Published on Dec 24, 2025

Artificial intelligence is widely seen as extremely energy-hungry, but it is also helping to generate new energy. In Chifeng in northern China, the company Envision is using AI to optimise its solar and wind power plants. By perfectly balancing renewable electricity, the facility is able to produce more green hydrogen than before.

Zhang Jian, Envision’s chief engineer for hydrogen energy, compares the AI model developed by the company, known as Tianshu, to a ‘conductor’ who fine-tunes the plant’s electricity consumption to moment-by-moment fluctuations in the weather.

When the wind blows particularly strongly on the edge of the Gobi Desert, the plant’s electrolysers are ramped up to maximum output in order to waste as little renewable energy as possible. When conditions weaken, the process is reversed.

There is probably no better training ground for embodied AI in energy systems than the harsh climate of Inner Mongolia, said Zhang Lei, chairman of the Shanghai-based Envision Group. At times there is too much sun, while at other times sandstorms block sunlight for days on end. This generates ideal data for training large AI models to regulate power loads.

Thanks to the integration of AI, it is possible to operate what is currently the world’s largest hydrogen production facility entirely on renewable energy, Zhang said.

Because highly accurate forecasts from an AI-based weather model are also factored in, Envision has been able to replace passive reactions to sun and wind in the plant’s energy system with proactive optimisation.

This is a concrete way in which AI is supporting China’s energy transition. Performance optimisation is already driving down the cost of green hydrogen and green ammonia, which are increasingly used in Chinese steel plants, chemical facilities and refineries as alternative clean energy carriers and as substitutes for coal or coke.

Green ammonia is also produced in Chifeng. Within two years, project representatives say, AI is expected to bring the price of green ammonia down to the level of grey ammonia. The production of grey ammonia from natural gas still generates large volumes of greenhouse-gas emissions. Even if this projection proves overly optimistic, the potential seems to be substantial.

Here, the world’s most advanced physical AI and AI power systems are taking shape, said Envision’s top executive. The solutions being developed on the edge of the Gobi Desert could be readily transferred to other industrial scenarios, such as green steel production or low-carbon data centres.

China has a particularly strong need to deploy AI in energy facilities and across the power grid, as the expansion of renewable energy has been especially rapid year after year. As ever more green electricity is fed into the grid, regulation requirements increase accordingly. AI can manage this in milliseconds and continues to learn continuously.

However, this demand for AI-enabled operating systems also exists in other countries expanding solar and wind power. In November, Envision sold what it described as the world’s largest AI energy system abroad, according to the Chinese industry outlet Huanqiu Lingtan. The company signed a contract with the UK-based firm Statera Energy to deploy its system at an energy-storage facility in Carrington.

What has been developed in Chifeng is therefore beginning to enable European utilities to fine-tune their systems using AI. Over the medium to long term, Zhang Lei recently said at a specialist conference on embodied AI and energy in Beijing, algorithms will fundamentally transform the entire energy industry.

At the conference, Zhang explained that AI-based energy system solutions create a closed loop of forecasting, dispatch, trading and self-learning. Energy-storage systems, he said, will no longer be simple stacks of equipment, but will evolve into ecosystems of intelligent agents.

This, many in China believe, will also reshape competition in the energy industry as a whole. Traditional assets such as physical infrastructure are likely to lose relative value compared with new AI assets.

At a time when discussions of AI often focus on energy-hungry data centres, the progress achieved at the Chifeng electrolysis facility is encouraging. According to Reuters, forecasts suggest that data centres in China alone could consume more than 1,000 terawatt-hours of electricity per year by 2030.

Even so, Chinese engineers hope that AI, thanks to its enormous potential in the energy industry, will ultimately deliver more benefits than harm. One way or another, they have begun to treat it as one of several core technologies driving the green transformation of China’s economy.