Michael Meding

This is a good read by @Bloomberg, breaking down all the components that go into building a theoretical 1-gigawatt AI data centre. We’ve all seen the headline numbers before, but the required gigadollars are still hard to digest. $30 BILLION is equivalent to X2 new USS John F. Kennedy Aircraft carriers... for just one gigawatt of data centre capacity. ....and the US is expected to add more than 60 gigawatts of data center capacity over the next four years. In copper terms alone, that's >2 million tonnes of copper (at 35,000 tonnes of copper per gigawatt; source: @BMO). So this AI boom is just a metals boom in disguise. Capital expenditure by @Meta, @Alphabet, @Amazon and @Microsoft this year alone is expected to total $0.75 TRILLION.... increasing by+2.3 times to $1.75 TRILLION in 2027. These estimates are based on already-outdated metal prices. The bill will keep inflating, as the cost of finding and mining the required copper, nickel, palladium, silver, scandium, gallium, germanium, niobium, and grain-oriented electrical steel (GOES), to name a few, keeps rising. Supply of these metals is ALREADY in deficit, and their supply chains are increasingly balkanizing. So, Bloomberg's numbers are likely to be closer to the floor than the ceiling. @ivanhoeelectric @IvanhoeMines_ @IvanhoeAtlantic bloomberg.com/features/2026-…