Understanding AI Edge SOC in Three Minutes

——Core Track of Edge Computing: Dual Main Lines of Localization and Scene Implementation

1. What is AI Edge SOC? Why is it Important?

Core Definition: AI Edge SOC (System on Chip) refers to a chip that integrates CPU, GPU, NPU (Neural Processing Unit), and ISP (Image Signal Processor) modules, specifically designed to provide localized AI computing power for terminal devices (such as smartphones, cars, robots, IoT, etc.) without relying on the cloud. Logic of Explosion:

1. Privacy and Latency Needs: Local data processing is safer in scenarios like healthcare and security;
2. Cost and Power Consumption Optimization: Edge computing reduces cloud transmission costs;
3. Deployment of Large Models on Edge: Devices like smartphones and PCs need to support lightweight large AI models (such as localized GPT-4).

2. Core Links in the Industry Chain and A-Share Listed Companies

1. Chip Design (Highest Technical Barriers)

• Leading SoC Companies:
  • Rockchip (603893.SH): Chips like RK3588 are widely used in robots, AR/VR, and edge servers;
  • Allwinner Technology (300458.SZ): Automotive-grade SoCs adopted by BYD and Xpeng;
  • Amlogic (688099.SH): Leading SoC for smart TVs and set-top boxes, extending into AIOT.
• NPU IP Licensing:
  • Cambricon (688256.SH): NPU IP licensing + cloud chips, with a comprehensive ecological layout;
  • Chipone Technology (688521.SH): AI chip design services + NPU IP, with clients including Intel and Samsung.

2. Application Scenarios (Implementation is Key)

• Smartphones/AI PCs:
  • Unisoc (688049.SH): SoCs for smart headphones and watches, entering edge AI voice;
  • Fuhang Microelectronics (300613.SZ): Leading ISP chip for security, extending into automotive smart cockpits.
• Automotive Smart Cockpits/Intelligent Driving:
  • Horizon Robotics (not listed): The Journey series chips compete with Nvidia, with extensive cooperation from domestic car manufacturers;
  • Desay SV (002920.SZ): Leading domain controller for smart cockpits based on SoC.
• Robots/IoT:
  • Hangzhou Huanxin Technology (688608.SH): Smart audio SoC, supplier to Huawei and Samsung;
  • Espressif Systems (688018.SH): IoT Wi-Fi/Bluetooth chips, with significant advantages in open-source ecosystems.

3. Supporting Industry Chain

• Storage: Beijing Junzheng (300223.SZ) automotive-grade storage chips;
• Packaging and Testing: JCET (600584.SH), Tongfu Microelectronics (002156.SZ) with increased demand for advanced packaging;
• Equipment Materials: SMIC (688012.SH), North Huachuang (002371.SZ) benefiting from domestic substitution.

3. Investment Logic and Risk Warning

Core Opportunities

1. Domestic Substitution: US sanctions accelerate the penetration of domestic chips in automotive and security scenarios;
2. Technological Iteration: NPU computing power competition (upgrading from 10TOPS to 100TOPS);
3. Ecological Binding: Companies collaborating with leading automotive and smartphone manufacturers are more likely to scale.

Risk Warning

• Technological route changes (e.g., storage-computing integration replacing traditional architectures);
• Intensified price wars (intense competition in mid-to-low-end chips);
• Fluctuations in downstream demand (consumer electronics recovery not meeting expectations).