AI data centres shift the metals question toward the power grid

A Reuters Open Interest analysis notes that AI data centres use roughly 60 to 75 metric tons of metals per megawatt, compared with much smaller conventional facilities. Gallium, germanium and tin could see the largest direct demand impact because of their roles in semiconductors and circuit boards.

The analysis argues that transmission lines, substations and new power generation could multiply copper and aluminium needs well beyond estimates for the data-centre buildings alone. It also cautions that grid connections can take years, announced projects may not all proceed and more efficient designs could reduce material intensity.