CITIC SEC: AI infrastructure accelerates demand explosion, leading AIDC companies may welcome a demand explosion

CITIC SEC released a research report stating that since 2025, domestic AI large models have accelerated their technological transition through algorithm and engineering innovation, with the low-cost DeepSeek -R1 inference capability rivalling OpenAI o1. Alibaba has released the billion-level MoE architecture model Qwen2.5-Max, while Douyin has improved its model capabilities based on ByteDance's ecological advantages. Tencent's hybrid large model has been deeply integrated into business scenarios such as WeChat. The progress of AI models and the explosion of AI demand have driven the overall increase in Capex for domestic cloud giants. AIDC is the core base of AI computing power. Currently, the scarcity of AIDC in first-tier cities is evident, and companies with a large reserve of high-quality resources are expected to experience a surge in demand. Key points from CITIC SEC: Introduction of AIDC: Core base of AI computing power. AIDC is a technology-based heavy asset leasing service, with initial Capex investment mainly including land and data center investments, as well as the purchase of power and temperature control equipment, with operating costs mainly consisting of depreciation and electricity fees. The construction of AIDC usually takes 1-3 years, and customer onboarding typically takes 6-18 months. Top manufacturers' IDCs can achieve over 90% onboarding rate, which can result in over 50% gross profit margin and over 30% net profit margin. Since 2024, the industry fundamentals have significantly improved. The emergence of high-quality AI models in China and the substantial increase in Capex of domestic cloud giants have ushered in a new era of domestic computing power infrastructure. The data center model is transitioning from IDC to AIDC, with high power and campus-level becoming the mainstream development in the industry. Leading AIDC companies are aggressively expanding their footprint, continuously winning large orders, and driving both onboarding rates and rental prices up. In the context of rapidly rising industry demand, AIDC market performance is soaring, and the continued growth of domestic AI infrastructure is promising, with AIDC expected to benefit significantly. Industry Analysis: Accelerated AI infrastructure construction leads to an explosion in demand, with core areas facing supply shortages. 1) Demand: The significant increase in AI investment by domestic cloud giants has contributed to a billion-yuan incremental market for AIDC. The AI capabilities of domestic cloud giants such as Alibaba, ByteDance, and Tencent have rapidly improved since 2024, with the daily active users of these models growing exponentially, driving a significant increase in Capex. Based on estimates, if 50% of a 100 billion Capex by cloud giants is used for AI chip procurement, it would correspond to a 700MW incremental demand for AIDC, equivalent to approximately 7 billion yuan in incremental AIDC orders. With the backdrop of the arms race in AI computing power domestically, it is expected that nearly 30 billion yuan of incremental AIDC demand will be generated in China by 2025. 2) Supply: Strict power consumption control in core regions, with policy support for smart computing construction. Due to strict power consumption control in first-tier and surrounding cities, there is still a significant imbalance between demand and supply in the IDC market in these cities. The overall layout of the Chinese IDC market is concentrated in core first-tier cities (Beijing, Shanghai, Guangzhou, Shenzhen), with uneven regional distribution. However, under the support of policies such as "East-West Big Data Corridor," several provinces and cities in China are encouraging the development of smart computing centers. The addition of core IDCs and smart computing nodes will be the main force in the market supply. 3) Competitive landscape: Companies with high-quality resources and customer binding advantages stand out. Third-party IDC companies are the main players in the IDC industry, with telecom IDCs and self-built IDCs by cloud providers playing a supporting role. Given the scarcity of power consumption indicators in first-tier cities, the massive upfront investment in IDC, and the professional capabilities required for IDC operation and maintenance, it is expected that industry orders will primarily tilt towards leading third-party IDC companies. Key Companies: Leading AIDC companies expected to benefit significantly. According to the industry estimates, companies with gigawatt-level energy reserves; additionally, some companies have resources reserves in the hundreds of megawatts. Beyond traditional IDC business, companies are further expanding their revenue boundaries. The positive improvement in AIDC industry demand, coupled with sufficient resource reserves and core customer binding, is expected to significantly benefit companies. Risk Factors: Risks of AI technology development and application not meeting expectations; risks of macroeconomic fluctuations; risks of intensified industry competition; risks of policy support falling short of expectations; risks of traffic growth not meeting expectations; risks of cloud providers' Capex growth falling short of expectations; risks of related companies' IDC cabinet expansion not meeting expectations; risks of related companies' cabinet onboarding rates not meeting expectations.