Industry “Chip” News
Wosh Electronics
Industry News
As the global low-altitude economy surpasses one trillion yuan in 2024, the large-scale application of drones in logistics inspection, security monitoring, and other scenarios places higher technical demands on core chips. Dedicated SoCs (System-on-Chip) utilize heterogeneous computing architectures to significantly enhance the computational efficiency of real-time image processing and obstacle avoidance algorithms. Compared to discrete solutions, SoCs can reduce the volume of peripheral circuits by 40%, lower power consumption by 20-30%, and shorten development cycles by more than 50%.
Leading International Companies
01
NVIDIA
The Jetson series provides edge AI computing power (e.g., AGX Orin with 275 TOPS INT8), supporting real-time visual SLAM and target recognition, applied in DJI industrial models, requiring low-power flight control chips to balance the 15-30W high power consumption.
02
STMicroelectronics
The STM32H7 series MCU features a dual-core Cortex-M7 (480MHz) with <2μs real-time response capability, occupying 70% of the global industrial flight control market, but its graphics processing (22.7 DMIPS) struggles to meet advanced AI demands.
03
Qualcomm
The Snapdragon Flight platform reduces overall costs by 30-40% through SoC integration, supporting 4K video processing and deep learning obstacle avoidance, previously used in ZeroTech’s Dobby, but subsequent collaboration was affected by heat dissipation issues.
04
Intel
Integrating Atom processors with RealSense depth vision modules, achieving millimeter-level environmental perception and 3D reconstruction, but the solution’s cost is 35-50% higher than the industry average, limiting penetration in industrial scenarios.
05
Samsung
The Artik series uses 28nm FD-SOI technology, with the Artik 1 package measuring only 12mm² and standby power consumption <5mW, suitable for micro-drones, but due to insufficient system stability and price competitiveness, its market share is below 8%.
Key Domestic Manufacturers
01
Rockchip
The flagship RK3588 uses 8nm technology, integrating a quad-core [email protected] + quad-core A55 architecture, with GPU performance reaching 6TOPS, supporting DJI Crystalsky and other professional devices for 4K60fps real-time video transmission, with energy efficiency comparable to Qualcomm’s mid-to-high-end solutions.
02
Huawei HiSilicon
The SS928 vision processor integrates NPU (4TOPS) + VPU + DPU three engines, supporting Transformer network acceleration, providing environmental semantic analysis capabilities for intelligent security drones.
03
Phytium
The Tenglong E2000D chip meets the extreme conditions required for emergency rescue drones through hardware-level security isolation and electromagnetic interference design, having established a domestic low-altitude data link ecosystem.
04
Chengdu Huayi Microelectronics
Launched a dedicated AI chip with 16TOPS, supporting 8K video encoding and decoding and complex environmental perception for cargo drones, with a computing density leading competitors by 20%, and the next generation 100TOPS chip targeting autonomous decision-making in swarms.
05
Aerospace Microelectronics
The heterogeneous SoC based on RISC-V architecture achieves real-time target detection at 1280×720@30fps under a power constraint of 2W, breaking through the computational bottleneck for high-dynamic inspection tasks.
06
Allwinner Technology
The T113-S3 processor integrates a dual-core HIFI4 DSP, achieving <10W overall power consumption for self-stabilizing flight control and 7km video transmission, previously used in Xiaomi consumer drones.
Gaps, Ecosystem, and Competitive Focus
01
Design Gaps
The NVIDIA Jetson AGX Orin uses a 7nm process and CoWoS 2.5D advanced packaging, achieving a unit computing weight of 0.8g/TOPS; in contrast, the domestic Chengdu Huayi’s 14nm process 16TOPS chip has a unit computing weight of 2.3g/TOPS, severely limiting the adaptability of small drones. The process bottleneck is even more pronounced—Qualcomm/Samsung have mass-produced 5nm drone-specific SoCs (e.g., Snapdragon Flight Gen2), while the most advanced domestic solution, Rockchip RK3588, remains at the 8nm node, with over 80% reliance on imports for high-performance chips below 14nm. This results in a 35% increase in size and a 40% rise in power consumption for domestic chips at the same performance level, directly impacting the competitiveness of industrial-grade long-endurance models.
02
Software Ecosystem
The current drone software ecosystem shows a bifurcation between closed-source and open-source: closed-source platforms like DJI Manifold establish technical barriers through deeply optimized CUDA architectures, leading to a 50% increase in the cost of porting third-party algorithms; while open-source ecosystems like PX4 are impacted by new domestic airspace management regulations, with compliance modification costs accounting for 30% of total investment. The localization process is accelerated through solutions like Huawei Ascend’s OpenCL compatibility layer (reducing porting cycles by 40%), Phytium’s national secret-level TEE data link, and Allwinner’s TinaOS lightweight SDK, but the maturity of the toolchain still lags behind ROS2 by two years, necessitating the establishment of standardized middleware to reduce the risk of ecosystem fragmentation.
03
Competitive Focus
The NVIDIA 275TOPS solution controls power consumption at 30W, while the domestic Chengdu Huayi’s 100TOPS target chip is expected to reach 45W, facing a heat flow density bottleneck of >8W/cm². In the low-power race, STMicroelectronics’ STM32H7 achieves 480MHz computation at 1W, while Hengxuan Technology’s BES2800 achieves ultra-low power consumption of 0.6W through architectural optimization, but sacrifices 20% of performance for energy efficiency improvement.
In terms of multi-modal perception capabilities, Intel’s RealSense D455 integrates TOF and RGB sensing, while Huawei’s SS928, despite integrating a three-engine heterogeneous architecture, still has sensor fusion delays of >50ms. The construction of open platforms is also crucial—Qualcomm’s Robotics RB5 provides a complete development kit, while Allwinner’s TinaOS edge computing framework has a documentation completeness of less than 60%, limiting the development of the developer ecosystem.
The Moment of Breaking Through Chinese Chips
International leaders like Qualcomm and Intel dominate the high-end drone control market with advanced processes (such as 7nm/5nm) and powerful AI computing capabilities. Their chips have high integration, supporting complex AI visual processing (such as real-time target tracking and scene understanding), but are susceptible to fluctuations in the international supply chain.
Opportunities and Challenges for Domestic Chips
Opportunities:
Supply chain security demand: Geopolitical factors accelerate the domestic substitution of chips, with Yangtze Memory (NAND), and Meixin (magnetic sensors) entering the DJI supply chain.
Scene customization advantage: Local companies are closer to niche markets such as agriculture and inspection, allowing for self-developed dedicated SoCs to optimize energy efficiency.
Policy and market dual drive: National integrated circuit industry support policies and the world’s largest drone market provide dual support.
Challenges:
High-end process limitations: Key equipment such as EUV lithography machines are restricted, hindering the development of advanced SoCs.
Design talent gap: There is a shortage of talent in complex SoC architecture design and AI accelerator development.
Insufficient ecosystem maturity: There are still gaps in toolchains, IP libraries, and software adaptation compared to international giants.
Future Competitive Core:1. Rise of the RISC-V Ecosystem: The open-source architecture provides domestic chips with an opportunity for “curve overtaking,” avoiding ARM licensing restrictions and accelerating the construction of independent IP.
2. Breakthrough in Process Technology: Advanced processes below 14nm are the threshold for high-end SoCs, and domestically, there is a need to accelerate breakthroughs in lithography, materials, and other technologies, or to make up for gaps through design innovations like Chiplet.
3. Industrial Collaborative Ecosystem: Drone manufacturers and chip companies (such as Rockchip, Allwinner, Horizon Robotics, and Black Sesame) need to work closely together to optimize integrated solutions of “chip-algorithm-hardware.”
4. Cross-domain Technology Integration: AI chip technologies from autonomous driving (such as Horizon Robotics) and robotics should be migrated to drones to promote upgrades in perception and decision-making capabilities.
Drone SoCs represent a dual game of computing power and security. International giants lead in technology, while domestic chips are beginning to shine in flight control and image processing. The future’s ability to break through bottlenecks depends on whether the entire chain of “process breakthroughs-architecture innovation-ecosystem collaboration” can be connected: using RISC-V as a sword to break through intellectual property barriers; using scene customization as a shield to build market moats; and requiring chip manufacturers, system integrators, and algorithm companies to work together to build an open and win-win technological ecosystem.
Disclaimer: The above content is compiled from online information and does not represent our company’s views and positions, and is for communication and learning purposes only. If there is any infringement, please contact us for deletion.
The future has arrived, and the breakthrough winners will win. We will continue to strive to provide better services and quality!
Follow Us
Wosh Electronics
