AI inference infrastructure startup Wuwenxinqiong (无问芯穹) has announced the closing of a new funding round of over 700 million yuan, led by Hangzhou Gaoxin Financial Investment Group and Huiyuan Capital, with participation from Guoxing Capital, Qinhuai Data, GF Qianhe, CSC Capital, and others. Existing investors Junlian Capital, Shanghai Guotou Futeng, and Yuanzhi Future also added follow-on investments. Combined with its previous nearly 500 million yuan Series A and nearly 500 million yuan Series A+, the company, founded in 2023, has now raised a total of over 2.2 billion yuan.
Wuwenxinqiong was founded by Wang Yu, a professor at Tsinghua University's Department of Electronic Engineering, and led by CEO Xia Lixue, also from the same lab. The company builds the middleware layer between large models and chips: it unifies scheduling across chips from Nvidia, Huawei's Ascend, Biren Technology, Moore Threads, and others, allowing models to run without being tied to specific hardware. It now covers 53 data centers across China, integrates over 15 chip types, and commands a total computing power of more than 25,000 PetaFLOPS. This kind of cross-chip orchestration puts heavy demands on the network — OpenAI recently open-sourced the network protocol that keeps its own massive AI clusters in sync.
The company's Agentic MaaS (Model as a Service) platform now hosts more than 160 models. As of the end of April, daily token call volume had grown more than 20x compared to the end of last year.