Amir Khamidullin

As AI is set to be the main driver of the world’s growth, with the U.S. and China at the forefront of the new innovation era, economies are allocating resources to data centres, servers, and cooling systems to run increasingly power-demanding artificial intelligence systems. All of these AI-related facilities depend on copper wiring, natural-gas turbines, rare-earth magnets, concrete, and a wide range of other inputs.

Copper has quietly become the backbone of the new industrial cycle. Every major technological shift in history has relied on a foundational material — steam on coal, electrification on steel, and now artificial intelligence on copper.

Production and Refinery

U.S. Global Investors fund recently said that “a single AI data centre that uses more copper than three conventional facilities combined” with estimations of AI data centres consuming “half a million tons of copper annually by 2030”.

The U.S., which is the number one investor and developer of AI at the moment, along with China, is lagging well behind in both production and refinement of copper globally.

As noted in Figure 1, while the U.S. is outperforming China in copper reserves, China leads the U.S. by more than 13 times in copper refining in 2024. Chile stands strong with a relatively enormous level of reserves and somewhat adequate refinery capabilities.

Refinement capacity and related infrastructure are crucial for AI and industries in general because raw copper is useless to the economy until it is purified, shaped, and delivered in forms that can meet the extreme performance standards of modern computing and power systems.

Moreover, according to the U.S. Geological Survey Scientific Investigations Report, the United States declined both in production and refinement of copper from 2022 to 2024, and this trend continued over to 2025 (Figure 2).

Across all years, total apparent consumption (1.66–1.96 million tons) is far higher than mine production (1.10–1.23 million tons), and primary refined output (0.84–0.93 million tons). This confirms the U.S. is a structural copper importer and cannot meet domestic demand through internal mining or refining. The gap is roughly 600–800 thousand tons every year, which means that about 40-45% of U.S. copper needs must be met by foreign supply. Indeed, according to the same U.S. Geological Survey Scientific Investigations Report, net import reliance as a percentage of apparent consumption is 41% in 2023 and 45% in 2024.

U.S. Imports and Tariffs

The U.S. copper import structure is extremely concentrated and highly segmented by product type. Refined copper dominates total copper imports, and Chile alone supplies nearly two-thirds, making the U.S. heavily dependent on one country for the final, ready-to-use copper that drives manufacturing (Figure 3). Indeed, Chile stands strong with 190,000 metric tons of reserves (Figure 1).

The U.S. copper supply chain relies on a few key trade partners: Chile for refined copper, Canada for upstream materials, and Mexico/Canada for scrap (Figure 3). This creates strategic vulnerability: any disruption in these handful of countries (strikes, policy shifts, logistics issues) could quickly impact U.S. copper availability and pricing.

President Trump announced in July that “an investigation under section 232 of the Trade Expansion Act had determined that ‘copper is being imported into the United States in such quantities and under such circumstances as to threaten to impair the national security of the United States». This is indeed supported by the data analyzed above.

The most straightforward yet most complicated solution to this structural issue is to boost domestic production and the refinement of copper. There are a few major copper producers in Arizona and Utah, but both are limited in their capacity to refine copper. Even if investments are made and effectively employed, the time lag is extreme. In other words, there is no quick way for the U.S. to break its dependence on copper imports.

Later, President Trump imposed 50% tariffs on imports of semi-finished copper products (copper wires, pipes, rods and sheets) and copper-intensive derivative products (cables, pipe fittings and electrical components). However, refined copper was not immediately included. Rather, there is a recommendation made for a phased universal tariff of 15% on refined copper, starting in 2027, and 30% starting in 2028. This gives a clear signal for the market that the U.S. wants to build domestic refining capacity, while providing importers time to adjust.

However, markets indeed fear imposition of further copper tariffs and question the ability of the U.S. to meet its local demand for copper through local production and refinement, given the time lag and higher import costs. This is evident in current U.S. trends in production, refinement, and consumption (Figure 2) as well as in futures markets.

Futures and forwards

Futures and forward contracts on copper are primarily traded on the Chicago Mercantile Exchange (CME) and the London Metal Exchange (LME)*.

There is a clear divergence in trends of CME Copper Futures prices and LME Copper Forward prices for the same expiration months. CME Futures experience a contango trend, meaning that futures prices are higher than the current spot price and each successive contract increases in price over time. LME Forward contracts are, in turn, in backwardation — futures prices are below the current spot price, and each successive contract decreases in price over time.

This provides a very neat yet interesting implication. Backwardation in LME Forward contracts implies current immediate tightness in supply and higher than usual demand (as firms are rushing ti move copper to the U.S. before possible tariffs) with expectation of easing demand and possible increased mine supply in London and globally. Indeed, according to Bloomberg, Mercuria Energy Group Ltd. recently ordered about “$500 million of copper for withdrawal from warehouses overseen by the London Metal Exchange, as the trading house positions itself for a global supply crunch fueled by possible US tariffs”.

Contango in CME futures contracts mirrors the same dynamic. Demand for copper in the U.S. is expected to grow further for years, as A.I. infrastructure will require even more of it, and domestic supply doesn’t present any clear signs of improvement now. Carrying costs (storage + financing) are also positive for the U.S. Anticipation of tariffs on refined copper makes the situation even more complicated, pushing prices up.

All in All, copper has become the central metal of the AI-driven industrial cycle, yet the United States enters this new era with a structural disadvantage. The U.S. now relies on foreign supply for roughly 40–45% of its copper needs, with refined imports dominated by Chile and upstream materials concentrated in Canada and Mexico. This dependency, paired with long lead times for developing new mines and refineries, makes the U.S. very vulnerable to supply disruptions and policy shocks. The recently announced tariffs on semi-finished copper products,  ombined with proposed future tariffs on refined copper, signal Washington’s intention to rebuild domestic capacity, but they also heighten near-term uncertainty over whether U.S. supply will tighten even further as AI infrastructure buildout accelerates.

The futures market reflects these emerging stresses. LME copper forwards sit in backwardation, signaling immediate global tightness, rapid warehouse drawdowns, and elevated short-term demand. In contrast, CME futures curve is in contango, shaped by expectations of sustained U.S. demand growth, positive carrying costs in a high-rate environment, and fears that future tariffs may push domestic prices higher. Their divergence underscores a widening split between global supply conditions and the uniquely bullish, structurally constrained outlook for copper in the United States.

---

*It is important to note that CME and LME curves are not perfectly comparable, as they reflect different delivery locations, inventory structures, and financing costs. Nevertheless, the divergence between them still highlights a meaningful split between U.S.-specific demand pressures and broader global supply conditions.