As Global Water Shortages Loom, China Is Gambling on a Radical Undersea Solution That Could Change How the World Handles AI Cooling
Not just another experimental prototype.
The world’s first wind-powered underwater datacentre is now fully operational off the coast of Shanghai, marking a major gamble by China to solve the energy and water crises plaguing artificial intelligence infrastructure. The Shanghai Lingang undersea datacentre demonstration project, which launched in May, is already proving that submerging servers in seawater can slash power consumption by more than a fifth while eliminating the need for freshwater cooling entirely.
The 24-megawatt facility, a collaboration between HiCloud Technology and state-owned China Communications Construction, is already handling real-world computing demands. Located about six miles offshore and submerged 33 feet below, the datacentre relies on a nearby offshore wind farm for its power, per The Guardian. That direct connection to renewable energy, combined with the natural cooling properties of seawater, makes this setup far more efficient.
If you’ve ever wondered why datacentres are such energy hogs, here’s the breakdown. Between 25% and 40% of a conventional datacentre’s electricity goes toward cooling servers with chilled water. That’s a massive drain on resources, especially as AI workloads continue to explode.
The new datacentre cuts power use significantly
The United Nations University Institute for Water, Environment and Health recently warned that datacentres could consume a staggering 9.3 trillion litres of water by 2030 – enough to meet the annual domestic needs of all 1.3 billion people in sub-Saharan Africa. China’s underwater approach sidesteps that problem entirely by using seawater for cooling, which not only cuts power use but also frees up freshwater supplies for other critical needs.
China isn’t entirely new to underwater datacentres. HiCloud already launched the world’s first commercial underwater datacentre in Hainan last year. But the Shanghai project is the first to combine subsea servers with offshore wind power, creating a model that could redefine how datacentres are built in coastal regions.
The wind farm powering the facility is visible from Lingang, a high-tech free-trade zone in eastern Shanghai that’s also home to Tesla’s gigafactory. That proximity to cutting-edge industries makes it an ideal testing ground for this kind of innovation.
The idea of underwater datacentres isn’t unique to China
Microsoft tested a similar concept in the waters around Orkney, Scotland, back in 2018. The company reported promising results two years later, but progress has since stalled. Dr. Hanjiang Dong of Hong Kong Polytechnic University explained why China pulled ahead.
“Microsoft was earlier in proving the concept, while China moved further on commercial deployment because it was able to bring together market demand, industrial capability, marine engineering, and policy support more quickly into a commercial project,” he said. That ability to rapidly scale from experiment to real-world application is a hallmark of China’s approach to AI infrastructure.
China has made AI a cornerstone of its economic and development strategy. Last year, the government released an AI action plan calling for accelerated datacentre construction, with a pledge to “significantly increase” clean energy supplies for AI infrastructure by 2030.
The Shanghai Lingang project received a hefty 1.6 billion yuan ($236.2 million) in investment, underscoring the government’s commitment to this vision. For China, this isn’t just about keeping servers cool. It’s about reimagining how computing power, energy, and geography intersect in the AI era.
The shift toward underwater datacentres isn’t without risks
Submerging servers in the ocean could disturb marine sediments or create localized heating effects in the surrounding water. However, experts believe these risks are manageable with proper monitoring.
Prof. Rick Stafford, a marine biologist at Bournemouth University, said the cooling process would result in some localized temperature increases but added, “An underwater datacentre is likely a good idea. While the cooling using seawater will result in some localised elevated temperatures, these will not be far reaching.” That’s a small trade-off for a solution that could drastically reduce the environmental footprint of AI.
The energy savings from this approach are nothing short of impressive. According to China Daily, Tsinghua University’s Professor Li Zhen noted that conventional datacentres typically use about one-third of their total electricity for cooling. In an underwater datacentre of the same scale, that figure drops to just one-tenth.
China’s datacentres currently consume around 250 billion kilowatt-hours of electricity annually, with 80 billion kWh going toward cooling alone. Li estimated that switching to underwater datacentres could cut cooling consumption to around 30 billion kWh, saving roughly 50 billion kWh of electricity each year. That’s the equivalent of not burning 15 million metric tons of standard coal annually, a massive reduction in carbon emissions.
The Shanghai project’s innovation goes beyond just cooling
It also introduces a “direct offshore wind connection” model, where electricity from offshore wind farms is transmitted straight to the submerged datacentre modules via subsea photoelectric composite cables. This bypasses traditional grid-routing systems, further improving efficiency. The design also uses a circulating copper-pipe heat exchange system, which maximizes the cooling potential of seawater while minimizing energy waste.
For coastal cities where land, electricity, and freshwater are in short supply, this model could be a game-changer. Shanghai, one of China’s leading AI hubs, is home to large-model developers, autonomous driving firms, biotech companies, and fintech groups – industries where even milliseconds of latency can make or break performance. The underwater datacentre’s proximity to these ensures low-latency, high-density computing power without the usual environmental trade-offs.
Globally, tech companies are scrambling to find ways to reduce the environmental impact of AI as model sizes and inference demands skyrocket. The combination of offshore renewable power and seawater cooling is emerging as a compelling solution, particularly in markets where resources are constrained. For China, this is about engineering a new generation of AI and industrial infrastructure where electricity, cooling, and computing are seamlessly integrated beneath the sea.
(Featured image: Daoducquan)
Have a tip we should know? [email protected]
Follow The Mary Sue:
