On July 28, 2026, China's Hangzhou semiconductor design company XinLianXin officially announced that its RISC-V-based AI accelerator chip "X200" has successfully taped out. According to the company, the chip is designed for edge computing scenarios, achieving a 50% performance improvement and 30% power reduction in typical visual inference tasks compared to the previous generation. XinLianXin CEO Li Wei said in an interview: "The tape-out of X200 marks a solid step forward for China's RISC-V ecosystem in the AI field, and we expect mass production to begin in Q4 this year."
Technical Breakthrough: Integration of Custom Instruction Set and Neural Network Acceleration
The X200 chip uses a 12nm CMOS process and integrates four custom RISC-V processing cores with a dedicated neural network accelerator (NNA). The NNA supports INT8 and FP16 precision with high programmability, optimized for tasks such as image classification, object detection, and speech recognition in edge devices. XinLianXin VP of R&D Wang Lei explained: "We designed the vector extension unit independently and optimized hardware for Transformer models, achieving 4 TFLOPS (FP16) AI computing power at similar power consumption, a leading indicator among similar RISC-V chips."
Notably, the X200 also includes a built-in hardware security module supporting end-side encryption and trusted execution environment, meeting data privacy requirements in industrial IoT and smart city scenarios.
Market Prospects: Edge AI Boom Creates New Opportunities for Domestic Chips
According to semiconductor research firm IC Insights, the global edge AI chip market is expected to reach $12 billion in 2026, with China accounting for over 30%. With the proliferation of 5G and IoT devices, terminal inference demand is surging, and RISC-V, with its open-source, flexible, and customizable features, is becoming the preferred architecture for many startup design companies.
Founded in 2021 and headquartered in Hangzhou, XinLianXin focuses on RISC-V AI chip design. Previously, the company launched the X100 series for smart cameras, with cumulative shipments exceeding 500,000 units. The newly taped-out X200 will target smart industrial cameras, edge servers, and low-power AI boxes. Li Wei revealed that five potential customers have signed procurement agreements, covering security and intelligent manufacturing sectors.
Industry Chain Collaboration: Domestic Foundry and Packaging Support
This tape-out received strong support from domestic foundry SMIC. The X200 completed production on SMIC's 12nm advanced process line with excellent yield. Packaging and testing were handled by JCET. XinLianXin supply chain director Zhang Ming noted: "Close cooperation with local foundry and packaging partners not only ensures capacity security but also shortens product time-to-market."
Analysts generally believe that the rapid development of RISC-V in China benefits from the open-source ecosystem and policy support. In 2025, MIIT issued the "Guiding Opinions on Promoting the Development of Open-Source Chip Industry", explicitly supporting RISC-V basic software and design tools. XinLianXin's successful tape-out is a direct reflection of this policy dividend.
Industry Impact: Challenges and Opportunities Coexist
Despite the bright prospects, XinLianXin still faces competition from established RISC-V vendors such as StarFive and T-Head, as well as deep software ecosystem barriers from ARM and x86 camps. However, Li Wei remains optimistic: "The essence of RISC-V is customization. We focus on edge AI as a niche track, delivering higher energy efficiency through vertical optimization. In the next three years, we plan to launch a series of chips covering endpoints, edge, and cloud."
Additionally, XinLianXin has joined the RISC-V International Foundation and invested heavily in basic software adaptation. The company expects X200 to officially enter mass production and delivery in Q4 2026, potentially breaking the monopoly of international vendors in the edge AI chip market.
- Chip model: X200 RISC-V AI accelerator chip
- Process node: 12nm CMOS (SMIC)
- AI computing power: 4 TFLOPS (FP16)
- Main applications: Edge inference, smart industry, IoT
- Expected mass production: Q4 2026
Overall, the successful tape-out of XinLianXin X200 injects a strong boost into China's RISC-V chip design, showcasing domestic companies' ability to achieve differentiated innovation under an open-source architecture. As mass production approaches, the market will watch its real-world performance with anticipation.
