As the artificial intelligence boom drives unprecedented demand for data centers, developers are testing a radical solution: submerging data centers on the ocean floor to improve energy use and cooling efficiency, and use less fresh water and land area. Yet, while marine-based infrastructure promises relief for overburdened electrical grids, it introduces complex marine thermal pollution risks and severe hardware maintenance hurdles.
The Genesis of Ocean-Based Server Pods
The push to relocate server infrastructure offshore is not entirely new. In August 2014, Microsoft launched an internal initiative known as Project Natick, according to IEEE Spectrum reporting. Spearheaded by engineers including Sean James—who drew on his background serving aboard a U.S. Navy submarine—the project aimed to limit the cost of cooling the machines.
By August 2015, Microsoft had deployed a prototype beneath the Pacific Ocean. Years later, in 2018, the company set up a waterproof data center holding 864 servers on the seafloor near Scotland’s Orkney Islands, linked to the mainland via an underwater cable. After two years, Microsoft reported a specific metric:
The underwater servers failed at about one-eighth the rate of comparable land-based hardware. Microsoft’s theory was that within a sealed marine setup, the machinery experiences less exposure to humidity, temperature swings, and oxygen, along with reduced physical bumping from personnel swapping out broken parts.
Despite these promising operational efficiencies, Microsoft ended the project in 2024 without expanding the initiative. Analyses by other observers indicate that factors like regulatory requirements—such as securing environmental permits—and the need for quicker equipment upgrades and component swaps might be responsible.
Global Deployments and Alternative Marine Architectures
Even as Microsoft pulled back, international engineering teams accelerated deployment schedules. China built what may be the world’s first wind-powered underwater data center in Shanghai, beginning full commercial operations in May 2026 following a $226 million investment.
This facility leverages seawater as a coolant rather than refrigerating fresh water. According to the project, the setup uses at least 30% less electricity than traditional data centers, while offshore wind turbines reduce reliance on fossil fuels and cut the data center’s carbon emissions.
Other nations are experimenting with floating marine infrastructure rather than total submersion:
- Japan: Opened a data center in shipping containers on a floating platform near Yokohama in 2025, utilizing onboard solar panels and battery storage. Testing on this platform runs through March 2027.
- Singapore: Infrastructure firm Keppel began building a four-story floating data center in 2026, scheduled for a 2028 opening to bypass limited land availability while utilizing seawater cooling.
- South Korea: Ulsan began planning in 2025 for an underwater data center designed to house more than 100,000 servers while using 30% less power than land-based centers.
- Canada: A submersible artificial intelligence facility called DeepGreen Western Passage was put forward for the Bay of Fundy, utilizing tidal turbines built to capture the region’s powerful water flows.
- Portugal: Sines plays host to the SIN01 AI data center, which channels Atlantic Ocean seawater to cool its computer servers before discharging it back into the sea.
The Thermal Footprint and Marine Ecosystem Risks
Moving workloads offshore does not make underlying environmental challenges such as energy consumption and carbon emissions disappear. While seawater cooling reduces grid reliance, the heat generated by servers must go somewhere.
Operational data from existing installations reveals measurable thermal outputs. HiCloud, the engineering contractor for China’s Hainan underwater data center, reported a temperature increase of less than 1 degree Celsius (1.8 degrees Fahrenheit) in the surrounding seawater. Similarly, the SIN01 facility in Portugal returns seawater about 1 C warmer.
While a one-degree delta may sound minor, many marine species depend on stable water temperatures for breeding, feeding, and migration patterns. Given that UNESCO estimates about 60% of marine ecosystems are already degraded or used unsustainably, cumulative thermal pollution from multiple underwater data centers could create localized thermal pollution and alter marine ecosystems.
The Logistics Bottleneck of Remote Infrastructure
Beyond ecological concerns, engineering a “lights out” environment—as described by IEEE Spectrum—introduces severe logistical constraints. When a computer fails underwater, it cannot be repaired or replaced on site.
The entire sealed data center module may need to be brought to the surface simply to service a single failed component. Furthermore, IEEE Spectrum notes that barnacles and other forms of sea life can cover a submerged vessel, which could interfere with the transfer of heat from the servers to the surrounding water.
Yet, geographic realities continue to drive interest in coastal placements. A Gallup poll conducted in March 2026 revealed that 70% of Americans oppose building AI data centers in their communities. With more than half of the world’s population residing within 120 miles of a coast, placing server infrastructure underwater could be another way to keep data centers physically close to users for speedy service.
Whether ocean-based infrastructure matures into a sustainable alternative for the AI industry depends on solving remaining economic, technical, and environmental problems.
