China has built the world's first commercial underwater data center, and it works — but the ocean is not about to swallow the entire global computing infrastructure.
The basic appeal is straightforward: data centers consume enormous amounts of electricity to keep their servers cool, and seawater is free, abundant, and excellent at absorbing heat. A Shanghai underwater data center project in the Lingang area has entered commercial operation with a 24 megawatt target capacity, using seawater cooling linked to offshore wind power. There are also underwater data center developments underway around Hainan. These are not laboratory experiments or prototype tanks. They are real infrastructure projects built by a country with proven expertise in turning engineered concepts into functioning steel, concrete, cables, and operating assets.
The technology itself is sound. Microsoft's Project Natick demonstrated this in 2018, when the company deployed a 12-rack, 864-server data center pod off Scotland's Orkney Islands. When Microsoft retrieved it in 2020, the underwater servers had a failure rate one-eighth that of identical land-based servers in a control group — a genuinely interesting result likely enabled by the sealed, dry, nitrogen-filled environment and stable temperatures. Yet Microsoft never commercialized Project Natick. That restraint is telling. Major technology firms do not leave obviously superior infrastructure sitting unused unless real constraints are attached to it.
Those constraints are substantial. Underwater data centers solve one problem elegantly — cooling — while creating or complicating almost every other problem a data center must manage. Serviceability becomes complex. A technician cannot simply walk across a floor, open a cabinet, and swap a failed server. Instead, modules must either be designed to run without repair until planned retrieval, or the entire pod must be brought back to shore on a vessel — both approaches carry significant costs and operational consequences.
Then there are the cables. Underwater data centers still need power and data connections tethering them to the grid and the wider internet. Subsea cables are mature technology, but mature does not mean simple. They require specific routes, landing points, permitting approval, specialized repair vessels, security protocols, and contingency plans for failure modes. Environmental review adds another layer. Power interconnection, fiber routing, latency, tax considerations, workforce needs, construction logistics, and increasingly political factors all remain as complex as they are on land.
China building underwater data centers proves the concept can be engineered. It does not prove that seabed computing will become the next dominant global architecture. For hot, land-constrained, coastal locations with strong marine engineering expertise and nearby clean electricity sources, underwater data centers offer genuine advantages. They can reduce mechanical cooling loads, free up scarce land, and reduce pressure on local freshwater systems compared with traditional evaporative cooling methods.
But infrastructure decisions are rarely made on the strength of one advantage. As artificial intelligence workloads push more electricity through denser chips, cooling matters more than ever — and so do all the other constraints that have long defined where and how data centers operate. China's commercial underwater projects deserve serious attention, not hype. They are impressive engineering achievements that solve real problems in specific contexts. That is exactly what infrastructure should do.