China taps offshore wind, subsea data centres to ease AI computing bottleneck

The pilot off Shanghai

China is testing the sea as a new frontier for AI compute. It has been reported that an underwater data centre (UDC) located 10 kilometres off Shanghai’s eastern coast and 10 metres under water has begun operation, and CCTV said it is the first such facility in the world to be directly linked to an offshore wind farm. The project, built by Shanghai HiCloud Technology (上海HiCloud科技), reportedly carries a total investment of 1.6 billion yuan (US$232.4 million) and a planned capacity of 24 megawatts to feed onshore cloud and telecoms infrastructure.

Who is behind it and what for

The UDC was developed by Shanghai HiCloud, a unit of Shenzhen-listed marine navigation and communications equipment group Highlander (海兰德). It is presented as a pilot to explore the technical and commercial viability of subsea data cabins connected to renewables. The company has said the cluster will be used for AI scenarios, embodied intelligence and autonomous driving — a claim that has been reported but remains to be independently verified. Highlander previously initiated a similar UDC project off Hainan province, signalling a broader push to prototype maritime compute hubs.

Why China is looking seaward

Why go underwater? Offshore wind offers abundant, low-carbon power close to coastal demand centres, and subsea placement improves cooling efficiency — both critical as China’s demand for AI compute surges. The move also needs to be read against geopolitics: Washington-led export controls and trade restrictions on advanced AI chips have heightened urgency to secure domestic compute capacity and diversify supply chains. Beijing is therefore experimenting with unconventional infrastructure as part of a larger strategy to close the gap with Western AI ecosystems.

The big questions

Pilot projects can prove technical concepts quickly. But can subsea centres scale economically and navigate regulatory, maintenance and data-governance challenges? It has been reported that authorities and industry will use these early sites to test viability. If successful, expect more maritime and even airborne compute experiments — but commercial rollout will hinge on costs, reliability and the wider chip-access picture.