The AI buildout story being sold to investors is fundamentally a software story dressed in hardware clothing. The narrative focuses on model capability, inference speed, and competitive positioning between foundation model labs. The physical infrastructure required to run those models at scale — and the material supply chains required to build that infrastructure — gets footnoted, if it appears at all.
Here are the numbers that belong in the headline.
Each of the 13-14 hyperscale data center campuses currently planned in the United States requires approximately 50,000 tons of copper just for electrical wiring and distribution infrastructure. That’s per campus. Multiply it out and you’re looking at 650,000 to 700,000 tons of copper for this buildout alone — before you account for the transmission infrastructure required to get power to these facilities, or the EV charging networks, or the re-shored manufacturing plants that are supposed to sit alongside them.
Global copper mine production runs at roughly 22 million tons per year. That sounds like plenty until you account for all the other demand: construction, automotive, consumer electronics, existing grid infrastructure. The copper market was already running structural deficits before the AI buildout was announced. The hyperscale data center program has added an enormous new demand category to a market with a 19-year supply response time.
Then there’s the power problem. You can’t run a data center without electricity. You can’t add electricity without transformers. Siemens’ transformer backlog is five years at current order rates. Gas turbines, required for dedicated on-site generation at many of these facilities, are fully allocated. The grid interconnection queue in most major U.S. markets runs 5-7 years.
Nvidia chips are being ordered and delivered. The buildings to house them are being designed. The copper to wire them doesn’t exist yet in sufficient quantity. The transformers to power them are five years out. Something in this chain is going to break, and when it does, the AI buildout narrative will collide publicly with the infrastructure reality that people paying attention have been watching build for two years. Position for that collision.
The Hedge 