[Trend Report] Global AI Development Drives Surge in Electricity Demand The Power Industry Is Expected to See a Revaluation

The three major Chinese A-share indices showed mixed performance today. By the close, the Shanghai Composite rose 0.39%, closing at 4,162.88 points; the Shenzhen Component declined 0.06%, ending at 14,495.09 points; and the ChiNext Index fell 1.04%, closing at 3,310.30 points. The combined trading volume of the Shanghai, Shenzhen, and Beijing markets reached 2.51 trillion yuan, a slight decrease of 50.4 billion yuan from yesterday. Industry sectors mostly gained, with small metals, rare earths, precious metals, energy metals, non-ferrous metals, coal, and power sectors leading gains. Glass fiber, components, and electronic chemicals sectors saw the biggest declines. On individual stocks, over 3,200 stocks rose, with nearly 100 hitting the daily limit-up.

Domestically, according to OpenRouter, the world’s largest AI model API aggregation platform, from February 9 to 15, China’s AI models recorded 41.2 trillion tokens in usage, surpassing the US models’ 29.4 trillion tokens for the first time during the same period. The following week, February 16 to 22, China’s weekly token usage surged further to 51.6 trillion, a 127% increase over three weeks, while US model usage fell to 27 trillion tokens. Data shows that electricity is the “fuel” and main cost for producing AI tokens. Electricity costs account for 60-70% of operating expenses for large AI models, far exceeding hardware depreciation and network costs. Some analysts believe that tokens are derivatives of electricity, and that the overseas expansion of tokens essentially reflects the export of electricity.

Internationally, under AI’s pressure, North America faces increasing electricity shortages. Trump plans to convene tech giants like Amazon, Meta, Microsoft, and Google next week to sign commitments ensuring these companies bear the costs of their energy-intensive data centers, further confirming that rising data center demand continues to drive North American power shortages. Estimates suggest that from 2025 to 2028, US electricity demand driven by AI will grow at a CAGR of about 55%, with over 150 GW of additional capacity needed in the next three years, creating substantial electricity demand.

CICC states that China’s power infrastructure is well-developed, and domestic computing power can leverage low-cost electricity to deliver cross-border value, promoting domestic power consumption and equipment demand. Changjiang Securities notes that by 2026, electricity consumption growth may rebound to 5.04%, with new loads from AI data centers and industrial electrification, potentially marking a turning point in the supply-demand and profitability of green electricity.

CICC: Multiple Factors Drive Demand for Power Equipment

On one hand, China’s mature power infrastructure and low-cost domestic computing power facilitate cross-border electricity value transfer, boosting domestic power consumption and equipment demand. On the other hand, domestic model providers deploying cloud services overseas are expected to benefit from deepening overseas markets, especially those with overseas capacity and stable partnerships with internet companies.

Changjiang Securities: Green Power Supply-Demand and Profitability May Reach Turning Point

By 2025, China’s power system will face a structural contradiction of “ample electricity supply but tight capacity”: thermal power generation will see negative growth for the first time in a decade, yet peak capacity demand and regulation pressures will continue to rise, pushing for high levels of new thermal capacity. Meanwhile, electricity consumption growth is expected to rebound to 5.04% in 2026, with additional loads from AI data centers and industrial electrification, potentially marking a turning point in green power supply-demand and profitability. Policy-wise, the National Development and Reform Commission has proposed establishing a reliable capacity compensation mechanism. Once implemented, the nationwide average capacity electricity price in 2026 is expected to increase by 0.041–0.050 yuan per kWh, further strengthening the “turnaround logic” for the green power sector.

Zhongyuan Securities: Suggest a Dumbbell Strategy for Power-Related Assets

From a medium- to long-term investment perspective, it is recommended to adopt a dumbbell approach: on the defensive side, focus on stability and shareholder returns, targeting large hydropower companies and high-dividend thermal power firms; on the offensive side, explore opportunities in virtual power plants, controllable nuclear fusion, and other themes.

Yingda Securities: Power Market Reform Enters a New Stage

In 2025, China achieved several historic breakthroughs in renewable energy development, with the main power supply gradually shifting to new energy sources. Newly installed renewable capacity reached 452 million kW, a 21% increase year-over-year, accounting for 83% of all new power capacity, with wind and solar surpassing thermal power in cumulative installed capacity for the first time. The construction of new power systems continues, and by the end of 2025, China’s energy storage projects had a total installed capacity of 54% higher than the previous year, with new storage capacity up 85%. A unified national electricity market top-level design has been introduced, marking a new phase of comprehensive, system-wide reform aimed at standardizing market rules and making the multi-dimensional value of electricity more explicit in the market.

Fangzheng Securities: Industry High Growth Already Reflected in Performance

Looking ahead to 2026, the AI data center industry is expected to remain highly prosperous. Major internet companies domestically and abroad have announced capital expenditure plans with high growth expectations. Meanwhile, leading overseas power equipment manufacturers are showing strong performance, indicating that the industry’s high growth is already evident in financial results.

AVIC Securities: Future HVDC and SST Projects Likely to Gradually Scale Up

The global AI investment boom is rapidly boosting the power infrastructure demand for AIDC (AI Data Centers). According to IEA forecasts, from 2025 to 2030, global IT loads will see a cumulative increase of 106 GW. Due to differences in grid structures and expansion difficulties across regions, overseas AIDC projects tend to adopt BYOG (Bring Your Own Generation) models, mainly using gas turbines, SOFCs, etc., for faster deployment. Whether using grid power or self-generation, distribution systems are critical for ensuring efficient and stable power supply. Upgrades in computing chips and servers are driving trends toward reducing losses, minimizing transformer stages, and better handling power fluctuations. Currently, most AIDC distribution systems use UPS (Uninterruptible Power Supply) solutions, but future projects are expected to see increased adoption of HVDC and SST schemes.

(This article does not constitute investment advice. Investors operate at their own risk. The market is risky; please invest cautiously.)

(Source: Dongfang Caifu Research Center)