Executive Summary
Key takeaways from this analysis:
- AI development is increasingly constrained by electricity supply rather than computing power, shifting the competitive landscape.
- Global power infrastructure investments are accelerating, with grid equipment sectors experiencing significant capital inflows.
- Chinese grid equipment ETF (159326) has attracted over 5.27 billion yuan in 20 days, reaching record highs as investors position for this trend.
- The transition from chip shortages to power shortages signals a new investment super cycle in energy and grid technology.
- Technological innovations like 800V DC architecture and solid-state transformers are emerging as critical solutions to power constraints.
The Unfolding Energy Challenge in AI Development
Microsoft CEO Satya Nadella (萨提亚·纳德拉) recently dropped a bombshell during an interview with OpenAI CEO Sam Altman (萨姆·奥特曼), stating that the biggest constraint in artificial intelligence development isn’t semiconductor availability but electricity supply. This revelation has sent shockwaves through global markets, particularly impacting Chinese equity markets where grid equipment stocks have surged in response. The power competition is intensifying as AI’s computational demands outpace energy infrastructure capabilities worldwide.
This power competition represents a fundamental shift in how investors and technologists view the AI ecosystem. Where once the conversation centered on GPU availability and computing architecture, the discourse has now expanded to include megawatt requirements, grid stability, and energy transmission efficiency. The recognition that data centers require massive, reliable power sources has triggered a reevaluation of energy infrastructure investments globally, with particular implications for Chinese manufacturers and technology companies positioned to benefit from this new paradigm.
Immediate Market Reactions
Following Nadella’s comments, A-share grid equipment components saw early session gains, with Shenma Power (神马电力) hitting new highs, Moen Electric (摩恩电气), Zhongneng Electric (中能电气), and Guancheng New Materials (冠城新材) reaching daily limits, and Siyuan Electric (思源电气) rising over 5%. The specialized Grid Equipment ETF (159326) briefly surged more than 2.5%, reflecting investor recognition of this emerging power competition dynamic. The ETF has attracted approximately 5.27 billion yuan over 20 trading days, with its scale reaching historical highs.
This market movement underscores a broader realization that AI’s advancement depends not just on algorithmic sophistication but on physical infrastructure capable of delivering unprecedented amounts of electricity. The power competition has moved from background concern to front-line investment thesis, with capital rapidly reallocating to companies positioned to address this bottleneck. Year-to-date, the Grid Equipment ETF has delivered impressive returns of 29.7%, demonstrating the market’s anticipation of sustained demand growth in this sector.
The Resurgent Power Shortage Crisis
Global electricity demand is entering a new growth phase after years of relative stability, driven primarily by exponential increases in AI computational requirements. This power competition is creating structural imbalances in energy markets worldwide. On Monday, Chinese markets experienced adjusted trading with the Shanghai Composite Index falling 0.41%, the Shenzhen Component Index declining 1.71%, and the ChiNext Index dropping 1.96%. The Beijing Stock Exchange 50 Index fell 2.45%, with total transaction volume across Shanghai, Shenzhen, and Beijing markets reaching 1,938.4 billion yuan, shrinking by 194.5 billion yuan from the previous day.
Sector performance highlighted this shifting landscape, with Fujian-related stocks, grid equipment, banking, ice and snow industry, film and cinema, and paper sectors leading gains. Meanwhile, non-ferrous metals, PEEK materials, weight-loss drugs, wind power equipment, robotics, and battery sectors declined. This sector rotation reflects investor recognition that the power competition will disproportionately benefit companies involved in electricity transmission and distribution rather than traditional technology components.
Nuclear Fusion’s Emerging Role
Controlled nuclear fusion represents a potential long-term solution to the power competition, with global research intensifying. According to International Atomic Energy Agency projections, the global nuclear fusion market could exceed $40 trillion by 2050. In early October, significant progress was reported at the Compact Fusion Energy Experimental Device (BEST) construction site in Hefei Future Science City in Anhui Province, where the first key component of the host machine—the Dewar base—was successfully installed.
This device began assembly in May and is expected to be completed by 2027, potentially demonstrating fusion power generation for the first time worldwide. Controlled nuclear fusion is regarded as the ultimate solution in the energy field, with commercialization potentially driving a technological revolution comparable to the advent of steam engines and electrification. The industrialization process of its main components warrants close attention from investors monitoring the evolving power competition landscape.
AI Competition Transforms into Power Competition
The core factor driving electricity shortages in the United States stems from AI computing explosions, with data center负荷快速增长 rapidly pushing North American power demand into a new growth cycle. On October 27, OpenAI submitted an 11-page proposal to the Trump administration’s White House Office of Science and Technology Policy (OSTP), urging the federal government to significantly increase investment in power infrastructure. OpenAI explicitly recommended setting a target of adding 100 gigawatts (GW) of generating capacity annually—nearly double the 51 GW added last year.
This power competition is reshaping national energy policies and corporate strategies alike. According to U.S. Energy Information Administration (EIA) predictions, annual U.S. electricity consumption will increase in 2025 and 2026, surpassing the record high set in 2024. This growth contrasts sharply with the relatively flat electricity demand trend from the mid-2000s to early 2020s. Recent and projected electricity consumption growth primarily originates from the commercial sector (including data centers) and industrial sector (including manufacturing enterprises).
Quantifying the Power Gap
Projections vary on the scale of the coming power competition:
- Data centers accounted for 4.2% of total U.S. electricity consumption in 2024
- McKinsey predicts data center electricity usage will reach 8.1% of total U.S. consumption by 2030
- The U.S. Electric Power Research Institute (EPRI) forecasts approximately 9% of U.S. generation will power data centers by 2030
- Boston Consulting Group (BCG) offers the most aggressive projection, suggesting data centers could consume 25% of U.S. electricity by 2030
From a supply perspective, while overall U.S. grid installed capacity has grown over the past decade, the increase has been modest and primarily driven by wind, solar, and storage, while high-quality power sources like hydropower, thermal power, and nuclear power have generally declined. Aging equipment (some over 40 years old) has led to insufficient transmission capacity. Additionally, lengthy and rigid grid connection processes historically averaging over three years make it difficult for new installations to quickly respond to power system demands.
Technological Responses to Power Constraints
OpenAI’s internal analysis suggests that over the next five years, the equivalent of 20% of the current U.S. technical workforce (such as electricians and construction workers) will need to be mobilized for data center and power facility construction. If power expansion cannot keep pace with demand, the result will be electricity shortages and price increases, potentially impacting economic stability.
NVIDIA is pushing for data center power infrastructure to transition to 800 VDC starting in 2027, aiming to support IT racks with power densities of 1MW and above. This architecture achieves more efficient power transmission by increasing voltage levels and integrates solid-state transformer (SST) technology to optimize system design. Solid-state transformers are intelligent power devices that achieve power conversion, electrical isolation, power regulation, and control through power electronic conversion technology and high-frequency electromagnetic induction principles.
In the AIDC field, rapidly increasing computing density places higher demands on power supply systems, making SST’s high power density and flexible AC-DC conversion characteristics an ideal choice. Adopting SST spatial architecture can save approximately 5% of peak demand and reduce power gaps by 5% to 10%. Beyond NVIDIA, companies like Alibaba, Tencent, Google, and Microsoft are actively planning data center power innovations.
Investment Opportunities in Grid Equipment
The global power competition is creating significant investment opportunities across the grid equipment value chain. According to the latest data from the General Administration of Customs, China’s transformer cumulative export value from January to September 2025 reached 35.092 billion yuan, a year-on-year increase of 52.73%, maintaining rapid growth since 2023. Chinese companies are actively expanding overseas to seek incremental growth, with the proportion of overseas business rapidly increasing in recent years.
For example, Siyuan Electric (思源电气) saw overseas revenue account for 33.68% of total revenue in the first half of the year, reaching a recent high, with overseas gross profit margin (35%) driving overall profitability improvement. This international expansion demonstrates the global nature of the power competition and the competitive advantages Chinese manufacturers have developed in grid equipment production.
ETF as Strategic Investment Vehicle
For investors seeking exposure to this power competition theme, the Grid Equipment ETF (159326) offers a diversified approach. As the only ETF in the market tracking the CSI Grid Equipment Theme Index, its components include ultra-high voltage weighting up to 63.35%, charging pile content of 61.17%, nuclear fusion content of 15.3%—all the highest in the market. Virtual power plant content reaches 42.57%, the highest among similar indices.
The index weighting stocks include industry leaders such as State Grid NARI (国电南瑞), TBEA (特变电工), Siyuan Electric (思源电气), and Zhongtian Technology (中天科技). Notably, as the only grid equipment ETF in the market, it has recently gained fund favor, with scale growth exceeding 400% in the past month, hitting a record high. This concentrated exposure to the power competition theme makes it an efficient vehicle for investors without specialized knowledge of individual grid equipment companies.
Global Implications and Strategic Positioning
As the global interest rate reduction cycle begins and AI computing power multiplies electricity load demand, the world’s power grids will enter a construction boom in the coming years. The power competition extends beyond the United States, with global grid investment currently in a relatively prosperous cycle. Chinese transformer export data confirms this trend, with 52.73% year-on-year growth in the first three quarters of 2025 indicating robust international demand.
Domestic companies切入布局 mainly come from three categories: traditional power equipment, AI power supplies, and optical storage and charging companies, each with distinct advantages. Thoroughly researching each track presents challenges for ordinary investors, making thematic ETFs an accessible alternative. The power competition necessitates a fundamental rethinking of energy paradigms, whether through nuclear energy innovations or 800V DC architecture upgrades.
SST Supply Chain Dynamics
In the SST supply chain, international giants dominate the high end, while domestic companies are rapidly catching up. Companies like Delta, Eaton, and Vertiv leverage first-mover advantages and technological accumulation to occupy premium markets in AIDC and industrial fields, providing one-stop solutions from SST to complete power distribution. Domestic companies such as Jinpan Technology (金盘科技), Sifang Company (四方股份), Sungrow Power (阳光电源), and China XD Electric (中国西电) achieve breakthroughs in specific application scenarios through differentiated competition.
Among them, Jinpan Technology has deeply cultivated overseas data center markets, obtaining benchmark customers like NVIDIA (Amazon, Microsoft). Sifang Company participates in China’s East Data West Computing project, promoting SST applications in new power distribution networks. This bifurcated supply chain approach allows investors multiple entry points into the power competition theme, depending on risk tolerance and geographic focus.
Synthesizing the Power Competition Landscape
The rapid development of AI and sharp increase in computing demand create urgent needs for further improvement in efficiency and space savings in AI data center power supply and distribution systems. Whether through energy black technologies like nuclear power or upgrades to 800V DC architecture, the essential transformation involves paradigm reconstruction of traditional energy systems. How to efficiently, economically, and reliably power AI clusters—these electricity-devouring monsters—has become the key problem for maintaining rapid computing growth in the AI era.
The power competition represents both a challenge and opportunity for global markets. Chinese companies stand to benefit significantly from this trend, given their manufacturing scale and technological advancements in grid equipment. Investors should monitor developments in power infrastructure spending, technological innovations in electricity transmission, and policy initiatives aimed at addressing electricity shortages. The Grid Equipment ETF (159326) provides a straightforward vehicle for gaining exposure to this theme, particularly for those who missed earlier opportunities in core AI technology tracks.
As Satya Nadella’s insight makes clear, the future of AI advancement depends not on how many chips we can produce, but on how much electricity we can reliably deliver. This fundamental shift in constraints—from semiconductors to electrons—will define the next chapter of technological progress and create winners across the energy and technology sectors. For sophisticated investors, understanding and positioning for this power competition may prove one of the most consequential investment decisions of the coming decade.