China’s push to clean up its energy system is increasingly being driven by artificial intelligence, not as a distant policy idea but as a practical tool embedded in everyday operations. From renewable-powered factories to citywide power networks, AI is starting to shape how electricity is produced, balanced, and consumed across the country.
A clear example can be found in Chifeng, a city in northern China, where a factory produces hydrogen and ammonia using electricity generated entirely from nearby wind and solar farms. Unlike conventional plants connected to the national grid, the facility operates as a closed system. That design avoids fossil fuels but comes with a challenge: renewable energy output rises and falls with the weather.
To manage that variability, the plant relies on an AI-driven control system developed by its owner, Envision. Instead of following fixed production schedules, the software continuously adjusts output based on real-time wind and solar conditions. When renewable power surges, production increases automatically. When conditions weaken, electricity use is scaled back to maintain stability. Envision’s chief engineer for hydrogen energy, Zhang Jian, described the system as acting like a conductor, coordinating supply and demand in real time.
Projects like this sit at the heart of China’s plans to expand hydrogen and ammonia, fuels viewed as key to cutting emissions in hard-to-abate sectors such as steelmaking and shipping. They also illustrate a broader strategy: using AI to manage the growing complexity of an energy system that is rapidly adding renewable capacity.
China already installs more wind and solar power than any other country, but integrating that energy efficiently remains difficult. AI is increasingly seen as a way to make the grid more flexible and responsive. Researchers say the technology can help forecast electricity supply and demand, track emissions, and optimize industrial energy use. At the same time, they caution that AI itself is becoming a major consumer of power, largely due to the rapid expansion of data centres.
That tension is now shaping policy. In September, Beijing unveiled an “AI plus energy” strategy that formally links artificial intelligence with the energy sector. The plan calls for the development of multiple specialized AI models tailored to tasks such as grid management, power generation, and industrial operations. By 2027, authorities aim to launch dozens of pilot projects and test AI across more than 100 use cases. Within three years after that, China wants to reach what it describes as a world-leading level of AI integration in energy.
According to analysts, the emphasis is on highly targeted tools rather than general-purpose AI. Systems are being designed for specific roles, such as managing wind farms, nuclear plants, or grid balancing. This approach differs from that of the United States, where much of the investment has focused on developing advanced large-language models.
One area where AI is already delivering results is demand forecasting. Power grids must keep supply and demand perfectly balanced to avoid outages. More accurate forecasts of renewable output and electricity use allow operators to store energy in batteries, smooth peaks in demand, and reduce reliance on coal-fired backup plants.
Shanghai offers an early example. The city has launched a virtual power plant that links dozens of operators, including data centres, building management systems, and electric vehicle chargers, into a single coordinated network. During a trial last August, the system cut peak demand by more than 160 megawatts, roughly equivalent to the output of a small coal-fired power station.
AI is also being explored as a tool to strengthen China’s national carbon market, which covers more than 3,000 companies in emissions-intensive industries such as power generation, steel, cement, and aluminium. Together, these sectors account for over 60% of the country’s carbon emissions. Analysts say AI could help verify emissions data, improve the allocation of allowances, and give companies clearer insight into the cost of cutting emissions.
Still, the rapid growth of AI brings risks. Studies suggest that by 2030, China’s data centres could consume more than 1,000 terawatt-hours of electricity annually, close to Japan’s current yearly power use. Because coal still dominates China’s energy mix, emissions linked to AI infrastructure are expected to keep rising well beyond the country’s 2030 emissions target.
Regulators are beginning to respond. A 2024 action plan requires data centres to improve energy efficiency and increase their use of renewable power by 10% each year. Other measures encourage new facilities to be built in western regions, where wind and solar resources are more plentiful. On the east coast, operators are experimenting with alternatives such as underwater data centres cooled by seawater, which could significantly reduce energy and water consumption.
Despite these challenges, researchers argue that AI could still play a net positive role in China’s climate transition if applied carefully. By optimizing heavy industry, power systems, and carbon markets, AI may help drive emissions reductions even as it adds new pressures that policymakers must manage.
