Does Copper Set the Speed Limit on Growth?

Copper is not a fashionable metal. It doesn’t sparkle like lithium, doesn’t frighten like uranium, and doesn’t come with the ethical theatre of cobalt.

Copper entered my thinking this week not through policy papers or price charts, but through a numerical claim. We now consume roughly 30 million tonnes of copper a year, recycle only a small fraction of it, and that sustaining something like 3% global GDP growth, even without further electrification, would require mining an amount of copper over the next two decades comparable to everything extracted over human history.

The framing is deliberately provocative… But it gestures toward a quieter, more unsettling question: Are our growth assumptions resting on a physical system that simply cannot respond on the timelines implied by our ambitions?

Start with demand. Global copper consumption already exceeds 25 million tonnes a year. According to the International Energy Agency, it is likely to rise above 35 million tonnes by the mid-2030s under current policies, and approach 40 million tonnes in a net-zero scenario. This isn’t a single technology gone wrong. It’s everything at once from power grids, to electric vehicles, charging infrastructure, renewables and buildings.

Copper isn’t just another “energy-transition metal”. It is THE platform metal. Roughly 60% of all copper ends up as wire and cable. Electricity doesn’t flow through ambition, regulation, or PowerPoint decks. It flows through copper.

Supply responds on geological and industrial timescales. New copper mines typically take 15 to 20 years from discovery to production. Ore grades have been falling for decades, particularly in Chile, which is still responsible for around a quarter of global production. This means more rock, more energy, and more capital per tonne of copper produced. As grades decline, energy intensity rises, quietly turning copper into an energy-price problem as well as a mining one.

Recycling helps, but it lags by design. Around 80% of all copper ever mined is still in use, locked inside buildings, grids, and machines doing exactly what they were designed to do. Only about a third of current demand is met by recycled copper. Circularity solves stock problems over decades; it does little for flow problems during rapid expansion.

Put these pieces together and the IEA’s warning becomes easier to understand. Without substantial new investment, global copper supply could fall around 30% short of demand by the mid-2030s. Not because copper is rare, but because scaling supply at the pace implied by our ambitions is slow, capital-intensive, and politically messy. That part tends to get waved away.

Then there is the part of the story rarely framed as a materials problem → AI.

Data centres are usually discussed as an electricity issue, and they are. Global data-centre electricity demand already sits around 400–500 terawatt-hours a year and is rising fast, driven largely by AI workloads. But electricity demand is also a copper story in disguise. More power means more grid capacity, more substations, more transformers, more cabling. A single megawatt of data-centre capacity can embed tens of tonnes of copper once power distribution and cooling infrastructure are included.

AI may live in the cloud. It still sits on copper.

What’s striking is how neatly these stories stack. Electrification, decarbonisation, and AI are often discussed as separate transitions. Physically, they converge on the same infrastructure, grids, transformers, cables, and the same material base. Copper is where these futures collide.

Layer geopolitics on top and the picture tightens further. Mining and processing are not the same thing. While copper is mined across multiple continents, refining and midstream processing are highly concentrated, with China playing a dominant role across copper and many other critical minerals. Western governments are only now rediscovering this reality under the banner of “strategic autonomy”. Rebuilding diversified supply chains is possible, but it moves at the same speed as mines, permits, and grids.

What copper exposes is a deeper misalignment. Our economic imagination now runs on software timelines, while the physical systems it depends on move at geological speed. Growth models, climate targets, and AI roadmaps all assume near-term acceleration.

This doesn’t mean the transition fails. It means it is slower, costlier, and more politically charged than our narratives tend to admit. The greener and more electrified the economy becomes, the more metal it needs, and metal does not scale like software.

Time, inconveniently, is the one variable we keep assuming we can bend.

Painting: Salvador Dali’s “The Persistence of Memory” (1931)

Sources

• European Commission – Data centres: an energy-hungry challenge

• International Copper Association: Energy Transition Will Support Global Copper Demand

• International Energy Agency (IEA): The Role of Critical Minerals in Clean Energy Transitions

• International Energy Agency (IEA): Global Critical Minerals Outlook 2024

• International Energy Agency (IEA) – Electricity 2024: Analysis and Forecast to 2026

• Northey et al. (2014) – Decreasing ore grades in global metallic mining (Journal of Cleaner Production)

• Princeton University – Energy Use in the Copper Industry

• US Geological Survey (USGS) – Copper Statistics and Information

• Reuters – Coverage on China’s role in critical minerals and refining

• World Bank – Mineral Intensity of the Clean Energy Transition