Series A financing comes to light
Qujing Technology disclosed its Series A financing on July 13. The round was led by Huirong Fund under Henan Investment Group, while existing investors including Zhenzhi Capital, Shangshi Capital, Xinglian Capital, Shanghai Guofang Innovation, Honghui Fund, Huakong Fund, and Hangzhou Fucheng made additional investments.
The company was founded by a Tsinghua University team. Founder and CEO Ai Zhiyuan and CTO Chen Xianglin both came from the High Performance Computing Institute under Tsinghua University’s Department of Computer Science. Chinese Academy of Engineering academician Zheng Weimin serves as chief scientific advisor, professor Wu Yongwei is chief scientist, and associate professor Zhang Mingxing is listed as a co-initiator involved in technical strategy and key research work.
The article says Qujing has completed a technology transfer equity arrangement with Tsinghua University and has become one of the university’s technology commercialization cases. It adds that the company has raised more than 1 billion yuan in the past six months.
Started with inference, not training
The report traces the company’s origin to 2023, when ChatGPT accelerated global interest in generative AI. At the time, many AI startups in China focused on large models, and even AI infrastructure startups were largely centered on training. Qujing chose to start with inference and positioned itself around high-quality AI Token production.
According to the article, professor Wu Yongwei and Zhenzhi Capital founder Ren Xuyang jointly initiated the company, with its technical roots tied to Tsinghua University’s High Performance Computing Institute. Qujing was formally established in late December 2023.
Ai, a PhD from Tsinghua’s High Performance Computing Institute, had previously led R&D functions in big data, digitalization, and AI application units at a listed company. Co-initiator Zhang Mingxing focuses on computer architecture research and had participated in infrastructure work for leading large-model companies. In March this year, Dr. Wu Wenjie became president of the company, adding finance management and global operations experience.
Ai said in the article that training is a cost item, while inference is the part that makes money. On that basis, Qujing set out to become a partner for building and operating Token factories in the AI era.
Focus on high-quality AI Token production
The company’s pitch is built around what it calls high-quality AI Token output. The article says enterprise customers are no longer looking only for a model that can chat. They want a system that can complete real business tasks with stability, efficiency, and acceptable cost.
For models with hundreds of billions or even trillions of parameters, the report says production-grade AI Tokens must meet several targets at the same time: low first-token latency, high concurrency, stable output quality, structured result generation, and function calling, while also keeping unit generation cost within an acceptable range.
Qujing says it addresses that challenge through technologies it describes as “full-system heterogeneous coordination,” “memory-for-compute substitution,” and “virtual-real isomorphism.” It also proposed the Token as a Service, or TaaS, concept and built its ATaaS platform to improve the conversion from hardware investment into AI Token output.
Funding timeline and disclosed metrics
The article lists several recent financing rounds. In February, Qujing completed an angel++ round invested by Parallel Technology. In May, it closed a Pre-A round led by Xinglian Capital and Huakong Fund, with Honghui Fund, Tianhao Energy, Shangshi Capital, Tianjin Renai Hongsheng, and Hangzhou Fucheng participating, while existing investor GL Ventures also added capital.
The newly disclosed Series A round was led by Huirong Fund under Henan Investment Group, with several existing investors increasing their positions again. The report also says the company’s next financing round is already in progress.
It attributes investor interest to rising inference demand as AI Coding, OpenClaw, and related applications spread more quickly. The company’s operating strategy is described as “fewer models, deeper optimization,” meaning it concentrates on a limited number of key large models and production scenarios rather than expanding model count.
One line in the article states that less than 10% of leading domestic models account for most of the AI Token market. Based on that view, Qujing says it focuses resources on a small number of top models and core scenarios.
Token factory model
The article cites National Data Administration data showing that China’s average daily Token calls had exceeded 140 trillion as of March 2026, up by more than 1,000 times from two years earlier. It presents that as evidence that inference demand is expanding and that AI Token factories are becoming a visible layer of AI infrastructure.
Qujing currently has two business models. One is a direct-operation model, in which it rents or secures computing resources and directly produces high-quality AI Tokens for leading model developers, internet platforms, AI application companies, and large enterprise customers. The other is a co-operation model, aimed at customers that plan to own or already own computing resources, where Qujing handles overall planning, system integration, construction delivery, and later co-operations for AI Token factories.
The article quotes Ai as saying that more listed companies, state-owned enterprises, central state-owned enterprises, and intelligent computing centers across China want to move from traditional computing power leasing to higher value-added AI Token production. Qujing’s offer, he said, is a full design and buildout solution for AI Token factories.
The report also disclosed operating data. Since the 2026 Spring Festival, average AI Token production efficiency per computing unit has increased by more than three times, while total high-quality AI Token output has grown by more than 30 times. It adds that one leading trillion-parameter model has already reached average daily capacity of trillions of high-quality AI Tokens. Revenue in June 2026 alone exceeded the company’s full-year revenue for 2025, according to the article.
This article was originally published by the WeChat public account Touzijie (ID: pedaily2012) and written by Wu Qiong. It was republished by MarsBit.