Yesterday (the 17th), the 2026 World Artificial Intelligence Conference commenced in Shanghai. During the sub-forum titled "Mutual Empowerment of Artificial Intelligence and Energy," the leader of the National Energy Administration announced that China is set to expedite the development of a new energy system to bolster the eco-friendly progression of artificial intelligence.

It has been reported that within the "15th Five-Year Plan" timeframe, inference loads are projected to outstrip training loads, emerging as the main consumers of computing power. Traits such as sensitivity to latency, significant load variability, and concentrated demand necessitate a shift in computing power and electricity cooperation from mere "spatial collaboration" to "multi-dimensional collaboration" between computing power loads and the novel power system. This transformation mandates a coordinated strategy for the integrated planning of essential components, including intelligent computing power, large-scale models, and corpus resources.

Wang Hongzhi, Director of the National Energy Administration, highlighted that approximately 40% of the total electricity consumption nationwide is derived from green power. With a plentiful and high-quality supply of green energy, the transition of green computing power from theoretical concepts to practical applications, and from laboratory settings to real-world scenarios, is being propelled. This ensures the efficient utilization of every kilowatt-hour of electricity and guarantees that every Token adheres to green principles.