首个「AI六小虎」上市,海淀收获「全球大模型第一股」
Sou Hu Cai Jing·2026-01-08 07:01

Core Insights - Beijing Zhiyu AI Technology Co., Ltd. (referred to as "Zhiyu AI") officially listed on the Hong Kong Stock Exchange on January 8, 2026, marking it as the world's first IPO of a large model company [1] - The listing signifies a critical breakthrough for China's large model technology, transitioning from laboratory to capital market, and highlights Haidian's role as a global innovation hub [1][10] - Zhiyu AI's IPO raised approximately HKD 4.173 billion, with a share price set at HKD 116.20, and the company experienced a market capitalization of HKD 55 billion shortly after listing [2][1] Group 1: Haidian's Innovation Ecosystem - Zhiyu AI's journey began in Haidian's academic labs, showcasing the integration of research and industry within this innovation-rich area [5] - The company was founded in 2019, evolving from the AMiner project at Tsinghua University, and has established itself as a leader in the large model sector [6][9] - The success of Zhiyu AI is attributed to Haidian's supportive ecosystem, which includes policies that facilitate technology transfer from research to industry [9][17] Group 2: Financial Performance - Zhiyu AI's revenue growth from 2022 to 2025 shows a compound annual growth rate of 130%, with revenues of CNY 57.4 million, CNY 124.5 million, and CNY 312.4 million respectively [8] - In the first half of 2025, the company reported a revenue of CNY 190.9 million, reflecting a year-on-year increase of 325% [8] Group 3: Haidian's Competitive Advantage - Haidian is home to a significant concentration of AI talent, with 37 universities and a large number of AI experts, which fosters innovation and collaboration between academia and industry [12] - The region has developed a comprehensive industrial ecosystem that spans from chip production to application development, enabling a full-cycle support for hard technology companies [13][12] - Market-driven innovation is supported by diverse application scenarios in Haidian, allowing companies to test and iterate their technologies effectively [14]