China has launched a new underwater AI data center off the coast of Shanghai, using ocean water to cool high-performance servers while tapping offshore wind for power.
As reported by Scientific American, the facility represents China’s efforts to align its expanding AI ambitions with climate-conscious infrastructure. Developers say the project “could influence the norms of global sustainable computing” by cutting energy and freshwater demand in data-heavy operations.
The environmental impact of this AI data center
Built by the tech firm Hailanyun as part of a $223 million initiative, the underwater facility stays cool without energy-hungry traditional systems, using seawater circulated through radiator-equipped racks.
The center is powered almost entirely by a nearby offshore wind farm, which supplies 97% of its energy and enables near-zero carbon emissions. Combined with natural ocean cooling, the setup significantly reduces the environmental impact of large-scale data processing.
Inside, it holds 198 server racks that can handle intensive AI workloads. According to Hailanyun, the system delivers enough computing power to train a large language model in just one day while operating in a stable, thermally regulated environment that requires minimal human intervention.
According to Scientific American, Zhang Ning, a postdoctoral researcher at the University of California, Davis, describes the project as a blueprint for next-generation low-carbon infrastructure, one that blends high-performance computing with sustainable design in a naturally efficient setting.
The high demands of dry grounds
Shanghai’s underwater site shows what traditional data facilities are missing.
Most of the world’s data centers, like Google’s and Meta’s, still operate on land. However, while dry air in arid regions protects hardware, it strains limited water supplies. Cooling can require hundreds of thousands of gallons daily, often drawn from sources shared with farms and communities. These consume massive amounts of power, with nearly 40% of their electricity going solely to cooling systems, as noted by Scientific American.
As AI workloads grow, so do the demands on these facilities. A new report from UNESCO and University College London warns that the energy toll of artificial intelligence, especially large language models, is becoming unsustainable.
Submerging data centers changes the equation. By using naturally cold seawater and clean offshore wind, ocean-based systems slash both energy and freshwater use.
Subsea systems come with trade-offs
Cooling with seawater may lower energy use, but it doesn’t come without impact. Warmer discharge can reduce oxygen levels around the capsule, potentially affecting marine life. Hailanyun spokesperson Li Langping has dismissed the concern, saying the temperature rise was under one degree and had no substantial effect.
Other challenges are less visible. Researchers at the University of Florida found that certain sound frequencies can disrupt underwater servers. They’ve developed a machine learning tool to detect and contain such interference early.
Then there’s repair. What’s routine on land becomes costly and slow underwater, where access is limited and failures are harder to fix.
The idea isn’t new — but this time, it’s moving
Since 2021, marine technology firm Beijing Highlander has tested sealed server capsules off the coasts of Zhuhai and Hainan, part of a years-long effort to push subsea computing from experiment to infrastructure.
Whether the strategy scales or not, China’s ambition is undeniable. It’s preparing to compete in AI far beyond dry ground.
While China builds the infrastructure, its open-source AI is earning global praise. See why NVIDIA’s CEO is calling it a catalyst for progress.