China’s $226 Million Undersea Data Center Cuts Land And Water Use For AI Infrastructure

China’s undersea data center near Shanghai has drawn about $226 million in investment and is designed for 24MW of total power capacity.
The project uses ocean cooling and wind power, with green electricity covering as much as 95% of operations.
Developers say the facility can save 61 million kilowatt-hours of electricity each year, while cutting land use by 90% and using no water.

Ten meters below the surface off Shanghai’s Lingang Special Area, China is testing a new answer to one of AI’s fastest-growing problems: how to power and cool data centers without pushing energy, water, and land use to unsustainable levels.

The undersea data center is powered largely by wind energy and cooled by the surrounding ocean. It houses 192 server racks across four levels and has now become fully operational after opening earlier this year.

The project has attracted around $226 million in investment. Its planned power capacity is 24MW, though only 2.3MW is currently in use. Developers say as much as 95% of the facility’s electricity comes from green sources.

For a sector under pressure to reduce its environmental footprint, the design is significant. It uses no water for cooling and reduces land use by 90%. According to CGTN, the facility is also expected to save 61 million kilowatt-hours of electricity each year.

AI Growth Is Raising Pressure On Energy Systems

The project comes as governments, technology companies, and investors face mounting concern over the environmental cost of artificial intelligence.

AI workloads require large volumes of computing power. That means more GPUs, more server racks, and more electricity. The energy demand does not stop at processing. Data centers also need major cooling systems to keep hardware stable and prevent equipment failure.

Traditional data centers can consume large amounts of power and water, especially in regions where cooling demand is high. That has turned digital infrastructure into a growing ESG issue. Companies need capacity for AI growth. However, regulators, investors, and communities are also asking how that capacity will be built.

China’s undersea model addresses several of those concerns at once. Ocean cooling reduces the need for conventional water-based systems. Offshore placement lowers pressure on scarce urban land. Wind power also helps reduce exposure to fossil-fuel electricity.

Governance And Infrastructure Stakes Are Rising

For China, the facility is more than an engineering experiment. It sits within a wider policy context where digital infrastructure, clean power, and industrial competitiveness are increasingly linked.

The Lingang Special Area has become an important zone for advanced technology and strategic industries. Placing an energy-efficient data center there supports China’s push to scale AI infrastructure while managing resource constraints.

For policymakers, the model offers a possible route to reduce the environmental burden of digital growth. For investors, it raises a practical question: will future data center assets be assessed not only on computing capacity, but also on power sourcing, water intensity, and land efficiency?

Those metrics are already moving higher on the ESG agenda. Data centers that can show lower operating emissions and reduced resource demand may have an advantage as reporting rules tighten and buyers seek cleaner cloud and AI services.

Undersea Data Centers Move From Experiment To Deployment

China is not the first country to test underwater data centers. Microsoft explored the concept through Project Natick, which involved placing a data center on the ocean floor. The experiment reported strong results, but Microsoft has not pushed the model into broad commercial rollout.

China’s latest project suggests the concept may now be moving closer to real-world infrastructure planning. Developers have already indicated that they intend to expand deployments in future.

That matters because the race to build AI capacity is accelerating. Some firms are even exploring data centers in space, a far more expensive and experimental route. Space-based infrastructure may attract attention, but commercial viability remains uncertain.

Undersea data centers offer a more immediate comparison. They use existing marine conditions, can connect to offshore wind resources, and may reduce pressure on urban energy and water systems.

For executives and investors, the takeaway is clear. AI infrastructure will not be judged only by speed, scale, and processing power. It will also be measured by how efficiently it uses electricity, water, and land.

China’s undersea data center shows how that calculation may shift. If the model proves scalable, it could influence how coastal economies plan the next wave of digital infrastructure. The wider question is whether global AI growth can align with climate and resource goals before power demand becomes a larger constraint.