The facility, located in Shanghai's Lingang Special Area between two phases of an offshore wind farm, became fully operational last week after a construction sprint that began in June 2025 and wrapped in October. HiCloud Technology, the operator, partnered with Shanghai Lingang Special Area Investment Holding Group, Shenergy Group, Shanghai Telecom, and CCCC Third Harbor Engineering to deploy pressure resistant subsea modules packed with GPU (graphics processing unit) clusters. China Telecom installed its own computing hardware inside to run AI (artificial intelligence) workloads, big data annotation, 5G infrastructure, and domestic LLM (large language model) development. The project started with a 2.3 MW (megawatt) demonstration phase before scaling to 24 MW total capacity. The cooling mechanism is brutally simple and eliminates the need for industrial chillers. Hot air from servers heats a refrigerant in copper pipes, turning it from liquid to gas. The gas rises through buoyancy to a cooling layer in the upper module, where it exchanges heat with seawater via a heat exchanger and condenses back to liquid. Gravity pulls the liquid refrigerant back down to the server room, creating a passive loop that requires zero additional power. Chinese state media claims the system hits a PUE (Power Usage Effectiveness) of around 1.15, meaning only 15% of total energy goes to non computing overhead. Conventional enterprise data centers typically run closer to 1.5 or worse, burning massive amounts of electricity just to keep equipment from melting. Developers say the underwater setup cuts electricity consumption by 22.8%, uses zero freshwater, and requires over 90% less land than equivalent terrestrial facilities. That last figure matters in coastal China, where real estate commands premium prices and industrial zoning fights are vicious. By sticking servers on the seafloor adjacent to wind turbines, the project sidesteps land acquisition entirely and taps renewable energy at the source, avoiding transmission losses and grid dependency. The direct wind to server connection is the key efficiency play here, not just the seawater cooling. But underwater computing is a maintenance nightmare. Saltwater corrosion, long term pressure sealing, subsea cable reliability, and hardware accessibility are all unsolved at scale. When a server fails on land, a technician swaps it in minutes. When a server fails 30 meters underwater in a sealed module, you're looking at an expensive recovery operation or accepting the loss until the entire module is hauled up for service. The Shanghai operators are betting on sealed modular systems, heavy redundancy, and remote monitoring to minimize physical intervention. That works until it doesn't, and the first major hardware failure will test whether this model can survive real world operational stress. Microsoft's Project Natick tested submerged data center capsules off California in 2015 and deployed a larger trial off Scotland's Orkney Islands in 2018. The Orkney deployment showed lower hardware failure rates underwater compared to land based equivalents, likely due to stable temperatures and the absence of oxygen and humidity fluctuations. But Microsoft quietly discontinued the project by 2024 without moving to commercial deployment, and the company has never publicly explained why. The prevailing theory is that maintenance costs and operational complexity outweighed the efficiency gains, or that hyperscale cloud economics just don't favor boutique underwater deployments when you can build massive land based facilities in cheap power regions. China's willingness to commercialize what Microsoft abandoned suggests either a different cost structure, a strategic push for energy independent infrastructure, or both.