AI rewiring how China's energy is managed

To generate a mere five-second, high-definition video out of the digital space — perhaps a lifelike sunset or a virtual city street — an artificial intelligence model does not simply crunch numbers; it guzzles electrons.

Producing those few fleeting seconds of digital reverie requires roughly the same amount of electricity as fully charging 10 smartphones, said Wang Hongzhi, head of the National Energy Administration.

"The AI boom is triggering a ravenous new appetite for electricity," Wang said.

In China, the AI sector is expanding at a breakneck pace. Projected to surpass a valuation of 10 trillion yuan ($1.48 trillion) by the end of the decade, the industry's explosive boom in computing power is driving a precipitous spike in power consumption.

The government is confronting this colossal appetite head-on. Yet, rather than merely throwing more coal into the furnace, Chinese planners are orchestrating an intricate, nationwide symbiosis between the machines that think and the grid that feeds them.

To prevent the AI boom from short-circuiting the nation's climate goals, Beijing is effectively redrawing its infrastructural map, which Wang describes as "powering computing with electricity and promoting electricity with computing".

In the sun-drenched, wind-swept expanses of the western provinces, national computing hubs are being deliberately tethered to colossal new renewable energy bases, while in the densely populated east, planners are nesting distributed computing facilities alongside local microgrids and virtual power plants so that data centers can scavenge power close to the source.

The government is also actively encouraging data centers to plug directly into green power and participate in green electricity trading, said Wang.

Conversely, delay-tolerant tasks — the massive, slow-cooking data-training runs — are nudged to operate during off-peak hours. This essentially uses the data centers as massive, flexible shock absorbers to balance the grid, he said.

But this relationship, Wang said, is far from a one-way street. While the power grid fuels the rise of algorithms, the algorithms are, in turn, rewiring how the nation's energy is managed.

The two industries have entered a fast lane of mutual empowerment, said Lin Boqiang, head of the China Institute for Studies in Energy Policy at Xiamen University.

While AI operations do consume a substantial amount of power, the technology has been acting as the nervous system for an increasingly complex energy landscape — optimizing grid operations, smoothing out renewable energy volatility and ultimately driving broader energy efficiency, Lin said.

In the petroleum sector for example, a large model known as "Kunlun" has accelerated the computational efficiency of oil and gas exploration by more than tenfold, cutting through geological guesswork to optimize drilling strategies, said Wang.

Operated by China National Petroleum Corp, the upgraded platform marks a crucial leap from passive question-answering to "active intelligence", allowing the AI to autonomously plan, dispatch tools, analyze data and execute tasks across the production line.

As the first large model in the domestic energy and chemical industry to achieve large-scale, full-chain application, Kunlun — currently deployed across 152 scenarios — boasts a drilling risk warning system with an accuracy rate exceeding 85 percent, issuing over 300 early warnings in the past six months to prevent accidents, CNPC said.

Ye Xiaoning, a senior engineer at the State Grid Energy Research Institute, said AI is proving to be the perfect antidote to the inherent chaos of renewable energy.

As immense volumes of wind and solar power flood the national grid, the system becomes jittery, subject to the whims of passing clouds and dying breezes, he said.

Here, an AI model dubbed "Yudian" acts as a tireless dispatcher. It predicts weather-driven fluctuations in real time and dynamically adjusts the grid to ensure that no clean energy goes to waste.

This intelligence is also moving from the control room out into the field, graduating from merely offering diagnostic advice to executing autonomous operations.

When a fault occurs in the power distribution network, an AI model called "Guangming Power" steps in like an attending physician. It senses line anomalies, pinpoints the exact location of a failure and coordinates intelligent repairs without waiting for human intervention.

Ultimately, it is a bold blueprint for an engineered feedback loop. The algorithms will consume the sun and the wind to grow smarter, and, in their newfound brilliance, they will figure out how to harness the Earth's elements ever more perfectly.

AI systems need an ocean of electricity to keep expanding. Over the last 10 years, China has built a colossal power grid perfectly scaled to power these machines.

By the end of May, China's total installed power generation capacity reached 4.01 billion kilowatts. To put that figure in perspective, it exceeds the combined power capacities of the United States, the European Union, India, Japan and Russia.

The pace of this expansion is compounding at a dizzying rate. It took the country eight years to double its capacity from 1 billion kW in 2011 to 2 billion kW in 2019. Reaching the 3-billion-kW mark took five years, achieved in April 2024. Adding this latest billion took a mere two years.

Overall, between 2010 and 2025, China's power generation capacity maintained an average annual growth rate of 9.7 percent — eclipsing the equivalent rates of the US (1.7 percent) and the EU (3.2 percent).