Copper is the cornerstone of AI. These 3 ETFs mine it
Artificial intelligence doesn't run on chips alone. Every new data center needs thousands of tons of copper in substations, cabling, transformers and cooling. That's why this industrial metal hit all-time highs this year. But supply is lagging behind demand, and tariffs, strikes and mine outages are also being priced in. Investors have three completely different ways to bet on the copper boom. Which one makes sense and where are the risks?
Key points
The AI boom isn't just about chips. Behind every new data center stands a huge amount of physical infrastructure, and copper plays a crucial role in it.
Three ETFs, but three distinctly different bets on the same megatrend. The difference isn't just in fees, but mainly in where the potential return should come from.
Miners can offer significantly more upside when commodity prices rise. But the same leverage works in the opposite direction, as this year's developments showed very quickly.
The record copper price has several drivers. AI is only one of them, and understanding the others can be crucial for what comes next.
The long-term trend looks strong, but the short-term is not so clear-cut. The difference between these two views can decide what exposure to choose.
When talk turns to AI boom winners, most investors think of chipmakers, hyperscalers or server suppliers. But at the end of the entire chain stand the raw materials without which no data center would ever be built. Copper is the most important of them. It has the best electrical conductivity, which is why it's found everywhere from the grid connection through transformers and busbars to the cables inside server racks.
The year 2026 showed how strongly this physical reality began to be reflected in prices. Copper traded at all-time highs in September and added roughly 46% over the past year. Meanwhile, copper miners, due to their operating leverage, earned even more on the stock market than the metal itself. At the same time, part of this year's rally isn't based on fundamentals but on expectations of US tariffs and inventory shifts between continents. That's why it's essential for investors to understand what type of exposure they're buying and what exactly their investment corresponds to. These ETFs make that easier for investors and have also earned them nice returns in recent years.
Why copper is the metal of the AI era
Traditional data centers have always been material-intensive, but they played only a marginal role in global copper consumption. AI is changing that. Compute clusters for model training have many times higher power density per rack, and thus significantly higher demands on power distribution, backup power and cooling.
By some estimates, a typical AI or cloud data center consumes 27 to 33 tons of copper per megawatt of installed capacity, and for hyperscale model-training campuses it's up to 47 tons. A single 1-gigawatt campus can thus absorb up to 50,000 tons of copper, even before counting the transmission grid reinforcements needed to supply it.
