The race for artificial intelligence supremacy is increasingly tied to the availability of electricity rather than computer chips alone. Rui Ma, an expert in Chinese technology and founder of the media company Tech Buzz China, highlighted this after taking her team around China to get a firsthand look at the AI advancements in the country. Fortune even noted in its headline that AI experts have entertained the idea that the United States is lagging in the AI race because its energy security is weak compared to China.
Beyond Chips and the Importance of Electricity Generation in the Artificial Intelligence Race: China Builds Energy Surplus While America Faces Shortages
The Energy Advantage of China
Advancing AI requires huge amounts of energy. An earlier report indicated that OpenAI consumed about 284 megawatt-hours of energy while training its GPT-3 large language model. The same report also noted that research and development activities related to AI and the utilization of AI applications at Google accounted for 15 percent of its total consumption in 2021. Training and running AI models consume significant amounts of energy.
China is an ideal place to develop and deploy advanced artificial intelligence systems. The energy infrastructure of the country currently enjoys high reserve margins ranging between 80 and 100 percent. This is according to Chinese electricity expert David Fishman. He explained to Fortune that the level of excess capacity allows China to reliably support the rapid construction of hundreds of data centers and provide AI developers with a stable foundation.
The United States electricity grid faces mounting strains by contrast. Forecasts from the U.S. Department of Energy revealed that data centers could account for between 6.7 and 12 percent of national electricity consumption by 2028. This will be a 4.4 percent increase from 2023. It is also worth noting that utilities already struggle with transmission constraints, aging infrastructure, and regional permitting delays that hinder the delivery of reliable capacity.
China is scaling up electricity generation at a striking rate. Reports indicate that it adds electricity demand equivalent to the entire annual electricity consumption of Germany. Investments are distributed across nuclear power, hydropower, and renewable energy sources. It is also interesting to note that its cola plants can be brought back online when necessary. This flexibility ensures that growth in its AI industry does not stall because of electricity shortages.
The Problem With American Energy
In the United States, the situation is markedly different. New data centers are often confronted with significant delays because regional grids cannot provide sufficient supply. Developers are now compelled to move to different locations or factor in geographic locations in their business decisions based on energy availability, while consumer power bills in some regions rise as utilities reallocate resources toward serving large-scale computing facilities.
Several U.S. technology firms have responded by investing directly in energy infrastructure. Big tech companies such as Microsoft, Google, Amazon, Oracle, and Nvidia are constructing private power generation facilities and energy-efficiency measures. Research and development efforts in advanced nuclear reactors are also underway. These strategies highlight the urgency with which the private sector is addressing supply shortages and energy security.
The strategy of China differs because it integrates surplus energy into national planning. Many data centers there remain idle or underutilized. The government is even developing a centralized marketplace to monetize excess power capacity. Although challenges persist, including latency and technical compatibility with foreign systems, the existence of surplus demonstrates the structural advantage China possesses in AI infrastructure scaling.
Observers have underlined the consequences and possible outcomes of these trends. Fishman has noted that the direction of China makes it capable of introducing breakthroughs in AI research. The situation of the United States can only get the country on base. It is also worth noting that Meta Platforms chief executive Mark Zuckerberg has similarly warned that insufficient power plant construction could restrict the ability of American companies to sustain AI growth.
Implications and Key Takeaways
This divergence raises long-term strategic concerns. The U.S. maintains advantages in hardware design and chip expertise. But these may be undermined if electricity cannot keep pace with computational demand. Researchers at SemiAnalysis reported that the CloudMatrix cluster from Huawei has surpassed the GB200 system of Nvidia. This shows how brute force power capacity can compensate for weaker individual hardware efficiency in performance.
Nevertheless, the availability of electricity has become the most critical factor in ongoing efforts and the future of artificial intelligence development. If the United States fails to accelerate grid modernization and transmission expansion, the secure and inexpensive power supply of China may tilt the balance of global AI leadership. The AI race is no longer only about chips. The availability of energy is an integral input in developing and deploying AI systems.
FURTHER READINGS AND REFERENCES
- Baranski, A. 2025. “Energy Consumption of Artificial Intelligence.” Profolus. Available online
- Ennes, J. 11 August 2025. “Big Tech, Power Grids Take Action to Reign in Surging Demand.” Reuters. Available online
- Patel, D., Nishball, D., Xie, M., Zhou, P., Chiam, I., Kourabi, A., Seifel, C., and OLaughin, D. 16 April 2025. “Huawei AI CloudMatrix 384 – China’s Answer to Nvidia GB200 NVL72.” SemiAnalysis. Available online
- Roytburg, E. 14 April 2025. “AI Experts Return From China Stunned: The U.S. Grid is So Weak, the Race May Already Be Over.” Fortune. Available online