Underwater data centers emerge as a solution to AI's land and cooling constraints
As AI demand grows, companies are exploring ocean-based data centers to improve cooling and reduce land use. Microsoft's experiment showed lower failure rates, but the project was discontinued. Challenges remain, including environmental impact and regulatory issues.
Microsoft's 2018 Orkney Islands trial housed 864 servers on the seafloor, reporting failure rates roughly one-eighth those of land-based equivalents, likely due to reduced exposure to oxygen, humidity, and temperature swings. Despite this promising result, the company discontinued the project in 2024, with analysts pointing to permitting hurdles and the practical difficulty of upgrading sealed equipment.
China's Shanghai facility, operational since May 2026, represents a $226 million investment combining seawater cooling with offshore wind power, cutting electricity use by at least 30 percent. Japan's floating container-based test near Yokohama runs through March 2027, while Singapore's Keppel is constructing a four-story floating data center slated for 2028 opening.
Underwater data centers could reshape where and how digital infrastructure is built, potentially easing pressure on land-scarce regions and reducing freshwater consumption. Coastal communities may see new economic activity, but marine ecosystems could face disruption from heat discharge and cable installation. Energy demands remain substantial, meaning the climate benefits depend heavily on renewable integration. Regulatory frameworks are still nascent, which may slow adoption or create uneven standards across jurisdictions.