招股价144.6港元，中国首家通用GPU企业天数智芯启动招股

Group 1 - TianShu ZhiXin officially launched its IPO on December 30, with plans to issue 25,431,800 H-shares at an offering price of HKD 144.60 per share, aiming to raise approximately HKD 3.7 billion and achieve a market capitalization of HKD 35.442 billion [2] - The company has established a diverse and collaborative shareholder matrix, attracting 18 top-tier investors, including ZTE Corporation (Hong Kong), UBS AM Singapore, and Sequoia China, indicating strong capital recognition [2] - The company is the first in China to achieve mass production of both training and inference general-purpose GPUs, with its "TianGai" training GPUs and "ZhiKai" inference GPUs already in large-scale delivery [3] Group 2 - From 2022 to 2024, the number of customers served by the company increased from 22 to 181, with over 290 customers having completed more than 900 actual deployments by June 30, 2025, demonstrating significant commercial success [3] - The company's shipment volume is projected to double from 7,800 units in 2022 to 16,800 units in 2024, with 15,700 units expected in the first half of 2025 [3] - Revenue for the years 2022 to 2024 is expected to reach CNY 189 million, CNY 289 million, and CNY 540 million, respectively, with a compound annual growth rate of 68.8%, and a revenue of CNY 324 million in the first half of 2025, reflecting a year-on-year growth of 64.2% [3] Group 3 - The valuation of domestic general-purpose GPUs is being restructured, focusing on sustainable technological assets and delivery capabilities, rather than short-term profit [4] - TianShu ZhiXin is recognized as an innovator in the domestic general-purpose GPU sector, effectively integrating performance and ecological compatibility, and establishing itself as a preferred choice in the market [4] - The IPO represents a significant milestone for China's AI infrastructure, providing a reliable and scalable domestic force in the general-purpose GPU industry, with expectations for continued leadership and long-term value [4]