Ai data center capacity will expand rapidly. With limited land and fresh water available for coolant, creative solutions will lower costs and save our limited resources.
China’s Highlander has launched an underwater data center in Hainan, the first project like it to be deployed at commercial scale. The energy demands of Ai data centers come largely from the cooling requirements of the servers, and China has answered this question by placing their Ai data centers underwater. This is something that’s been successfully experimented with by Microsoft among others but hasn’t been deployed commercially before now. The benefits of submerged data centers include a reduction of 90% in cooling costs as the ocean’s currents do what fresh water would otherwise; this translates into 40% more compute than a comparable land-based system. When coupled with renewable energy the costs drop further as this project is largely powered by an adjacent offshore windfarm; the company says that it’s 95% powered with renewable energy. Ai and Water
Microsoft and others have built underwater data centers, but no projects of this caliber are currently operating in the West. Anything that saves water is welcome as current land-based data centers are often built in arid regions and require massive water withdrawals that stress already water stressed regions. The current trajectory is toward continued water disregard, as the recent batch of data centers announced by OpenAI will be built in Texas, New Mexico, and somewhere in the Midwest. These states are not water abundant, and any withdrawals will come at the expense of local residents’ drinking water and competing industries such as agriculture.
