In the city of Chifeng, in northern China’s Inner Mongolia region, Reuters reports an ultra-modern factory humming to life not on coal or gas, but on wind, solar power, and algorithms.
Owned by green-energy company Envision, the facility produces hydrogen and ammonia using renewable electricity, guided by an artificial intelligence system designed to solve one of the hardest problems in the global energy transition: how to run energy-hungry industrial processes on power sources that are inherently unstable.
The plant operates on a standalone grid that feeds electricity directly from Envision’s wind and solar farms into its production lines. That electricity supply can swing sharply depending on weather conditions, a problem for chemical manufacturing, which traditionally depends on steady, uninterrupted power. To manage this, Envision built an AI-driven operating system that continuously adjusts production levels to match real-time renewable output.
Zhang Jian, Envision’s chief engineer for hydrogen energy, described the system as “a conductor” that synchronizes electricity use with nature. When wind speeds rise, the AI automatically ramps up production to absorb as much green power as possible. When output drops, the system scales back electricity demand almost instantly, Zhang told China Energy News, a state-run publication.
The Chifeng facility is Envision’s blueprint for large-scale renewable hydrogen and ammonia production, fuels that could play a critical role in decarbonizing hard-to-abate sectors such as steelmaking, chemicals, and shipping. It is also a glimpse into how China is deploying AI not just as a digital tool, but as infrastructure — a way to make its sprawling renewable energy system work at scale.
“AI can play a hugely important role in China’s climate action and energy transition,” said Zheng Saina, an associate professor specializing in low-carbon transition at Southeast University in Nanjing. Beyond industrial optimization, she noted, AI can help calculate and project carbon emissions, forecast electricity supply and demand, and manage increasingly complex power systems.
This push comes at a critical moment. Energy has become one of the biggest bottlenecks in the evolution of artificial intelligence globally. As AI models grow larger and data centers proliferate, electricity demand is surging at a pace that is straining grids from the United States to Europe. In China, AI data centers alone are projected to consume more than 1,000 terawatt-hours of electricity annually by 2030 — roughly equivalent to Japan’s total yearly power consumption.
That looming demand surge presents a paradox. AI is being positioned as a key enabler of China’s green transition, yet its own energy appetite risks undermining climate goals.
“AI data centers are expected to cause explosive growth in electricity demand,” Zheng warned. “This is a problem that urgently needs addressing.”
Where China believes it has an edge over the United States is in energy diversity and scale. China leads the world in installed wind and solar capacity, and continues to add renewable power at a pace unmatched elsewhere. While the U.S. has focused much of its AI investment on building advanced large-language models and cloud infrastructure, China has placed parallel emphasis on integrating AI into physical systems, particularly energy.
President Donald Trump is opposed to green energy initiatives and has recently ordered the halt of wind energy programs.
“China is developing very specific, tailored AI solutions that support the grid and individual energy sectors like wind, solar and even nuclear,” said Cory Combs, associate director at Beijing-based research firm Trivium China. “That’s different from the U.S. approach, which has largely centered on model performance rather than system-level integration.”
As renewable energy expands, grid flexibility has become essential. Wind and solar power are intermittent, and without smart coordination, excess electricity is often wasted. AI is increasingly seen as the missing link. In September, Beijing launched an “AI+ energy” strategy aimed at deeply integrating artificial intelligence across the power system.
By 2027, the plan calls for more than five large AI models dedicated to energy applications, at least 10 replicable pilot projects, and over 100 real-world use cases. Within three more years, China wants to reach what it describes as a “world-leading level” in combining specialized AI technologies with the energy sector.
One of AI’s most critical roles is demand forecasting. Power grids must balance supply and demand second by second to avoid blackouts. Fang Lurui, an assistant professor of power-system planning at Xi’an Jiaotong-Liverpool University in Suzhou, said accurate AI-driven forecasts allow grid operators to plan ahead, deciding how much electricity to store in batteries or when to call on backup generation.
“If AI models can accurately predict renewable output and electricity demand throughout the day, the grid can operate more efficiently and safely,” Fang said.
Better forecasting also reduces reliance on coal-fired backup plants and allows more renewable power to be absorbed rather than curtailed.
Some Chinese cities are already experimenting at scale. Shanghai has launched a citywide virtual power plant backed by a digital platform developed by State Grid. The system aggregates electricity generation and load-reduction capacity from 47 operators, including data centers, building heating and cooling systems, and electric vehicle charging networks, allowing them to function as a single flexible resource.
During a trial run in August, the platform successfully flattened a demand spike by shedding 162.7 megawatts of load — roughly the output of a small coal-fired power plant.
Beyond grid management, China is also exploring how AI can strengthen its national carbon market, which covers more than 3,000 companies across power, steel, cement, and aluminum smelting. These sectors account for over 60% of the country’s total carbon emissions. AI could help regulators verify emissions data, improve allocation of free allowances, and allow companies to calculate compliance costs more precisely, according to Chen Zhibin, a senior manager for carbon markets at think tank Adelphi.
Yet challenges remain formidable. Studies suggest the lifecycle carbon emissions of China’s AI industry could double between 2030 and 2038, peaking at nearly 700 million tons — higher than Germany’s total emissions in 2024. China’s grid still relies heavily on coal, complicating efforts to green AI at speed.
To address this, Beijing has mandated that data centers improve energy efficiency and raise renewable power usage by 10% annually. It is also encouraging new data centers to be built in western regions rich in wind and solar resources. On the East Coast, operators are experimenting with unconventional solutions. Near Shanghai, a data center is being built underwater, using cold seawater for cooling and drawing more than 95% of its electricity from a nearby offshore wind farm, according to its developer, Hailanyun.
While AI’s energy footprint is now a central concern, many researchers argue that the technology remains indispensable. Xiong Qiyang, a PhD candidate at Renmin University of China who co-authored one study on AI emissions, said the trade-off is unavoidable but manageable.
“AI’s energy use is a real concern,” he said. “But it does much more good than harm in helping key sectors reduce emissions. That makes AI an essential tool in China’s green transition.”
China’s bet amid intensifying competition in artificial intelligence is that by pairing AI with a vast and diversified renewable energy system, it can stay ahead in both the AI race and the energy transition, even as other economies struggle to keep the lights on.
