China Merchants Securities has released a research report stating that the export of tokens is driving the expansion of computing power, which in turn is leading to a sharp increase in electricity demand. The firm estimates that by 2030, electricity consumption by data centers could account for approximately 4% of China's total societal electricity usage. Across economic cycles, utility companies characterized by heavy assets and low obsolescence (HALO) are experiencing a value resurgence. As a typical HALO sector, utilities are expected to become a reliable choice for capital allocation amid an environment of AI disruption and geopolitical instability. The main viewpoints of China Merchants Securities are as follows:

The rapid growth of AI computing power is causing a surge in electricity demand from data centers. Data from the China Academy of Information and Communications Technology shows that from 2019 to 2024, electricity consumption by data centers in China grew from 82.4 billion kWh to 166.0 billion kWh, with a compound annual growth rate of 15.0%. Although the proportion of data center electricity consumption relative to total societal electricity usage is currently low—only 1.7% in 2024—its growth rate is substantial. Concurrently, the accelerated overseas expansion of domestic large AI models means that an increasing amount of inference for overseas data is being processed in domestic data centers. Data from OpenRouter, the world's largest AI model API aggregation platform, shows that from February 16 to 22, the total token consumption of the top ten models on the platform was approximately 8.7 trillion, with Chinese models accounting for 5.3 trillion tokens, or 61% of the total. The export of tokens essentially represents the cross-border delivery of electricity value, which is expected to further drive up electricity consumption in data centers.

By 2030, the proportion of electricity consumption by data centers is expected to reach around 4%. According to IDC forecasts, China's intelligent computing power (FP16 precision) will grow from 1,037.3 EFLOPS in 2025 to 2,781.9 EFLOPS in 2028, with an average annual growth rate of 38.9%. To study the impact of computing power expansion on electricity consumption, the firm constructed a model for calculating data center electricity usage based on power consumption per token, while also accounting for the continuous decline in Power Usage Effectiveness (PUE). Assuming average annual growth rates of 25%, 35%, and 45% for China's intelligent computing power, by 2030, the country's intelligent computing power would reach 3,166 EFLOPS, 4,651 EFLOPS, and 6,649 EFLOPS, respectively. Correspondingly, electricity consumption by data centers would reach 390.7 billion kWh, 574.1 billion kWh, and 820.6 billion kWh. Assuming an average annual growth rate of 5.2% for China's total societal electricity consumption from 2026 to 2030, under these different computing power growth scenarios, the proportion of data center electricity consumption in 2030 would reach 2.9%, 4.3%, and 6.1%, respectively.

The operation of data centers relies on stable and sufficient power supply, while the "Dual Carbon" goals impose requirements for clean, low-carbon electricity. In 2023, the National Development and Reform Commission issued the "Implementation Opinions on Deepening the 'East Data, West Computing' Project and Accelerating the Construction of a National Integrated Computing Power Network," proposing that by the end of 2025, newly built data centers in national hub nodes should source over 80% of their electricity from green power. Under the "Dual Carbon" targets, data centers are expected to become a significant force in promoting the consumption of renewable energy. In recent years, the green electricity trading market has become increasingly mature, with green power premiums gradually returning to a reasonable range. By securing green power resources through long-term power purchase agreements, data centers can effectively hedge against future price volatility of fossil fuels, achieving a win-win situation for both computing cost efficiency and green value.

Across economic cycles, utility companies with heavy assets and low obsolescence (HALO) are witnessing a value resurgence. In the AI era, technological disruptions at the software level are constant. However, HALO assets are tangible, with high entry barriers, stable business models, predictable cash flows, and resilience to AI disruption, offering defensive value. The utilities sector is a typical HALO industry and is poised to be a definitive choice for capital in an environment characterized by AI disruption and geopolitical turmoil.

Investment recommendation: Maintain a positive outlook on traditional power leaders such as China Yangtze Power and China National Power. It is also advised to pay attention to quality regional power companies including Anhui Energy, Fujian Funeng, Shenergy, Guiguan Electric Power, and Fuling Electric Power.

Risk warnings: AI computing power demand may fall short of expectations; policy support might be weaker than anticipated; macroeconomic fluctuations pose a risk.