China launches the first commercial subsea data center powered by offshore wind

Home » Python » China launches the first commercial subsea data center powered by offshore wind

On the seabed of the East China Sea, just off the coast of Shanghai, China has put into operation a new type of data center that combines subsea deployment and renewable energy. The Lingang Subsea Data Center is not a lab prototype or a short-lived pilot: it is a commercial facility that already serves real workloads for artificial intelligence, cross‑border e‑commerce and digital logistics.

By placing computing modules underwater and feeding them almost entirely with offshore wind, the project brings together two long‑term national goals: digital sovereignty and carbon neutrality. In a context where AI and big data push electricity and water consumption to record levels, this installation tests how far data infrastructure can go in cutting its environmental footprint without sacrificing performance.

Location, scope and commercial nature of the project

The Lingang Subsea Data Center is located in the Lingang Special Area of the Shanghai Pilot Free Trade Zone, roughly ten kilometres from the shoreline. The first construction phase has already been completed and is in service, providing 2.3 MW of installed IT capacity housed inside a vertical subsea module connected to an onshore control center.

Unlike earlier experimental pods, this facility is designed from the outset as commercial infrastructure. From day one it has been processing live traffic related to AI systems, cross‑border online retail and logistics services, rather than synthetic or test workloads. Chinese authorities have recognised the installation as a national low‑carbon computing model, underlining its role as a reference project rather than a small‑scale demo.

The long‑term roadmap foresees scaling from the initial 2.3 MW to a total of 24 MW across two phases, backed by an investment of around 1.6 billion yuan (in the range of 200-226 million US dollars). For comparison, that still puts it below the largest hyperscale facilities on land, which can reach 50-500 MW, but it already moves clearly beyond a mere proof of concept.

Behind the project is Shanghai Hicloud Technology (also known as Hailan Cloud or HiCloud), supported by the Lin‑gang Management Committee and Lingang Investment Holding Group. The initiative has also attracted state‑linked partners such as Shenergy Group, China Telecom Shanghai, INESA and CCCC Third Harbor Engineering, highlighting its strategic importance for Beijing’s broader industrial and energy policies.

Subsea architecture and high‑density computing

At the heart of Lingang, a large steel cylinder sits on the seabed of the East China Sea, acting as a pressurised, watertight hull for clusters of high‑performance servers. These servers are organised in modular racks, with aggregate computing power roughly comparable to tens of thousands of high‑end gaming PCs, enough to sustain thousands of parallel AI conversations or other intensive applications.

The interior of the module is filled with inert gases to reduce corrosion and fire risk, while the structural design seeks to maximise usable volume and minimise the impact of waves and currents. Radiators placed behind the racks evacuate heat, assisted by pumps circulating seawater through dedicated channels, turning the surrounding ocean into a large, stable heat sink.

Installation on the seabed was a complex marine engineering operation. The support structure’s legs and the steel piles driven into the seabed had to align with clearance tolerances of just a few centimetres. Using GPS guidance and the heavy‑lift vessel Sanhang Fengfan, the engineering team managed to lower the module with virtually zero deviation relative to the planned position, according to local media reports such as Shanghai Daily.

Communication and power are provided through two 35 kV submarine cables that link the subsea module with offshore wind turbines near Shanghai and with onshore control and grid infrastructure. This dual‑cable layout ensures both redundancy and enough transmission capacity for data and electricity.

Cooling with seawater: efficiency without freshwater

One of the least visible but most pressing issues in the data‑center industry is water consumption for cooling. A conventional mid‑scale facility on land can evaporate millions of litres of freshwater per year just to keep servers within their operating temperature range. By moving the hardware underwater, the Lingang project takes advantage of the ocean’s thermal properties.

Here, the surrounding seawater acts as a free, continuous heat sink. Instead of relying on large industrial chillers and cooling towers, the system pumps cold seawater through radiators that absorb heat from the server racks before returning the water to the sea, having transferred warmth away from the electronics. This design removes freshwater from the equation entirely.

The impact on efficiency shows up clearly in the sector’s standard indicator, Power Usage Effectiveness (PUE), which compares total facility power use with the portion consumed directly by IT equipment. While typical modern land‑based centers operate around PUE 1.5-1.6, Lingang has been engineered to stay at or below a PUE of 1.15. That figure translates into a reduction of around 22.8% in overall electricity use compared with efficient terrestrial sites.

Cooling costs, which can account for 40-50% of the electricity bill of a traditional data center, are estimated to fall by up to 90% thanks to passive and semi‑passive seawater cooling. Beyond cutting costs, this setup also helps stabilise operating temperatures, a key factor in extending hardware life and reducing unexpected failures.

Offshore wind as primary power source

The other pillar of the Lingang model is its energy supply. Instead of drawing primarily from the terrestrial grid and supplementing with renewable certificates, the facility is physically tied into adjacent offshore wind farms. More than 95% of its electricity comes from these marine turbines, making it one of the first data centers worldwide to be powered almost exclusively by offshore wind.

In practical terms, the subsea data center functions as a constant baseload consumer for nearby wind assets. One longstanding problem with wind power is that its generation pattern depends on weather rather than demand, which often leads to curtailment when there is no immediate use or storage capacity. Directly attaching high, steady computing loads to the turbines helps soak up production that might otherwise go to waste.

The link between generation and consumption is reinforced by close coordination between wind‑farm operators, grid managers and the data‑center control systems. Workloads can be shifted or modulated to better match the availability profile of wind output, turning power management into a joint optimisation problem rather than a one‑way supply arrangement.

Beyond the initial 24 MW plan, project partners have signed cooperation agreements around offshore wind development of up to 500 MW tied to digital infrastructure. Exact timelines and site details for this potential larger build‑out have not been disclosed, and for now the figure looks more like a strategic ambition than a fully committed construction schedule.

Land, urban density and marine economy

Space is another piece of the puzzle. In a metropolis like Shanghai, urban land is a scarce and expensive resource. Data centers require large footprints, secure perimeters and access to robust power and connectivity, which often clashes with other development priorities in crowded metropolitan areas.

By moving the server modules to the seabed, the Lingang project reduces land occupation by more than 90% compared with an equivalent onshore site. Only the control building, grid interconnection point and some auxiliary equipment remain on land, freeing up valuable real estate and easing conflicts between infrastructure and housing or commercial uses.

The initiative also fits neatly into China’s ambition to grow what it calls the “marine economy” alongside the digital and new energy economies. Converting offshore areas into hubs for both energy and computing turns previously underutilised marine space into productive infrastructure zones, while still leaving room for other maritime activities under appropriate planning frameworks.

Institutionally, the subsea data center has been included in the Green and Low‑Carbon Technology Demonstration Project list of the National Development and Reform Commission (NDRC), China’s top economic planning body. This listing signals that the central government views the project as a showcase for how advanced infrastructure can align with climate and industrial policies.

Comparison with Microsoft’s Project Natick

For observers outside China, the obvious reference point is Microsoft’s Project Natick, an experiment that immersed a sealed server pod off the coast of Scotland between 2018 and 2020. Natick delivered notable technical results: out of 864 servers, only eight failed during the trial, a much lower failure rate than similar land‑based systems, and the experimental pod achieved an impressive PUE of around 1.07.

Despite those metrics, Project Natick remained in the realm of R&D. Once the test period ended, Microsoft recovered the pod and eventually shelved the concept, citing open questions around long‑term costs, logistics and maintenance at scale. The company never moved to roll out commercial subsea facilities for customers.

Lingang, in contrast, has been conceived from the start as a revenue‑generating, production‑grade data center. It operates with real client workloads, plugs into the regional digital ecosystem and is integrated with China’s broader push for green, resilient infrastructure. That distinction—experimental versus operational—marks a significant qualitative shift.

There are other differences as well. Natick did not rely on a tightly coupled offshore renewable supply and remained limited to a single pod of roughly a dozen racks. Lingang begins at multi‑megawatt scale, targets future expansion to tens of megawatts and is directly linked to dedicated offshore wind assets. Both projects exploit seawater for cooling, but the Chinese installation embeds that principle into a complete commercial design, from energy flows to regulatory approvals.

Technical and operational challenges under the sea

Even with its advantages, operating high‑value infrastructure on the seabed comes with significant challenges. One of the most obvious is maintenance and physical access. Any intervention that cannot be handled via remote management systems may require specialised vessels, remotely operated vehicles (ROVs) or even diver support, which raises both costs and lead times for repairs.

Corrosion is another key concern. The marine environment is aggressive for metals, seals and electronic components. Although the steel cabins are filled with inert gas and built with corrosion‑resistant materials and coatings, the long‑term behaviour of such structures over 10-20 years under continuous exposure remains partly untested in commercial conditions.

Geography also limits where this model can be replicated. Successful deployment requires the right combination of suitable coastline, sufficient and relatively steady offshore wind, appropriate seabed conditions and proximity to demand centres. Not every country or region will tick all of these boxes, which may confine large‑scale adoption to specific coastal corridors.

Connectivity and latency issues cannot be ignored either. While fibre‑optic cables can deliver high throughput, distance to urban hubs and backbone networks still shapes response times. For latency‑sensitive applications, careful planning is needed to ensure that subsea installations do not introduce unacceptable delays compared with land‑based edge facilities located closer to end users.

Economic scale, roadmap and open questions

From a capacity standpoint, the first 2.3 MW stage of Lingang is modest compared with large hyperscale data centers, which often start at tens of megawatts and can grow several times larger. In that sense, the current configuration can be seen as a bridge between a pure trial and a fully mature subsea campus.

The roadmap calls for a phased expansion to 24 MW of total capacity, but there is no publicly confirmed date for completing the second construction stage. Beyond that, the cooperation agreement around 500 MW of wind‑linked computing projects points to an ambitious horizon, though details on locations, financing and technical configurations remain sparse.

Cost structure is another area with limited transparency. While developers highlight substantial savings in energy and water, long‑term maintenance, periodic overhauls and eventual retrieval or replacement of subsea modules will all add to the total cost of ownership. To date, no comprehensive public breakdown of these lifecycle expenses has been released.

Unresolved questions also extend to operational protocols for hardware upgrades and decommissioning. Servers typically have three‑to‑five‑year refresh cycles in standard data centers. Adapting that rhythm to a sealed, underwater environment will likely require careful planning to avoid frequent, costly marine operations, or a shift to longer replacement cycles with more robust equipment.

Implications for AI, cloud providers and startups

Globally, the digital economy is pushing infrastructure to its limits. Data centers already account for roughly 1-2% of total electricity consumption worldwide, and that share is climbing as AI, streaming and cloud services expand. Training and running large‑scale language models, recommendation systems and real‑time analytics all demand huge volumes of compute.

In that context, the Lingang approach offers a concrete template for scaling computing power without linearly increasing carbon emissions or freshwater use. For cloud providers, this kind of facility could lower operating costs and improve environmental metrics, both of which are increasingly important for regulators, investors and enterprise clients with climate commitments.

For startups building on top of cloud infrastructure—particularly those focused on generative AI, real‑time processing or high‑availability services—the emergence of low‑PUE, renewables‑powered data centers may eventually influence pricing and sourcing decisions. Over time, competitive pressure could nudge more providers to adopt similar designs or at least invest more heavily in clean‑energy supply and advanced cooling.

Sustainability has also shifted from a nice‑to‑have to a core requirement in many markets. Environmental, social and governance (ESG) criteria, regional carbon rules and customer expectations are converging to push infrastructure operators toward greener models. A subsea center that combines PUE around 1.15, nearly 100% renewable power and zero freshwater use fits squarely into that trend.

Regions with strong offshore wind potential and major coastal cities—such as parts of Latin America, Northern Europe or East Asia—may look closely at the Lingang case. While replicating the exact Chinese setup would require tailored regulation, financing and local engineering capacity, the basic idea of pairing offshore renewables with marine data infrastructure could inform future national or regional strategies for digital and energy planning.

As things stand, China’s Lingang Subsea Data Center functions as a high‑profile testbed for the next generation of green data infrastructure. With its underwater server modules, mostly wind‑powered electricity, seawater‑based cooling and multi‑stage expansion plan, the project shows one possible pathway for reconciling the rapid growth of AI and digital services with tighter constraints on energy, water and land—while leaving open technical, economic and regulatory questions that other players will be watching closely.