In May in Shenzhen, the heat wave hit people. Outside the venue of the National Energy Administration's on-site promotion meeting for artificial intelligence and energy, a war without gunpowder has already begun. At 20:21 on May 25, 2026, the load of Southern Power Grid broke the historical record, reaching 259 million kilowatts. Faced with the early peak of electricity consumption, it is no longer a traditional human wave tactic, but billions of algorithms are running at high speed in the cloud, commanding the precise flow of electricity thousands of miles away. An energy revolution driven by computing power is quietly reshaping China's power supply system.

Liquid heat dissipation in the intelligent computing center

Entering the Nanshan Intelligent Computing Center of Shenzhen Telecom, there was a hint of coolness amidst the buzzing sound of operation. There is no loud air conditioning roar from traditional computer rooms here, as the servers are immersed in non-toxic and harmless fluoride solution. This fully immersed liquid cooling technology has achieved a leap in heat dissipation efficiency, reducing cooling power consumption by over 90%. Faced with the reality that electricity costs account for half of the operation, the center is vigorously procuring 50 million kilowatt hours of green electricity from Yunnan and Guangxi, striving to achieve 100% green electricity supply for new projects by 2026.

AI dispatcher precise control

In the dispatch hall of Southern Power Grid, a large power model known as Big Watt is staring at the screen. It can not only predict load fluctuations under high temperatures 14 days in advance, but also schedule resources in real-time and accurately. Since the beginning of summer in 2026, AI has accurately released 6.6 million kilowatts of new energy storage power, equivalent to the electricity consumption scale of a second tier city. At the pumped storage power station in Guangzhou, AI has become a remote doctor, performing over 10000 online pulse checks on the units every minute, replacing 90% of manual inspections and increasing the daily power generation of a single unit by 11% year-on-year. Keywords: high-tech news and information, computing power empowerment

Buying computing power is as simple as buying electricity

Computing power is becoming a commodity that can be traded like electricity. In the southern region, a new trading platform is emerging. A Shenzhen based artificial intelligence company has released a training requirement of 1 billion words. The system predicts electricity prices through AI models and automatically schedules tasks to Guizhou data centers with lower electricity prices. This new model of computing collaboration allows users to place orders on demand and pay as you go, which not only reduces the computing power cost of enterprises, but also ensures that every kilowatt hour of green electricity is used on the cutting edge.Editor/Gao Xue