AI companies look to the ocean as a place to put more data centers
Moving servers into the ocean does not make underlying environmental challenges disappear. And it creates new concerns as well.
The rise of artificial intelligence is driving a surge in demand for data centers, leading companies to seek alternative locations that could address energy consumption, water use, and carbon footprint concerns. While traditional land-based data centers face challenges, the ocean has emerged as a potential new site for AI infrastructure.
Microsoft's underwater data center experiment in Scotland demonstrated that servers in such an environment experienced fewer failures compared to land-based counterparts, suggesting potential energy and cooling efficiency benefits. However, the project was halted in 2024 due to regulatory concerns and the desire for faster equipment upgrades.
Meanwhile, China, Japan, and South Korea are exploring different approaches to ocean-based data centers. China's wind-powered underwater data center in Shanghai uses seawater for cooling and reduces electricity consumption by 30% compared to traditional data centers. Japan is testing floating data centers housed in containers that utilize seawater for cooling and solar power, freeing up scarce land space.
Singapore plans to build a four-story floating data center that will also use seawater for cooling and take advantage of floating capabilities. South Korea is planning an underwater data center with over 100,000 servers that would use 30% less power than land-based facilities by employing seawater cooling. Meanwhile, Portugal's SIN01 AI data center in Sines employs seawater cooling from the Atlantic to reduce power consumption before returning the water to the ocean.
These projects highlight the potential of ocean-based data centers to improve energy efficiency and reduce environmental impact, but also highlight the need to address regulatory challenges, maintain and expand infrastructure, and ensure sustainable practices.
Written by urgent.news from The Conversation's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
