中国的智算中心布局

Global Data Center Landscape - The global data center construction is expected to enter a sustained growth cycle, with an estimated additional capacity of 47GW from 2023 to 2026, increasing from a base of 49GW in 2023 to over 96GW by 2026. Approximately 85% of this new capacity (40GW) will be directed towards intelligent computing centers (IDC), indicating a surge in demand for AI computing power [1] - North America will dominate the market, expected to absorb 65%-70% of the new capacity (approximately 30.55-32.9GW), establishing itself as the core hub for global computing infrastructure. China's market share is about 15% (7.05GW), with Europe and Southeast Asia contributing around 10% (4.7GW) [1] Chinese Market - In China's data center market, the compound annual growth rates for traditional general computing (通算) and artificial intelligence computing (智算) are 3%-4% and over 20%, respectively. The current market shows an oversupply in general computing facilities, with core node cabinet utilization rates between 40%-50% [2] - The majority of China's intelligent computing centers are located in the central and western regions, with significant hubs in Ningxia, Inner Mongolia, Gansu, Chongqing, and Guizhou, each hosting large-scale computing networks exceeding 30,000 PF. The largest center in the eastern coastal region is in the Beijing-Tianjin-Hebei area, with a capacity of 22,800 PF [2] Investment Landscape - The investment landscape is characterized by several key players: 1. Government and urban investment platforms account for about 35%, primarily responsible for land supply, energy support, and project coordination, typically using a "design-procurement-construction" model [4] 2. Telecom operators (China Telecom, China Mobile, China Unicom) represent about 20%, leveraging collective procurement systems to reduce costs, separating equipment procurement from construction [4] 3. Internet cloud vendors, including major players like BAT, make up approximately 40%, possessing significant influence over equipment selection and employing a mixed model of "general contracting + specialized subcontracting" [4] 4. Other entities, including research institutions and self-built enterprises, account for about 5% [4]