Most discussions of artificial intelligence (AI) and copper focus on how many more tonnes of copper data centers will need. It is the right place to start, but it misses what may be the more consequential change.
AI is bringing into the copper market a new class of buyer with an unusually high willingness to pay. As these buyers account for a growing share of incremental demand in a supply-constrained market, they could increasingly influence copper prices, reshaping who consumes the metal, how it is used and how the market evolves.
A material-intensive start
Despite its digital nature, AI rests on a substantial physical foundation that includes data centers, transformers, substations, cooling systems, new electricity generation and transmission lines. The International Energy Agency projects that global data center electricity consumption will rise from about 415 TWh in 2024 to around 945 TWh by 2030, more than doubling in six years.
Industry estimates suggest that data centers may require roughly 30 tonnes of copper per megawatt of installed capacity, depending on design and the infrastructure included, before accounting for the wider electricity networks needed to power them.
AI is not arriving alone. It coincides with electrification, the energy transition and a geopolitical push to rebuild domestic manufacturing, diversify supply chains and expand military capacity. These trends have different motivations, but all require large investments in physical assets before their economic benefits materialize. Together, they point to an unusually material-intensive phase of global investment.
A supply side that cannot keep up
This demand meets an industry that struggles to grow. Declining grades, deeper deposits, more complex ores, water constraints and lengthy permitting have made new supply slower and more expensive. Bringing a new mine from discovery to production can now take well over 15 years.
Even established mining regions illustrate these constraints. Chile, the world’s largest copper producer, has struggled to expand output despite its vast resource base and substantial investment pipeline. Across the industry, sustaining production at mature operations increasingly requires major capital expenditures, while developing new deposits involves growing technical, environmental and regulatory challenges.
This matters for prices. When supply responds slowly, even relatively modest increases in demand can have significant market effects. As the cost of developing new production rises and high-value technological applications expand, the combination could support structurally higher copper prices, without eliminating the cycles inherent to commodity markets.
Not all buyers feel the price the same way
Here lies the less-discussed shift. In an AI data center, where total investment can exceed $30 million per megawatt, copper may represent only around 1% of the overall cost, depending on the facility’s configuration and the infrastructure included. For hyperscalers racing to deploy computing capacity, securing infrastructure on time matters far more than fluctuations in the price of the metal inside it.
Traditional copper consumers face a different reality. Manufacturers of cables, construction products, appliances and electrical equipment are often more exposed to raw material costs, although their sensitivity depends on their ability to pass those costs on to customers and on the availability of substitutes.
Because both groups participate in the same global market, their different willingness to pay matters. As high-value technological applications account for a growing share of incremental demand, their willingness to pay could increasingly shape the marginal price of copper, putting pressure on traditional users with less capacity to absorb higher costs.
The result need not be lower aggregate demand, but a shift in its composition, with faster material efficiency improvements, product redesign, recycling and substitution, particularly by aluminum where technically and economically feasible.
For forecasters, aggregate consumption projections are no longer enough. Understanding who can pay, and who will seek alternatives, is becoming as important as estimating how much copper the world will consume.
The Jevons question
A second, longer-term uncertainty concerns the relationship between AI-driven productivity and material consumption. AI promises efficiency gains across the economy, from engineering design to logistics and resource allocation, potentially reducing the materials required per unit of output.
History, however, counsels caution. In 1865, William Stanley Jevons observed that more efficient steam engines had not reduced Britain’s coal consumption but helped expand it, as cheaper energy encouraged new uses. In copper’s case, improvements in material efficiency have not prevented global consumption from increasing as electrification, industrialization and new applications have spread.
Greater efficiency does not necessarily mean lower aggregate consumption. By reducing costs and creating new economic opportunities, AI could stimulate additional investment, electricity demand and infrastructure development.
The question is how these forces will interact over time. This uncertainty argues against both extrapolating today’s demand growth indefinitely and assuming that technological efficiency will necessarily reduce future copper consumption.
AI as a supply-side tool
Most attention has gone to the copper AI will consume. Less has gone to the copper it could help produce. Machine learning can improve exploration targeting; advanced analytics can optimize mine plans and equipment utilization; AI-based process control can raise grinding efficiency and flotation performance.
The gains can be material. At a mine producing 300,000 tonnes of copper annually with an 85% metallurgical recovery rate, improving recovery by just one percentage point would generate approximately 3,500 additional tonnes per year, potentially with limited incremental capital investment.
Capturing these benefits requires more than buying software. It demands reliable data, integrated systems, skilled people and organizations able to act on what the models indicate. The real test of AI in mining will not be the number of pilots announced, but the additional copper produced and the costs reduced.
A twofold challenge
AI is likely to strengthen copper demand, but its most lasting effect may be on the structure of the market. Some consumers may be able to absorb substantially higher prices, while others face growing incentives to substitute copper. At the same time, technological innovation is changing the economics of producing the metal itself. In such a market, high prices can also accelerate the efficiency gains and substitution that erode copper’s traditional markets.
Companies that use this window to raise productivity and build technological capabilities will be better placed to weather future cycles. The same applies to producing countries, whose opportunity extends beyond exporting more metal to developing technologies, engineering capabilities and mining services that can compete globally.
The challenge is understanding a demand whose composition is changing and using AI itself to build a more productive mining industry.
The future of copper will depend not only on how much the world needs, but on who can afford it and how effectively the industry can produce it.
Juan Carlos Guajardo is the executive director of Chile-based consultancy Plusmining.