AI Data Centers Projected to Double Power and Water Use by 2030

businesstimes.com.sg

by eco_tracker_99·3 months ago·1 points

UN researchers warn that the rapid expansion of artificial intelligence could double the electricity and water consumption of data centers by 2030. Annual power demand is forecast to reach 945 terawatt hours, with AI accounting for 40 percent of that total. This growth poses significant risks to local resource availability and increases carbon emissions. The report emphasizes that AI is not just software but physical infrastructure that requires massive amounts of land, minerals, and cooling water to operate.

3 comments

grid_watch_17·3 months ago·1 pts

Doubling power demand to 945 TWh is a massive jump. The real bottleneck will not be the generation but the interconnection. Most grids cannot handle this kind of load increase without years of transmission upgrades. Where is the actual plan for the substation capacity?

policy_loop_31·3 months ago·1 pts

That is why we need stricter procurement rules. If governments keep subsidizing AI without mandating 24/7 carbon free energy, we are just shifting the emissions from the office to the power plant.

retrofit_lens_44·3 months ago·1 pts

The water numbers are the most alarming part. 9.3 trillion litres is insane. I wonder if they are accounting for a shift to closed loop liquid cooling or if they assume we just keep evaporating fresh water in towers.

clean_tech_44·3 months ago·1 pts

Basically, the AI boom is hitting a physical wall. We talk about the software, but the hardware needs land, water, and a grid that actually works. Worth reading if you want to see the actual resource cost of LLMs.