Reading time: 5 min
LME copper closed at $12,075 per tonne on March 26, down 17% from the all-time high of $14,527 struck on January 29. The sell-off has been relentless. The fundamentals behind the rally have not changed at all.
That’s the disconnect your book needs to price. Copper didn’t fall because the world suddenly needs less of it. It fell because the Iran war repriced energy costs, crushed risk appetite, and forced leveraged longs to liquidate into a market that was already nervous about Chinese demand. On COMEX, the front-month contract sits at $5.56 per pound, still trading at a record premium to the LME because American importers spent all of 2025 stockpiling cathodes ahead of potential Section 232 tariffs. That premium hasn’t collapsed. The metal locked in U.S. warehouses isn’t going anywhere.
What the Chart Says
The January spike to $14,527 came on the back of the Grasberg force majeure, Chile’s production miss, and a wave of speculative buying tied to AI infrastructure spending. Then the correction started before the first missile hit Iran. By mid-February, copper had already pulled back to $13,000 on profit-taking and weak Chinese PMI prints. The war accelerated the move. Brent above $100 means higher smelting costs, tighter margins, and weaker industrial activity in import-dependent Asia. S&P Global’s March commodity watch flagged aluminum as the metal most directly hit by the conflict, but copper caught the second-order effects through demand destruction fears and a surging dollar.
The 61.8% Fibonacci retracement from the October 2025 low to the January high sits right around $12,000. That’s where copper found a floor last week. If you’re watching the technicals, that level matters.
The Deficit Hasn’t Disappeared
Strip away the war noise and the supply picture is tighter than it was a year ago. J.P. Morgan projects a global refined copper deficit of 330,000 tonnes in 2026, driven by mine supply growth of barely 1.4%, roughly 500,000 tonnes below what the bank was forecasting at the start of 2025. JPM’s Shearer pointed to the Grasberg shutdown, which took 70% of the Block Cave’s output offline after September’s mudslide, and warned that full production won’t resume until the second quarter. Freeport says 85% of operations should be restored by the second half of the year. That’s optimistic guidance from a company that has missed timelines before.
Chile is trying to fill the gap. Cochilco expects national output to reach 5.61 million tonnes in 2026, up roughly 100,000 tonnes from the current run rate. Thirteen projects worth $14.8 billion are hitting milestones this year, per Mining.com, including Codelco’s Rajo Inca, Capstone’s Mantos Blancos, and the Collahuasi upgrade run by Anglo American and Glencore. Seven of those projects are scheduled to begin production, adding nearly 500,000 tonnes of annual capacity on paper. But Guzmán at GEM warned that none of them will ramp to full capacity immediately. Community relations, not commodity prices, remain the binding constraint. Peru’s output is roughly flat at 2.7 million tonnes, per INN data. The $18 billion Argentine copper pipeline is still years from first production.
The AI Demand Engine
Here’s the part that gets lost in the Iran headlines. J.P. Morgan estimates that data centre installations alone will consume 475,000 tonnes of copper in 2026, an increase of roughly 110,000 tonnes year-on-year. Amazon Web Services signed a two-year offtake agreement with Rio Tinto in January for domestically produced copper from an Arizona mine. That was one of the first direct links between low-carbon copper and AI infrastructure. It won’t be the last.
The International Energy Agency estimates that data centres currently consume about 1.5% of global electricity supply, roughly equivalent to the entire United Kingdom. By 2030, that figure is expected to more than double, with AI responsible for most of the increase. Every megawatt of new capacity needs copper for cabling, bus bars, transformers, and cooling systems. There is no viable substitute at scale. Goldman Sachs called copper the commodity with the highest growth potential this year, though their price target of $5.17 per pound, equivalent to roughly $11,400 per tonne, sits well below current spot. The gap between Goldman’s caution and JPM’s $12,500 target tells you that even the desks can’t agree on how much of the AI premium is already priced in.
The Positioning
COMEX warehouse stocks stand above 453,000 tonnes, a record, but that inventory is effectively locked in by the CME-LME premium arbitrage. It isn’t available supply in any meaningful sense. Chinese smelters, meanwhile, have agreed to cut refined output by more than 10% in 2026 after treatment and refining charges fell into negative territory, per Goldman. That means smelters are literally paying miners to receive concentrate. When the processing fee goes negative, supply is not abundant. It is desperate.
The International Copper Study Group sees a 2026 deficit of 150,000 tonnes, more conservative than JPM’s 330,000 but still a shortfall in a market where new mine approvals have averaged under 300,000 tonnes annually for three years running, per Wood Mackenzie. Annual demand growth requires 600,000 to 700,000 tonnes of new supply. The maths don’t work. They haven’t worked for a while.
If you’re long copper here, the case hasn’t broken. The $12,000 floor held, the deficit is structural, and the AI demand story is accelerating faster than any mine can be permitted. If you’re waiting for a cleaner entry, watch the Strait of Hormuz. The moment energy costs stabilise and the industrial demand outlook clears, the metal that fell 17% on sentiment is going to reprice on fundamentals. The deficit doesn’t care about geopolitics. It cares about tonnes. And there aren’t enough of them.
