AI Chip
Focusing on three cutting-edge areas: AI SoC, Edge AI, and Next-Generation Storage.
The Topology team is actively expanding its AI chip product line. We welcome colleagues with relevant expertise to connect and collaborate.
AI SoC
AI System-on-Chip
AI SoCs integrate compute units such as CPUs, GPUs, and NPUs into a single chip, delivering edge AI inference and training capabilities for robotics, autonomous driving, smart terminals, and more. As the primary computing engine for edge intelligence today, they power next-generation applications on-device.
Sweet Potato Robot focuses on AI chips and development platforms for robots, delivering end-to-end solutions from chip to algorithm. Widely used in service robots, educational robots, and industrial inspection scenarios.
RDK X3, RDK X5, and other robot development platforms
- Cost-effective robotic computing power
- Robust developer ecosystem
- Supports ROS 2 Framework
SpacemiT is dedicated to developing AI computing chips based on the RISC-V architecture, delivering high-performance, low-power AI inference solutions for edge computing and data center scenarios.
K1 Series RISC-V AI Processors
- Innovation in the RISC-V Architecture
- High-Efficiency AI Inference
- Edge Computing
Cixin Technology focuses on Arm-based general-purpose computing chips, delivering high-performance, low-power edge computing solutions for PCs, tablets, and edge devices.
Arm-based General-Purpose Computing SoC
- ARM Architecture Ecosystem
- High Performance, Low Power
- For PC and Edge
Qingwei Intelligence specializes in reconfigurable computing architecture AI chips. Its CGN series offers superior energy efficiency and flexibility, targeting use cases in autonomous driving, edge computing, and data centers.
CGN Series Reconfigurable AI Chip
- Reconfigurable Computing Architecture
- High energy efficiency
- For intelligent driving and edge
Moore Threads is a leading domestic GPU chip company offering full-featured GPUs for graphics rendering, AI computing, and intelligent video processing, building a domestic GPU ecosystem.
MTT S Series Full-Featured GPU
- Full-featured GPU architecture
- Graphical + AI Fusion
- Domestic GPU Ecosystem
Cambricon is a pioneer in China's AI chip industry, offering products for the cloud, edge, and terminal to provide AI computing power for data centers and smart devices.
Siyuan 370/590 Cloud AI Chip, MLU Series
- All-Scenario AI Computing Power
- Proprietary Instruction Set Architecture
- Comprehensive software ecosystem
On-device AI
Edge AI Chip
Edge AI chips perform AI inference locally on devices without relying on cloud computing, offering low latency, high privacy, and low power consumption. They are widely used in smart security, smart homes, industrial inspection, and other scenarios.
Aftermo Intelligence specializes in developing compute-in-memory AI chips. By integrating computation and storage, we significantly reduce data movement power consumption, delivering high-energy-efficiency computing for autonomous driving and edge AI.
Houmo Hongtu Series Processing-in-Memory Chips
- In-memory computing architecture
- Ultra-high efficiency
- For Autonomous Driving
Rockchip is a leading domestic AIoT SoC vendor. The RK1820 series targets edge AI inference scenarios, integrates NPU computing power, and is widely used in smart vision, industrial control, and smart retail.
RK1820 Series AIoT SoC
- Integrate NPU computing power
- Rich AIoT Ecosystem
- High value
Hailo is an Israeli AI chip company specializing in edge AI accelerators. Its Hailo-8 series delivers high-performance AI inference at ultra-low power and is widely used in smart cameras, robotics, and industrial inspection.
Hailo-8/Hailo-15 Edge AI Accelerators
- Ultimate energy efficiency
- Native AI Architecture
- Extensive partner ecosystem
Sophon, an AI chip company under Bitmain, delivers comprehensive AI computing solutions spanning from cloud to edge. Its BM1684X series chips are widely deployed in video analytics and intelligent inference.
BM1684X/BM1688 Edge AI Chip
- Full-Stack AI Solution
- High computing density
- Complete development toolchain
DeepX is a South Korean AI chip company specializing in edge AI inference accelerators. Its DX series chips deliver exceptional performance in vision tasks such as object detection and semantic segmentation.
DX-M1/DX-H1 Edge AI Accelerator
- High-performance visual AI inference
- Low-power edge deployment
- Extensive model support
Axera Technologies focuses on visual AI chips, delivering high-performance AI perception solutions for smart cities, intelligent transportation, and industrial vision.
AX650N/AX630C Visual AI Chip
- High-Performance Visual AI
- Multi-channel video processing
- Smart City Solutions
Next-Gen Storage
In-Memory Computing and Emerging Storage Technologies
Next-generation storage chips break through the von Neumann bottleneck with innovations like in-memory computing, ReRAM, and MRAM to deliver higher energy efficiency and lower latency for AI inference and edge computing.
Zhicun Technology is a pioneer in China's compute-in-memory AI chip sector, leveraging Flash storage for in-memory computing to significantly reduce power consumption and latency in AI inference.
WTM2101 Computing-in-Memory AI Chip
- Flash Compute-in-Memory
- Ultra-low-power AI inference
- Wearable Device Application
Bowei Logic specializes in designing next-generation storage controller chips, delivering high-performance control solutions for emerging memory technologies such as ReRAM and MRAM.
New Storage Controller Chip
- Next-Generation Storage Control
- High-Reliability Design
- Multi-protocol support
Jiutian Ruixin specializes in computing-in-memory AI chip development, delivering ultra-low-power AI inference through analog CIM technology for wearables, IoT, and edge intelligence applications.
ADA Series Compute-in-Memory AI Chips
- Analog in-memory computing
- Ultra-low-power AI inference
- For Wearables and IoT
Yizhu Technology specializes in RRAM (Resistive RAM) computing-in-memory chip development, delivering high-efficiency AI computing solutions tailored for large model inference scenarios.
ReRAM Compute-in-Memory Large-Scale Chip
- ReRAM Technology Roadmap
- Optimized for Large Model Inference
- High energy efficiency