Qujing Technology closed a Series B of more than $195 million, taking its 2026 fundraising to about $360 million. Several participants in the round are also disclosed partners in the Beijing-based company’s regional buildout of infrastructure for running trained AI models.
The six co-leads included Xiaoshan Capital, ICBC Investment and Zhongguancun Science City. Qujing said the proceeds would expand computing capacity, upgrade its ATaaS platform, develop inference technology and products, and improve commercial delivery. Figures are in US dollars, converted at October 23, 2026 reference rates.
Qujing turns chips, memory, networking and software into the computing service that produces a trained model’s responses. It either secures computing resources and sells that output to model developers, internet platforms, AI applications and large enterprises, or helps customers with their own resources plan, integrate and operate dedicated systems it calls token factories.
The investor list overlaps with Qujing’s project map. SenseTime and Wangsu Technology participated in the round and are also disclosed project partners. Another co-lead, Qianjiang Century City, is associated with a planned token factory in Hangzhou designed around roughly 10,000 accelerator cards.
Qujing concentrates its engineering on a smaller set of heavily optimized models rather than maximizing compatibility. The company and chipmaker Moore Threads said in September that a mixed-hardware system had entered production for the official service of an unnamed leading model developer, although they did not identify the customer or independently substantiate the claimed cost advantage.
Qujing says its daily output rose more than 100-fold over six months and that 2026 revenue to date is nearly 10 times its full-year 2025 figure. It did not provide absolute revenue, gross margin or per-token economics with the announcement. A July China Business Journal analysis of SiliconFlow’s 2025 filing found that computing resources represented 86.9% of revenue costs and its overall gross margin was negative 24%, showing why higher output alone does not establish profitability.