LingRui ZhiXin raises hundreds of millions of yuan for RISC-V chips for AI agents

Chinese processor developer LingRui ZhiXin (灵睿智芯) has closed a new round of funding valued at several hundred million yuan — the deal was disclosed by the 36Kr portal citing the company itself.

Who invested in the round

Several industry investors participated in the financing: Zhangjiang Hi-Tech (张江高科) — the investment platform of the famous Zhangjiang technology park in Shanghai, Zhongke Chuangxing (中科创星), Oriental Fortune Capital (东方富海), Chengwei Capital (成为资本), Shixi Capital (石溪资本), as well as AI accelerator developer Biren Technology (壁仞科技) and a number of industry partners. Existing shareholders — funds Yuanhe Puhua (元禾璞华), Shanghai Guotou Futeng Capital and Huashan Capital — also increased their stakes in the company.

• The round is valued at "several hundred million yuan" (人民币数亿元), with the exact amount not disclosed
• Among new investors are Zhangjiang Hi-Tech, Zhongke Chuangxing and AI chip developer Biren Technology
• Existing shareholders Yuanhe Puhua, Shanghai Guotou Futeng Capital and Huashan Capital bought additional shares
• The company is already in negotiations about the next round

Where the money will go

The funds will be directed toward commercializing high-performance processor cores based on the RISC-V architecture, accelerating development of proprietary chips for autonomous AI agents — so-called "agentic CPUs" — and developing AI-native unified computing architecture.

RISC-V is an open and free processor instruction set that has been rapidly gaining weight in China in recent years: against the backdrop of US export restrictions on x86 and ARM technologies, local chip developers are betting on it as a path to technological independence. One of the round's co-investors, Biren Technology, is a notable player precisely in the AI accelerators segment: the company develops GPUs for training large models and has already gone through its own rounds of financing amid demand for domestic alternatives to Nvidia chips.

Why a separate agentic CPU is needed

The emergence of a separate class of processors for AI agents reflects a broader industry trend: the industry is seeking computing architecture not only for training and inference of language models, but also for new types of workloads — when the model doesn't just answer a query, but independently plans and executes chains of actions, appeals to external tools and maintains a long execution context for a task. Such workloads impose different requirements on the processor than classical inference — hence the attempts by startups like LingRui ZhiXin to single out this direction as a separate product.

The Zhangjiang technology park in Shanghai is one of China's oldest and largest centers of the semiconductor industry, around which chip design startups and state investment funds have historically concentrated; the participation of an investment platform associated with it in LingRui ZhiXin's round fits into the general practice of Chinese regional authorities co-financing directions critical to technological sovereignty through their own venture structures.

What this means

Funding of a niche Chinese developer of RISC-V chips for AI agents with participation of state technology parks and specialized venture funds shows that the competition for computing architecture for agentic AI is unfolding not only among giants like Nvidia and AMD, but also at the level of startups systematically supported by Chinese industrial policy.