China has begun operations of the world's first undersea data center directly powered by offshore wind, as the country races to solve the soaring energy demands of artificial intelligence with greener and more efficient infrastructure.
The Shanghai Lingang undersea data center demonstration project, built by a subsidiary of China Communications Construction, officially entered operation in the waters off Shanghai's eastern coast in May.
The project combines offshore engineering, renewable energy and AI-focused digital infrastructure in a model Chinese officials and engineers describe as a potential template for next-generation computing systems.
Located about 10 kilometers offshore in Shanghai's Lingang area, the project has a planned capacity of 24 megawatts, which is enough to power roughly 20,000 households.
Its core innovation is what developers call a "direct offshore wind connection" model. Electricity generated by offshore wind farms is transmitted directly to submerged data modules through subsea photoelectric composite cables, bypassing traditional grid-routing systems.
The system also uses seawater as a natural cooling source through a circulating copper-pipe heat exchange design, reducing electricity consumption by 22.8 percent, eliminating freshwater use entirely and cutting land usage by more than 90 percent.
The move is not only an engineering breakthrough, but also a paradigm shift in the relationship between computing power, energy and geographic space in China, industry experts said.
The launch comes as China's AI boom fuels a rapid rise in demand for low-latency, high-density computing infrastructure. Shanghai has become one of China's leading AI hubs, home to large-model developers, autonomous driving firms, biotech companies, fintech groups and advanced manufacturing enterprises — industries where milliseconds can determine commercial performance.
The project also reflects a growing global scramble to tackle the energy and cooling crisis facing AI infrastructure.
Data centers have become one of the world's fastest-growing electricity consumers as companies expand AI model training and inference capacity. Cooling alone accounts for a large share of energy consumption in conventional data centers, particularly in densely populated urban markets.
Tsinghua University Professor Li Zhen said conventional data centers typically use about one-third of their total electricity consumption on cooling systems.
"For an undersea data center of the same scale, the electricity used for cooling would only account for about one-tenth of total power consumption," Li said.
China's data centers currently consume around 250 billion kilowatt-hours of electricity annually, Li said, with roughly 80 billion kWh used for environmental cooling.
"If data centers of the same scale were placed underwater, even allowing extra margins, cooling consumption could fall to around 30 billion kWh," Li said. "That would save about 50 billion kWh of electricity each year."
He estimated the reduction would be equivalent to not burning roughly 15 million metric tons of standard coal annually, significantly lowering carbon emissions.
Globally, major technology companies are searching for new ways to reduce the environmental footprint of AI infrastructure as model sizes and inference demand expand rapidly. The combination of offshore renewable power and seawater cooling could become increasingly attractive in coastal markets where land, electricity and freshwater resources are constrained.
For China, Li said, the country that has built the world's largest manufacturing supply chains is now attempting to build a new generation of industrial infrastructure for the AI era — one where electricity, cooling and computing are engineered as a single integrated system beneath the sea.