Mine output shrinking, costs leap...
COPPER set another record this month, writes Frank Holmes at US Global Investors.
Three-month metal on the London Metal Exchange (LME) touched $14,779 a ton while New York futures crossed $3.74 a pound. The red metal is up roughly 24% this year and about 51% over the past 12 months, beating Magnificent 7 stocks.
Ask around and you'll probably hear the same explanation for why this is happening: US tariffs.
The Commerce Department has proposed a 15% duty on refined copper imports beginning in 2027, rising to 30% in 2028, though the administration has neither confirmed nor ruled it out. The story goes that traders are racing metal into US warehouses ahead of the deadline. July imports hit a record 225,094 tons.
It makes sense on paper, but I no longer think it's the right explanation.
Take a look at the chart below. When traders genuinely expect a tariff, New York has to trade at a premium to London because that premium is the only thing that pays the duty.
You can see that's exactly what happened last summer. In July 2025, COMEX copper averaged nearly a 23% premium over the LME price, touching 30% at the peak.
Today the premium has averaged about 1% all year. In February and March, it actually turned negative, with New York trading at a discount to London. As I write, it's roughly $148 a ton.
So copper has gained nearly half its value in a year while the arbitrage that supposedly explains the move sat flat on the floor.
If this were a tariff trade, the spread would be blown out the way it was 14 months ago. But it's not, which tells me something else is doing the work. That something is deteriorating supply.
The International Copper Study Group (ICSG) reports that global mine production fell 1.1% in the first half of 2026.
Output dropped in Indonesia and the Democratic Republic of Congo, two of the largest producers on earth, and Chile − the world's largest copper producer, responsible for a little under a quarter of total global supply − just posted its weakest second quarter in at least 19 years and cut its full-year forecast for a second consecutive quarter, now guiding to a 2.6% annual decline.
Codelco and Freeport-McMoRan both reported double-digit production declines. Morgan Stanley now projects global production running flat or slightly lower for the year after beginning 2026 expecting mine supply to expand. That would mark the first annual decline in global copper mine supply since 2017.
A weak year for output is a headline, but underneath it sits what I see as a generational problem.
S&P Global Market Intelligence tracks major copper discoveries back to 1990. In 36 years, the industry has found 263 of them, holding about 1.4 billion tons of copper.
That sounds like plenty...until you sort it by decade. The industry found 715 million tons in the 1990s alone. Everything discovered since the year 2000 adds up to 687 million tons.
Let's put a price on it. In the 1990s, the industry spent roughly $6 billion looking for copper, which worked out to about $8 for every ton discovered. But so far this decade, since 2020, it's spent $16.4 billion and found only 8.7 million tons. That's roughly $1,889 a ton, a 225-fold increase in the cost of finding the metal.
The easy copper deposits were found generations ago. Average drilling depth has risen nearly 50% since 2010, to around 600 meters, and ore grades keep slipping.
Permitting has grown slower. S&P puts the average timeline from discovery to production at a whopping 17.5 years, meaning a deposit found today doesn't deliver copper until the 2040s.
While supply is slowing, demand keeps surging. S&P projects copper demand will climb from 28 million tons in 2025 to 42 million by 2040, a 50% increase. The firm warns of a potential 10 million ton shortfall without meaningful supply expansion.
Of course, artificial intelligence (AI) is the latest copper driver. Recent research published in the peer reviewed journal Resources Policy found that copper accounts for an incredible 82% of the total mineral mass in AI data center construction, and that grid transmission and distribution, not compute hardware, accounts for the largest share of it.
Meanwhile, Bank of America estimates that each incremental megawatt of data center capacity embeds 60 to 75 tons of metal, mostly copper, noting that metals run under 5% of total data center capex. That makes the demand highly price inelastic.
You're also seeing capital behave strangely, which in my experience has often been the tell.
Ivanhoe Mines, founded by my friend Robert Friedland, raised the resource estimate at its Western Forelands project in the Congo by 30% this week, to 12 million tons of contained copper. Robert says he's received calls from sovereign wealth funds and Silicon Valley hyperscalers, and he describes it as interest he's never seen across 45 years in mining.
It's important to remember the copper deficit everyone's talking about is just a forecast. The refined market has been running close to balanced, flipping between small surpluses and deficits month to month. Bloomberg Intelligence's own model shows surpluses in 2025 and 2026 and doesn't turn negative until later this decade.
Copper mining stocks have moved too. Freeport-McMoRan is up 44% year-to-date, while Southern Copper and Teck Resources both up around 45%. Freeport's own sensitivity model puts each 10-cent move in copper at roughly $390 million in annual EBITDA.
Copper's scarcity is set by geology, drilling depth, permitting queues and a 17.5-year lead time. None of those respond to a rally.
Just don't tell yourself this one is about tariffs. That trade appears to have ended last summer.
