How two major developments in geopolitics and technology in the third decade of the 21st century—the escalation of U.S.-China competition and breakthroughs in artificial intelligence technology—are jointly shaping the international landscape. The November 2025 U.S. National Security Strategy emphasizes increasing investment in critical technologies such as artificial intelligence while ensuring energy supply. The Trump administration's second-term energy policy reverses the Biden-era priority on new energy, shifting toward expanded utilization of fossil fuels and nuclear energy. Although lacking a systematic electricity strategy, the contours of U.S. electricity strategy have emerged based on executive orders, congressional legislation, and departmental policies: with the goals of promoting economic prosperity and national security, it supports fossil fuels and nuclear energy, curbs green energy, and enhances grid resilience, with the core driving force being the need to meet the explosive electricity demand of artificial intelligence. Academic research indicates that power supply capacity has become a critical foundation for U.S. artificial intelligence technology and industrial development, with the "AI-energy nexus" relationship becoming increasingly prominent. During the Trump 2.0 period, energy, electricity, and artificial intelligence strategies are deeply interconnected, with government and tech giant cooperation attempting to transform "energy advantage" into "power advantage," using oil and gas leverage to shape the global economic landscape. Analyzing this strategy and its structural constraints helps understand the limitations of U.S. artificial intelligence competition with China, China's advantages in energy infrastructure, and the evolving trends in U.S.-China competition.

I. The Power Supply-Demand Contradiction Facing U.S. Artificial Intelligence Development

This section examines the power supply-demand contradiction facing U.S. artificial intelligence development, analyzed across three dimensions. First, the expansion of AI computing power has led to a sharp surge in electricity demand. Large language model training and inference consume massive amounts of power. The number of U.S. data centers is growing rapidly, and their share of electricity consumption continues to rise, projected to consume 4.6%–9.1% of national electricity generation by 2030. However, net electricity generation growth in the U.S. during the same period is extremely limited, making power generation inadequacy increasingly prominent. Second, U.S. grid infrastructure is outdated and fragile. Most power plants and transmission lines have been operating for excessively long periods. The lengthy planning cycles and gradual growth characteristics of power supply make it difficult to respond to the sharp surge in electricity demand. The contradiction between the highly concentrated electricity consumption characteristics of data centers and the carrying capacity of aging grids is particularly acute in places like Northern Virginia, where grid capacity shortfalls may force companies to consider relocating overseas. Finally, power supply bottlenecks directly constrain corporate expansion. Even when tech giants invest billions in building hyperscale data centers, power connection delays of up to five years have become the biggest obstacle. Energy security has been regarded as a competitive element equally important as semiconductor supply, and power supply capacity is becoming a hard constraint on the continued expansion of the U.S. artificial intelligence industry.

II. U.S. Layout for Supporting Artificial Intelligence Development Through the Power Industry

This section discusses the multiple pathways through which the Trump administration supports artificial intelligence development through power industry layout. In the fossil energy sector, the government has strengthened natural gas and coal power generation through the "Unleashing American Energy" executive order and the "One Big Beautiful Bill," with dozens of new gas-fired power plants approved for construction in 2025. Meanwhile, environmental standards for coal-fired plants have been relaxed to delay retirement, and the Department of Defense has been pushed to sign long-term power purchase agreements with coal plants. Although federal policy suppresses renewable energy, solar power generation has maintained relatively rapid growth due to prior investment plans and power shortages. In nuclear energy revival, the government has signed multiple executive orders requiring nuclear power project deployment within 30 months. Small modular reactors have become the power source of choice for tech giants, with companies such as Amazon and Meta actively investing in nuclear power and signing long-term power purchase agreements. Nuclear fusion power generation has also attracted capital market enthusiasm due to strategic expectations, while the government explores using artificial intelligence to improve nuclear power infrastructure operational efficiency. In grid upgrading, the "AI Action Plan" lists substantially increasing power generation and upgrading the grid as core pillars. The Department of Energy launched the "Powering Acceleration" program, with grid investment reaching $115 billion in 2025, but this remains severely insufficient relative to demand. The government has had to adopt emergency administrative interventions requiring grid operators to prioritize power supply to data centers. In cost-sharing mechanisms, against the background of data center expansion causing sharp increases in residential electricity bills and more than half of states considering legislative restrictions, the Trump administration has pressured tech giants to sign "Electricity Consumer Protection Commitments," requiring them to build their own power generation resources, bear transmission and distribution upgrade costs, and activate backup power sources. This has pushed AI companies to increasingly rely on self-built "shadow grid" systems that are even independent of the public grid. This strategic layout reflects the pragmatic orientation of the U.S. under energy bottleneck constraints, attempting to use fossil fuels to underwrite short-term demand, nuclear energy to guarantee long-term supply, and corporate investment to compensate for insufficient public investment, in order to maintain artificial intelligence industry competitiveness.

III. Constraints Facing U.S. Efforts to Enhance Power Supply Capacity to Support Artificial Intelligence Development

This section analyzes three major constraints facing the Trump administration's efforts to enhance power supply capacity to support artificial intelligence development. First, insufficient supply capacity in the power equipment supply chain. The U.S. and its allies have limited manufacturing capacity in key equipment such as gas turbines, high-voltage direct current cables, and large transformers, with lengthy delivery cycles. Meanwhile, the Trump administration's tariff policies on China further constrain power infrastructure construction, while China, with its comprehensive industrial chain advantages, has become the only country globally capable of rapidly responding to large-scale orders. Second, power industry policy orientation limiting growth space. The Trump administration relies on fossil fuels and nuclear energy rather than renewable energy, yet solar and wind power are the fastest-building and lowest-cost forms of new power generation. The "One Big Beautiful Bill" imposes tariffs on Chinese clean energy equipment and cancels tax credits, which will cause numerous renewable energy projects to stall, significantly pushing up electricity prices and exacerbating the energy bottleneck for artificial intelligence development. Third, a market ecosystem pursuing short-term returns eroding industrial foundations. U.S. capital markets expect returns within three to five years, which severely mismatches the investment recovery periods of over ten years for power projects. Nuclear power and other sectors have seen divergence between stock price speculation and project delays. If tech giants divert massive capital into power infrastructure, it will divert R&D funding, while the "financialization" and "circular financing" models of the artificial intelligence industry contain enormous bubble risks against the background of insufficient power supply. These structural contradictions will delay the integrated development of the U.S. power system and artificial intelligence industry, limiting its competitive capacity against China.

IV. Conclusion

This section analyzes how power supply bottlenecks constrain U.S. artificial intelligence competitiveness, noting that U.S.-China artificial intelligence competition is shifting from a technology race to a contest of comprehensive energy infrastructure capabilities. With its rapid growth in power generation capacity, advantages in renewable energy installed capacity, and systematic layouts such as "East Data West Computing," China has formed a competitive advantage over the U.S. in power infrastructure resilience, providing more stable, green, and low-cost energy security for artificial intelligence development. Meanwhile, U.S. lags in grid resilience and clean energy transition have placed its artificial intelligence industry in the predicament of potentially being forced to become "energy operators."