The Oldest Metal in the World Just Became AI's Newest Problem

Copper is on an absolute tear.

Last week, copper futures climbed to a record high of nearly $6.90 per pound. The metal is now up nearly 50% in the last 12 months, making it one of the best-performing major commodities in that time frame.

I get what you’re thinking. Futuristic innovations expert talking about… copper.

Is it exciting? Not really.

Humans have been using the metal for thousands of years and there’s nothing revolutionary about a piece of copper wire.

But that’s exactly what makes this story so interesting.

Some of the most advanced technologies in the world increasingly depend on one of the oldest metals humans have ever used. And as trillions of dollars pour into AI and other next-generation technologies, copper could become one of the most important (and profitable!) bottlenecks of the next decade.

Earlier this year, Amazon (AMZN) signed an interesting agreement tied to its massive AI buildout.

It wasn’t with a semiconductor company or one of the up-and-coming neoclouds. It was with mining giant Rio Tinto (RIO).

Under the two-year agreement, Amazon Web Services will use copper produced at the Johnson Camp mine in Arizona for components in its U.S. data centers. The copper will be extracted using Rio Tinto’s new Nuton bioleaching technology.

Think about that for a moment.

We’ve spent the last several years talking about GPUs, large language models, and companies like Nvidia (NVDA). Those are obviously critical pieces of the AI boom. But all those chips need to be housed somewhere. They need electricity. They need cooling. And they need to be connected to an enormous physical infrastructure.

That means electrical cables, transformers, motors, circuit boards, cooling systems, substations, and much more. Copper runs through virtually all of it.

AI may live in the cloud. But the cloud needs an awful lot of copper.

And AI is only one piece of the demand story.

Electric vehicles can use roughly four times as much copper as traditional vehicles. Renewable energy requires it. The electrical grid needs it. Defense systems need it. Factories need it. And as we discussed last week, the coming robotics boom will require even more electrification.

Now we’re throwing the largest data-center construction boom in history on top of all that existing demand.

S&P Global expects global copper demand to increase roughly 50%, from about 28 million metric tons today to 42 million by 2040. Data-center demand alone is projected to increase from 1.1 million metric tons in 2025 to 2.5 million by 2040.

That’s a lot of copper.

And there’s a problem.

Silicon Valley is accustomed to moving fast. Mining moves at a very different speed.

If Amazon decides it wants another data center, it has the money to start building. If a software company suddenly needs more computing power, it can order more servers.

But if the world suddenly needs another million tons of copper every year, we can’t simply flip a switch and produce it.

Major mines can take well over a decade to discover, permit, finance, and construct. At the same time, ore grades have been declining in many important mining regions. Miners must move more rock and spend more money just to produce the same amount of copper.

That’s why I’m actually more intrigued by the supply side of this story than I am by demand.

S&P Global projects global copper production will peak around 33 million metric tons in 2030. Without major changes, it estimates the gap between supply and demand could reach 10 million metric tons annually by 2040.

That’s an enormous potential shortfall.

AI infrastructure can be planned in months. New copper supply can take more than a decade.

You don’t need an economics degree to figure out what happens when demand keeps rising and supply can’t keep up.

Now, I’m not telling you copper is going straight up from here. Nothing does. Copper has always been cyclical, and with prices sitting near record highs, there will almost certainly be pullbacks along the way.

But when I look beyond the next few months, I really like the setup.

Investors looking for exposure can start with the industry’s giants, including Freeport-McMoRan (FCX), Southern Copper (SCCO), Rio Tinto (RIO), and BHP (BHP).

Personally, I’m more interested in digging deeper.

I want to find the smaller companies sitting on large copper deposits that could become far more valuable if the shortages being projected today actually materialize.

Ideally, we’re looking for projects in mining-friendly jurisdictions with good infrastructure and a realistic path toward production.

That’s where things could get really interesting.

And the bigger lesson here goes well beyond copper.

For years, investors have viewed AI primarily as a digital revolution. But the deeper I dig into the next phase of this boom, the more physical it becomes.

We’re going to need power plants, transmission lines, data centers, factories, transformers, robots, cooling systems, and enormous amounts of raw materials to build all of it.

That’s why some of the biggest winners from AI over the next decade may look nothing like traditional AI companies.

The most advanced technological revolution in human history still needs things dug out of the ground.

And right now, copper is looking increasingly difficult to ignore.

Here’s to the future, 

Matt McCall
Editor, Market Insights