«Данное сообщение (материал) создано и (или) распространено иностранным средством массовой информации, выполняющим функции иностранного агента, и (или) российским юридическим лицом, выполняющим функции иностранного агента»
Topic: Artificial Intelligence (AI), and Digital Infrastructure Blog Brand: Techland Region: Americas, and Asia Tags: China, Data Centers, Hailanxin, Microsoft, North America, South China Sea, United Nations Convention on the Law of the Sea (UNCLOS), and United States China’s Subsea Ambition: Powering the AI Race with Global “Thermal Debt” April 15, 2026 By: Jeanette Tong
Share
China’s underwater data centers advance AI dominance while exporting environmental costs, forcing the United States to balance competitiveness with accountability and global regulatory leadership.
China has deployed its first commercial underwater data center, developed by Hailanxin (Beijing Highlander Digital Technology Co., Ltd.), off the southeastern coast of Hainan Island, in shallow coastal waters of the South China Sea. Roughly 35 meters (almost 115 feet) down, it is close enough to shore to connect by submarine cable to a land-based relay station. The facility now provides data storage and computing services to internet companies working in artificial intelligence (AI) and big data. Its inaugural clients include Hainan Telecom and Atlas, with Tencent, Alibaba, JD.com, and Pinduoduo slated to come online later this year.
Hailanxin was once a designated supplier to the Chinese Navy, with core competencies in intelligent ship systems, marine information data, and seabed mapping — making it a central player in China’s maritime industrial strategy. In 2022, the US Department of Commerce (DOC) blacklisted the company, alleging that its ocean surveillance systems, built using American technology, may have helped Russia monitor infiltration activities—involving submarines, divers, and warships—along Ukraine’s coastline.
This deployment is more than a feat of engineering. It represents a new front in great-power competition—one fought not with weapons, but with computing infrastructure quietly laid on the ocean floor. China is using the sea as both a resource and a geopolitical tool: offloading the environmental costs of AI expansion onto a shared global commons while racing to become the world’s dominant exporter of cheap AI compute. The United States and its allies must decide whether to match this ambition, constrain it through international law, or—as has too often been the case—simply watch. The answer demands both competitive urgency and environmental accountability, and it requires both at the same time.
From Microsoft’s Lab to China’s Seabed
The concept of underwater data centers traces back to Microsoft’s Project Natick, launched in 2014. When Microsoft declared the experiment a success in 2020, it chose not to commercialize the technology at scale—in part due to significant environmental concerns. Hailanxin spotted the opportunity. In 2019, the company acquired Canadian deep-sea equipment firm OceanWorks, obtaining technology with roots in the same lineage as Natick. It then partnered with the China National Offshore Oil Corporation (CNOOC) to begin constructing the main pressure vessels for its underwater data centers.
The facility sits on the seabed off Lingshui—a county on the southeastern coast of Hainan Island, facing the South China Sea, with a 57.5-kilometer (approximately 36-mile) coastline and sea areas connecting directly to the broader South China Sea. This is not remote deep ocean; it is the coastal water of a populated tropical island, at a depth a recreational diver could reach. Each deployment unit weighs 1,300 metric tons—equivalent to roughly 1,000 automobiles—and sits at a depth of approximately 35 meters (115 feet). The system uses ambient seawater to cool servers while simultaneously storing, processing, and transmitting large volumes of data. Company president Shen Wanqiu stated that, unlike Microsoft’s design, Hailanxin solved the challenge of natural seawater cooling in high-temperature ocean zones through a rack-level split cooling system combined with an external thermal circulation loop—achieving a power usage effectiveness (PUE) as low as 1.07.
The Hidden Cost of Underwater Data Centers: Thermal Debt
The efficiency figures, however, do not tell the whole story. A standard underwater data center pod consumes between 500 kilowatts (kW) and 1 megawatt (MW) of power. Hailanxin’s roadmap calls for 100 pods—implying a total output of 50 to 100 megawatts. When 100 MW of waste heat is continuously discharged into the sea, approximately 100 million joules of energy enter the ocean every second. Even with liquid-cooled rack technology and split cooling systems, that level of thermal discharge is sufficient to raise the temperature of tens of thousands of cubic meters of seawater noticeably within a single hour.
China’s record on environmental stewardship during its economic ascent has been deeply troubling. The country’s dominance in rare earth production—which effectively controls the throat of global high-tech supply chains—was purchased at an enormous ecological price: the sacrifice of tens of thousands of square kilometers of land and lakes across Jiangxi’s Ganzhou region and Inner Mongolia’s Baotou, along with severe health consequences for surrounding populations. Each ton of rare earth extracted generates roughly 2,000 tons of tailings and vast quantities of toxic wastewater.
While the current underwater data centers are located in China’s coastal waters at some distance from neighboring countries such as Vietnam and the Philippines, the ocean is not a closed system. Heat exchanged into these waters eventually integrates into global marine circulation. Hailanxin claims that the maximum temperature rise at the outlet is only 2°C and is manageable—but under the framework of global commons, does unilaterally using the ocean as a “free heat sink” violate the international order? That question deserves a serious answer.
Soon, China will begin exporting low-cost AI tokens to the world, powered by cheap energy and naturally cooled undersea infrastructure. These tokens will become China’s next celebrated high-tech export—following solar panels and electric vehicles (EVs)—and buyers worldwide will rush to adopt them, just as few questioned the environmental devastation behind silicon wafer slicing or EV battery production. The warming of China’s coastal seas may seem distant and inconsequential, but as hundreds of underwater data center pods are submerged, each “locally negligible” temperature rise will accumulate into an irreversible deep-sea thermal burden.
Winning the AI Race Without Losing Ourselves
How should the United States and other major powers navigate the tension between environmental responsibility and great-power competition?
For years, American and Western policymakers have been constrained by the moral absolutism of extreme environmental advocacy, which has impeded economic development across nuclear energy, automotive manufacturing, and petrochemicals—ceding ground to a China that operates with far lower environmental and labor standards. In just over two decades, China leveraged that asymmetry to become the world’s sole manufacturing superpower across nearly every critical supply chain.
Facing a rival that controls the commanding heights of global industry, the United States must adopt flexible, adaptive regulation. That means abandoning rigid legacy rules, streamlining approval processes, and embracing real-time digital environmental monitoring—not choosing between competitiveness and accountability, but demanding both simultaneously.
On the international front, nations should press the United Nations (UN) Convention on the Law of the Sea (UNCLOS) to clarify its ambiguous treatment of thermal pollution and move urgently toward a Global Thermal Footprint Tracking Protocol to fill the legal vacuum created by rapid technological change.
If UN institutions remain paralyzed by bureaucratic inertia or geopolitical gridlock, then the AI industry itself—OpenAI, Google, Meta, and their peers—must step up. These companies should establish transparent thermal emissions reporting, treating the environmental cost of compute as a labeled, accountable metric. If artificial intelligence can solve humanity’s most complex problems, it should start by solving the regulatory gaps its own heat generates.
America will not—and must not—become another China: a nation that wins by ignoring environmental consequences and abandoning commercial ethics. We must win. But we must win better.
About the Author: Jeanette Tong
Jeanette Tong is the editor of Yibao.net and a research fellow at Citizen Power Initiatives for China. Her current research tracks China’s broad high-technology expansion—spanning digital infrastructure, energy, and biotechnology—and its role in great-power competition, with a focus on how Beijing leverages these advances to reshape the international order.
The post China’s Subsea Ambition: Powering the AI Race with Global “Thermal Debt” appeared first on The National Interest.
Источник: nationalinterest.org