
AI data centers and the power infrastructure supporting them could lift demand for metals, but grid delays, political opposition and more efficient designs may limit the build-out. A World Economic Forum estimate puts data-center demand in 2035 at 6% of global gallium use, 2.4% of germanium use and 1.2% of tin use.
Data centers consume an estimated 60 to 75 metric tons of metals per megawatt, and AI facilities can exceed 100 megawatts. Gallium, germanium and tin stand out in projected data-center demand: the World Economic Forum estimates they will account for 6%, 2.4% and 1.2%, respectively, of global annual use in 2035. Copper is projected to account for less than 0.9%.
Those estimates exclude the metals used to connect data centers to power grids and build energy infrastructure. Wood Mackenzie estimates that including power connectivity would make copper and aluminium demand three or four times higher than projections based only on data-center buildings, while annual power-capacity additions for data centers could rise from 15–20 gigawatts today to 30–33 gigawatts in the early 2030s.
Grid connections, political opposition and policy changes could slow construction. A study by the Oxford Smith School and Marex Group described some announced AI capacity as “bragawatts” and said deployment was likely to be delayed, uneven and constrained; the average US wait for a grid connection was four years from 2018 to 2023. The column also points to potential longer-term metal demand from robots using AI in physical settings.