China has switched on the world’s first wind-powered underwater data center off Shanghai, using seawater cooling and offshore wind to cut energy use, eliminate freshwater consumption, and shrink its land footprint dramatically.
The race to keep artificial intelligence running without draining the planet’s water and power grids has taken a genuinely strange turn. Off the coast of Shanghai, in the Lin-gang Special Area, China has completed and activated what it’s calling the world’s first commercial-scale underwater data center powered primarily by offshore wind. The facility sits roughly 35 meters beneath the surface, is projected to reach up to 24 megawatts of capacity, and represents an estimated $226 million investment backed by a consortium of state-linked firms. As data centers worldwide strain municipal water supplies and electrical grids to feed the AI boom, China’s decision to sink its servers into the sea, cooling them with ocean water and powering them almost entirely with wind, is being closely watched by engineers, environmental researchers, and rival tech powers alike. The Lin-gang unit is described as a sealed, seawater-cooled cluster designed to run on offshore wind, and it is backed by a consortium including Shenergy, China Telecom’s Shanghai branch, INESA, and CCCC Third Harbor Engineering.
Ready to lead the China’s underwater data center?
Discover the regional trends and growth factors shaping the industry. We’re here to assist with expert, personalized data.
Call +1 303 800 4326 or Send us a message for a personalized consultation.
A Landmark Launch Backed by Growing Momentum
China has finished building and switched on a wind-powered underwater data center off Shanghai’s coast. The Lin-gang facility, submerged 35 meters beneath the surface, represents an estimated $226 million investment and is projected to reach up to 24 megawatts of capacity, drawing roughly 95% of its electricity from offshore wind. The project launched in June 2025 through a cooperation agreement with the Lin-gang administrative committee, and construction was completed by October 2025, with the facility achieving a power usage effectiveness (PUE) below 1.15. This builds on earlier momentum: China had already launched an underwater data center off Lingshui Li, Hainan Province first deployed in December 2022 which became the world’s first commercial underwater data center, distinct from Microsoft’s earlier test facility off Scotland. Together, these projects mark a shift from research curiosity to industrial deployment, with Lin-gang representing the larger step of pairing underwater computing directly with an offshore wind farm.
The Engineering: How Water Replaces Chillers
The engineering innovation of the Lin-gang underwater data center lies in its seawater-based thermal management system, which uses the surrounding ocean as a heat sink rather than relying primarily on conventional mechanical chillers. In a conventional land-based data center, servers generate substantial heat that must be removed through chilled-water systems, refrigeration compressors, cooling towers, pumps, and air-handling equipment. These auxiliary systems can represent a significant portion of total facility power consumption.
In the underwater design, computing equipment is housed inside sealed, pressure-resistant server modules. Heat generated by processors, GPUs, power supplies, and other electronic components is transferred through a closed-loop heat-exchange system incorporating copper piping. Copper is suitable for this application because of its high thermal conductivity, allowing heat to move efficiently from the internal coolant circuit toward the external seawater environment.
The thermal pathway can be simplified as:
IT equipment → internal coolant → copper heat exchanger → seawater → ocean
Rather than mechanically lowering the temperature of the coolant through energy-intensive refrigeration, the system transfers waste heat directly toward the comparatively stable surrounding seawater. The ocean therefore functions as a passive external heat sink, reducing the cooling-system electrical load.
According to reported project figures, this architecture can reduce electricity consumption by approximately 22.8% compared with an equivalent land-based facility. The facility is also targeting a Power Usage Effectiveness (PUE) of approximately 1.15. PUE is calculated as:
PUE = Total Facility Energy ÷ IT Equipment Energy
A PUE of 1.15 means that for every 1.00 kWh consumed by computing equipment, approximately 0.15 kWh is used for supporting infrastructure, including cooling, power distribution, pumps, monitoring, and other overhead. A theoretical PUE of 1.0 would mean virtually all facility energy is delivered to IT equipment.
The key technical advantage is therefore not simply the underwater location, but the integration of sealed computing modules, conductive heat-transfer materials, closed-loop cooling, and seawater heat rejection. By eliminating or substantially reducing dependence on compressor-driven chillers, the architecture lowers cooling overhead and enables a more energy-efficient environment for high-density AI and data-processing workloads.
Unlock exclusive market insights. Blog news—Download the Brochure now and dive deeper into the future of the Market
Market Stats
China’s AI compute: China hit 2,185 EFLOPS of AI compute by June 2026, with 52 ten-thousand-accelerator clusters nationwide but only 71.4% utilization.
Data center energy consumption: China’s data centers consumed roughly 260 TWh of electricity in 2024, about 2.67% of national demand, up from 161 TWh in 2018.
Global share: China accounts for 25% of global data center electricity consumption, second only to the US at nearly half.
Coal dependence: Coal supplies about 60.5% of China’s overall power mix, with roughly 70% of electricity in its data-center-heavy eastern regions still coal-based.
Renewable capacity: China’s renewable energy capacity reached 1.889 billion kW, or 56% of total capacity, by the end of 2024.
National strategy: China’s 2022 “East Data West Computing” plan aims to shift eastern data processing loads westward to align with renewable-rich regions.
Industry cooling benchmark: Hyperscale data centers average a PUE of 1.55 globally, while AI-specific clusters run closer to 1.2.
Liquid cooling gains: Liquid cooling techniques can cut data center energy use by around 30% compared to traditional air cooling.
Global AI energy trajectory: Global data center electricity use hit about 485 TWh in 2025 and is expected to roughly double to 950 TWh by 2030.
Two-country race: The US and China together are responsible for around 80% of all global data center growth.
The Green Pitch: Wind, Zero Freshwater, and a Smaller Footprint
The environmental proposition of the underwater data-center architecture combines offshore renewable power, seawater-based heat rejection and reduced land requirements. The computing modules are positioned near offshore wind-generation infrastructure, enabling electricity to be supplied from renewable generation with shorter transmission distances. Project specifications indicate that approximately 95% of electricity consumption can be supplied by offshore wind, while the cooling system is designed to account for less than 10% of total power consumption. Unlike conventional evaporative cooling systems, the underwater architecture does not require freshwater for heat rejection because waste heat is transferred through heat exchangers into the surrounding seawater. This can eliminate freshwater consumption associated with cooling operations. The system also substantially reduces the physical footprint required for computing infrastructure. Project estimates indicate more than 90% lower land requirements compared with a comparable conventional data center. Technically, the model therefore integrates renewable-energy generation, subsea thermal management and high-density computing into a compact infrastructure platform, addressing three major data-center constraints: electricity demand, cooling-water consumption and land availability.
Who’s Actually Building It
Projects of this scale don’t happen without serious institutional muscle behind them, and Lin-gang’s underwater data center is very much a state-aligned effort. The project is backed by a state-heavy consortium including Shenergy, China Telecom’s Shanghai branch, INESA, and CCCC Third Harbor Engineering. Notably, an executive overseeing the engineering side has been candid about how early this technology still is. Wang Shifeng, chairman of China Communications Construction Company’s Third Harbor Engineering division, acknowledged that construction of underwater data centers remains in its initial stages even as the sector moves toward commercial deployment. That kind of institutional honesty is notable, and it underscores that this is less a finished, mature technology than an evolving industrial bet one that fits neatly into China’s broader “Eastern Data, Western Computing” strategy aimed at rebalancing the country’s computing infrastructure across regions with different energy and land constraints.
What It’s Powering and Why the World Is Watching
The underwater data-center concept is being developed to support AI-intensive and data-heavy workloads, including big-data processing, data annotation, high-performance computing and the development of large language models (LLMs). The facility is designed around high-density GPU computing, with computing clusters from telecommunications and local computing-service providers supporting commercial workloads. Its planned capacity is reported at around 24 MW, while the initial demonstration phase has a capacity of approximately 2.3 MW. The technology combines subsea computing modules with seawater-based heat rejection, reducing dependence on conventional mechanical chillers. Reported performance figures indicate approximately 22.8% lower electricity consumption than comparable land-based data centers, while the system targets a PUE of around 1.15.
The project is also exploring integration with offshore renewable energy, creating the possibility of combining wind-generated electricity with naturally cooled computing infrastructure. Microsoft’s earlier Project Natick demonstrated the technical potential of subsea computing. Its Northern Isles module contained 864 servers across 12 racks and operated underwater for more than two years. Microsoft reported that the underwater servers experienced a failure rate one-eighth that of its land-based control group. However, Microsoft ultimately kept Natick as a research initiative rather than pursuing it as a commercial product. The newer commercial approach reflects growing interest in using underwater environments to address AI computing’s rising demands for electricity, cooling and land.
Unlock exclusive market insights. Blog news—Download the Brochure now and dive deeper into the future of the Market
Future Outlook
China’s underwater data center push is likely to scale rather than remain a one-off experiment, fitting neatly into a broader national strategy of pairing computing infrastructure directly with renewable energy sources. As AI compute demand keeps climbing, expect more coastal deployments tied to offshore wind farms, tighter efficiency targets, and continued government backing for this approach. If the Lin-gang facility’s performance holds up over time, China could position undersea, wind-powered data centers as a genuine export model and a template other countries look to as they grapple with the same water and power constraints. This could meaningfully reshape how the world thinks about sustainable AI infrastructure going forward.