The Copper Squeeze Behind the AI Buildout

The global copper market is heading toward one of its largest supply deficits on record, and the driver this time is not a mining strike or a war. It is artificial intelligence. A single large data center can require tens of thousands of tonnes of copper once the surrounding grid and power infrastructure are counted, and forecasters now expect data centers alone to consume hundreds of thousands of tonnes of the metal annually. The headline framing is a story about technology stocks. The more grounded signal is that a very old, very physical metal has become the binding constraint on a very modern buildout, and that constraint does not resolve itself the way a software bottleneck does.

The Historical Echo

The United States has had to ration copper for a strategic priority before, though the priority in question was war rather than computing. When the country entered the Second World War, copper became essential for shell casings, communications wire, and naval equipment, and the demand from the war effort collided directly with the everyday uses of the metal, right down to the one-cent coin in every American's pocket. Congress asked the Mint to find a substitute, and in 1943 the United States struck its pennies not in copper but in zinc-coated steel, the only year in the coin's history that copper was removed from it entirely. A tiny handful of copper planchets were struck by accident that year and remain among the rarest error coins in American numismatic history, a small physical record of just how tightly the metal was being rationed.

The parallel is not exact, a data center is not a wartime shell factory, but the underlying dynamic is the same one playing out again now. A strategic priority, whether winning a war or building the computing infrastructure of the next decade, creates a call on a physical resource that outpaces the ordinary channels supplying that resource. In 1943 the response was substitution at the margins, steel in place of copper for a low-value use, while the metal itself was redirected to where it mattered most. Today's response is unfolding through price, as copper has already posted some of its sharpest rallies in modern trading history, and through years of underinvestment in new mine supply now colliding with demand that shows no sign of slowing.

Where Patient Capital Is Positioning

For long-horizon capital, the copper story reinforces something that gold and silver have already been signaling in their own markets: physical scarcity in a foundational metal does not announce itself gradually. It tends to arrive as a structural deficit that persists for years, because new mines take the better part of a decade to permit and build, while demand from a technology shift can arrive within a few product cycles. The 1943 steel penny is a curiosity today, but it captured a real moment when a physical resource was scarce enough that the smallest, most routine use of it had to be sacrificed for the larger priority.

The AI buildout is creating a version of that same dynamic at a much larger scale, and it is happening alongside the increased industrial demand already stressing the silver market and the reshoring push already underway in rare earth minerals. None of these are speculative bets on a single company or technology. They are claims on physical metals that a modern, electrified economy cannot function without, regardless of which specific companies end up winning the AI race itself. That distinction, between owning the technology and owning the tangible inputs the technology depends on, is exactly the kind of signal long-horizon capital is built to notice.

Copper will not disappear from circulation the way it briefly did from the penny in 1943. But the same underlying principle applies across both eras: when a strategic priority outruns the supply of a physical resource, the resource itself becomes the scarce and valuable thing, whatever headlines happen to be written about the technology sitting on top of it.

The Capital Memo