling total performance up to 2,048 TOPS for even more demanding use cases such as complex urban navigation and higher automation functions.
Safety and reliability have been central to the design. The AD1000 platform includes a top-tier ASIL-D safety island, aligning with ISO 26262 functional safety standards — a must for systems that handle critical tasks like collision avoidance, lane keeping and automated steering. Chip features also include cryptographic hardware security modules, broad connectivity support (including PCIe 4.0, 10GbE time-sensitive networking and CAN-FD interfaces), and high memory bandwidth for fast data handling.
SiEngine’s chairman highlighted that the AD800 serves as a more cost-effective chip in the same family, aimed at mainstream ADAS applications where extreme computing power isn’t essential but reliability and scalability still matter. Together, the AD1000 and AD800 position SiEngine to meet a wide spectrum of vehicle electrification and autonomy needs — from advanced driver aids to future traffic-aware autonomous platforms.
The chips have been optimized for both traditional convolutional neural networks and next-generation transformer-based models, enabling efficient execution of large neural networks that vehicle perception systems increasingly rely on. With support for up to 20 camera inputs, broad sensor integration and compatibility with mainstream AI frameworks, the platform is designed to shorten development cycles for Tier-1 suppliers and OEMs alike.
SiEngine is closely connected to China’s broader automotive ecosystem — the company was co-founded by Ecarx (backed by Geely Holding) and ARM China, and it already supplies smart cockpit solutions to a number of major brands including Geely Galaxy, Lynk & Co, Volvo, Hongqi and Changan NEVO. The mass production of these chips not only strengthens China’s local semiconductor capabilities but also enhances its competitive stance in global ADAS and autonomous technology markets.
Automakers are racing to integrate higher levels of autonomy into next-generation vehicles, and powerful, automotive-grade chips like the AD1000 and AD800 are becoming indispensable. By delivering scalable performance, robust safety compliance and flexible development support, SiEngine’s new offerings could become a key building block for future intelligent vehicles that rely on AI-driven perception and control systems.
