The Shanghai Lingang undersea data center demonstration project, located about 10 kilometers off Shanghai's eastern coast, officially entered operation in May, according to China Daily.

Developed by a subsidiary of China Communications Construction, the facility combines offshore engineering, renewable energy and AI-focused computing infrastructure. The project has a planned capacity of 24 megawatts, enough to supply electricity to around 20,000 households.

Its core innovation is a "direct offshore wind connection" system, in which electricity generated by nearby wind farms is transmitted straight to submerged data modules through subsea photoelectric composite cables, bypassing conventional grid networks.

Shanghai Lingang undersea data center demonstration project. Photo courtesy of China Communications Construction

The facility also relies on naturally cold seawater to cool its servers. Developers said its circulating copper-pipe heat exchange system reduces electricity consumption by 22.8%, eliminates the need for freshwater cooling and cuts land use by more than 90% compared with conventional data centers.

With a planned investment of 1.6 billion yuan ($235 million), the project is part of China's broader effort to address the mounting energy demands created by AI, CGTN reported.

According to the International Energy Agency, global electricity consumption by data centers is expected to more than double by 2030, with AI becoming the main driver of that growth.

Li Zhen, a professor at Tsinghua University, said China's data centers currently consume around 250 billion kilowatt-hours of electricity annually, including about 80 billion kWh used for cooling.

Li said that if data centers of the same scale were built underwater, cooling-related electricity consumption could fall to around 30 billion kWh annually, saving about 50 billion kWh of power each year. He estimated the reduction would be equivalent to avoiding the burning of roughly 15 million metric tons of standard coal annually, helping lower carbon emissions.