In the workshop of a transformer production enterprise in Nanchang, Jiangxi, machines roar and workers rush to produce overseas orders. By 2026, the company's on hand orders have approached 700 million yuan, of which exports exceed 600 million yuan, accounting for over 90%. The company leader introduced that the overseas incremental growth is explosive, and the company will continue to expand its coverage in overseas markets, vigorously promote the research and industrialization of products with high voltage levels above 500 (kV), and occupy the high-end market with technological leadership. In the production workshop of a technology company in Ganzhou, Jiangxi, transformer cores are neatly arranged and automated equipment is running at high speed. The company's main products are transformers and complete sets of high and low voltage equipment, which are exported to Africa, America and other places. In order to keep up with the growth of orders, they are expanding. Outside the factory building of an electrical company in Pingdingshan, Henan Province, transformers awaiting shipment are neatly arranged; In the 5G intelligent workshop, precision components are being produced intensively. Since the beginning of this year, an average of over 30 transformers have been shipped from here to markets such as Russia, Vietnam, and Mexico every day.
Why did the global order surge for power equipment such as transformers suddenly emerge? The answer lies in the huge demand for electricity generated by the development of artificial intelligence. Data shows that the electricity load of a large-scale AI data center has exceeded 1 gigawatt (1GW), equivalent to the peak summer electricity load of a medium-sized city. In addition, AI big models are accelerating from the "training" stage to the "inference" stage, which means that their electricity consumption is shifting from one-time input to continuous consumption. Professor Ding Zhaohao from the School of Electrical and Electronic Engineering at North China Electric Power University introduced that as large models become more and more useful, the energy consumption of inference will become more and more significant, and the proportion will also increase. As a result, the demand for electricity in AI data centers will naturally increase.