Copper futures on the London Metal Exchange have climbed to around 14,700 dollars a metric ton, with Comex prices setting record highs near 6.74 dollars a pound, taking the metal's gain this year to nearly 18 percent. Rallies in industrial metals usually tell a story about one thing: a strike, a smelter outage, a demand spike. This one is telling three stories at once, and that is what makes it durable.
Demand: the grid behind the data centre
The first story is the AI buildout. Every data centre campus arrives with a power problem attached, and power problems are copper problems: transmission lines, transformers, switchgear and cabling. Layered on top are grid modernisation programmes, renewable-energy expansion and vehicle electrification, all of them copper-intensive. S&P Global expects global demand to surge from 28 million tons in 2025 to 42 million tons by 2040, and warns of a potential 10 million ton annual shortage by 2040 without significant new supply.
Supply: older, deeper, slower
The second story is that supply is going the wrong way. Global mine production fell 1.1 percent in the first half of 2026, with output at Chile's state giant Codelco down 11 percent year on year. Average ore grades have declined roughly 40 percent since 1991, development costs have risen 65 percent since 2020, and a new mine now takes an average of 17 years from discovery to production. Even at record prices, only around 60 percent of identified new projects pencil out as profitable. The International Energy Agency projects a 25 percent supply deficit by 2035.
The third story is policy distortion. The prospect of US copper tariffs, potentially 15 percent from January 2027 rising to 30 percent in 2028, has pulled inventory into American warehouses ahead of the deadline, tightening availability everywhere else and adding a geopolitical premium to a market already short of metal.
The WBW view
Treat this as a structural repricing, not a spike. For manufacturers, the practical agenda is procurement: longer-dated supply contracts, hedging programmes sized to multi-year exposure, and design-stage substitution where aluminium can do the job. For investors, the bottleneck economics favour incumbent producers with existing, permitted assets over the promise of new discoveries that remain 17 years from cash flow. And for anyone modelling AI infrastructure costs, the lesson of 2026 is that the constraint list has grown: first chips, then power, now the metal that connects the two.