When CEO Dan Roberts said the company is only using ~2% of its total capacity, it instantly clarified why $IREN is one of my core horses in the AI race. If that tiny slice of the footprint can already support 23K GPUs then scaling toward full utilization means they can expand compute by an order of magnitude without rebuilding anything.  That is a completely different challenge than the rest of the sector which has to construct, permit & power entirely new infrastructure just to keep up.