The rapid growth of artificial intelligence is creating unprecedented demand for data centers, intensifying concerns over electricity consumption, water use and carbon emissions. In response, technology companies and infrastructure developers are increasingly considering locations beyond traditional land-based facilities, including the ocean.
Underwater and floating data centers could offer several advantages, such as more efficient cooling, reduced freshwater consumption and a smaller land footprint. However, moving computing infrastructure offshore also introduces new technical, environmental and regulatory challenges that could determine whether the concept becomes a viable long-term alternative.
Researchers studying the environmental and social effects of AI infrastructure say ocean-based facilities could help accommodate the technology’s expansion, but they would not eliminate the industry’s fundamental environmental problems.
Microsoft’s underwater experiment
Microsoft was among the first major technology companies to investigate underwater data centers. The company launched its research project in 2015 and placed a sealed data center containing 864 servers on the seabed near Scotland’s Orkney Islands in 2018.
The facility was connected to land through an underwater cable. After two years of operation, Microsoft reported that its servers experienced failures at roughly one-eighth the rate recorded at comparable land-based facilities.
The company suggested that the sealed environment may have reduced exposure to oxygen, humidity and temperature changes. The absence of workers regularly entering the facility to replace equipment may also have reduced physical disturbances.
Microsoft eventually ended the project in 2024 and did not proceed with additional underwater facilities. Although the company did not disclose the reasons, analysts have pointed to possible regulatory requirements, including environmental permits, as well as the difficulty of rapidly upgrading or replacing equipment underwater.
Microsoft has instead continued developing land-based data centers, which can be expanded more easily and provide simpler access for maintenance.
China, Japan and Singapore pursue new models
Other countries have continued experimenting with ocean-based computing infrastructure.
China’s Shanghai facility, which began commercial operations in May 2026 after launching in June 2025, is described as potentially the world’s first wind-powered underwater data center. The US$226 million project uses seawater to cool its servers, reducing the need for energy-intensive freshwater cooling systems.
According to the project, the facility consumes at least 30% less electricity than conventional data centers, while offshore wind power reduces its dependence on fossil fuels and lowers emissions.
Japan is testing floating facilities instead. In 2025, a data center housed in shipping containers was installed on a floating platform near Yokohama. Solar panels on the platform provide electricity, while batteries store energy. The system uses seawater for cooling and is scheduled to remain under testing until March 2027.
Singapore is also developing a larger floating facility. Infrastructure company Keppel began construction in 2026 on a four-story floating data center that is expected to open in 2028. The project will use seawater for cooling and is designed to avoid consuming Singapore’s scarce land resources.
South Korea has also begun planning an underwater data center in Ulsan. The proposed facility could accommodate more than 100,000 servers and is expected to consume 30% less power than conventional land-based facilities by using seawater for cooling.
In the United States, DeepGreen Western Passage has proposed placing a submersible AI data center in Maine’s Bay of Fundy. The facility would be powered by tidal turbines designed to take advantage of the area’s strong tidal currents.
Seawater cooling could also benefit land-based facilities
Ocean-based infrastructure is not the only way companies are attempting to use seawater for data center cooling.
In Portugal, the SIN01 AI data center in Sines draws seawater from the Atlantic to cool its servers before releasing the water back into the ocean.
These approaches could reduce both electricity consumption associated with cooling and the amount of freshwater required by data centers.
Ocean locations could also help address opposition to large data centers in residential areas. A Gallup poll conducted in March 2026 found that 70% of Americans opposed building AI data centers in their communities.
At the same time, more than half of the global population lives within 120 miles of a coastline. Locating computing facilities offshore could therefore keep them relatively close to users while reducing their presence in densely populated communities.
Maintenance remains a major obstacle
Despite the potential benefits, maintaining underwater computing infrastructure presents significant difficulties.
A failed server cannot simply be accessed and replaced by a technician when it is located beneath the sea. In some cases, an entire sealed module could have to be brought to the surface for repairs, even if only a single computer has malfunctioned.
Such limitations could make underwater facilities less flexible than conventional data centers, particularly as AI hardware develops rapidly and operators need to upgrade servers more frequently.
Environmental concerns move underwater
The biggest environmental question surrounding ocean-based data centers may be the effect of their cooling systems on marine ecosystems.
Using seawater to remove heat from servers means that warmer water is eventually released back into the surrounding environment. This can potentially affect local oxygen levels, pH and marine organisms.
At China’s Hainan underwater data center, engineering contractor HiCloud has reported that seawater temperatures near the facility have risen by less than 1 degree Celsius. The SIN01 facility in Portugal also returns seawater approximately 1 C warmer than when it entered the cooling system.
Although these temperature changes may appear small, marine species often rely on relatively stable water temperatures for breeding, feeding and migration. A larger number of underwater data centers could therefore create localized thermal pollution and potentially disrupt marine ecosystems.
The concern is particularly significant because marine environments are already under substantial pressure. UNESCO estimates that around 60% of marine ecosystems have been degraded or are being used unsustainably.
Meanwhile, oceans are already warming as a result of climate change. Additional heat from data centers could add to existing pressures on coral reefs, mangroves, seagrass and other parts of marine ecosystems.
Rising ocean temperatures could also eventually make seawater a less effective cooling resource for data centers in some parts of the world.
Ocean-based data centers could provide AI companies with a way to reduce pressure on land, electricity supplies and freshwater resources. Whether they can deliver those benefits without creating a new set of environmental problems in the world’s oceans will determine whether underwater computing becomes a lasting part of the AI infrastructure landscape.
Be First to Comment