The chokepoint is refining, not mining

The gap between what China mines and what China refines

World refined copper output is 27,000 kt a year. China accounts for 12,000 kt of that, or 44.4%.

Its share of copper mining is a small fraction of the same number. China buys ore from everywhere and turns almost half the world's copper into metal.

So the figure people reach for when they discuss copper concentration, the mining share, is measuring the wrong step.

The mining figure is also milder than usually reported, which widens the mismatch. On USGS 2024 estimates, Chile at 23.0%, the DRC at 14.3% and Peru at 11.3%, together with China and the United States, account for 14,100 kt of a 23,000 kt world total. That is 61.3%, not the roughly 70% commonly cited. The step everyone measures is less concentrated than the headline suggests. The step almost nobody measures is more.

Why the pattern repeats

This is not a copper story. It happens wherever the processing step is capital-intensive and the ore is not.

China holds a reported 90%-plus of battery-grade spherical graphite processing, and dominant rare-earth separation capacity. Both sit far above its corresponding mining share. Graphite is mined in several countries and turned into anode material in essentially one.

The reason is structural rather than strategic, at least to begin with. Refining is expensive, slow to build, hard to permit and commercially unglamorous. It attracts none of the political attention a mine does and earns thinner margins per tonne. Concentration accumulates at the stage with the highest capital intensity and the lowest political salience, because that is the stage nobody writes supply-security policy about.

Indonesia ran the same mechanism deliberately, and it is the clearest case available.

The government banned raw ore exports, comprehensively from 1 January 2020, so anyone who wanted Indonesian nickel had to process it inside Indonesia. Refined nickel output went from 24,000 tonnes in 2014 to 636,000 tonnes by 2020. Operational smelters went from 2 to 43, on roughly $30bn of largely Chinese investment.

That took six years. A country with ore and no processing industry became the processing industry, by making the ore unavailable in any other form. The mechanism is available to any resource-holding state willing to wear the transition cost.

What would prove this wrong

Two things would falsify the argument. Neither has happened.

The first is cobalt easing, as Western policy intended. A decade of diversification effort has been aimed at reducing DRC concentration.

It went the other way. On USGS figures the DRC reached 76% of world mine production in 2024, up from around 70% in 2019. CMOC alone produced 114,000 tonnes, roughly 33% of the world total from a single company.

The price behaviour should give diversification advocates pause. Over the same window cobalt fell 59.5%, from $41 to $16.62 a pound between May 2022 and May 2025. A dominant low-cost producer expanding through a price trough is not a market correcting itself. It is a market consolidating, because high-cost entrants outside the DRC cannot survive the trough the incumbent is causing. A glut of that kind increases fragility rather than reducing it, which is the opposite of what a buyer watching prices fall would assume.

The second is the Indonesian precedent being struck down. A 2022 WTO panel ruled the export ban inconsistent with trade rules. Indonesia appealed, and the dispute is unresolved.

So the concentration Indonesia built rests on a contested legal basis, and we would not want to overstate how durable the model is. But a ruling against it would be a risk to the mechanism, not evidence the mechanism fails. It plainly worked. The open question is whether others can copy it.

One caution applies to every figure here quoted to a decimal place. Gold production is disputed by about 10.94% between the World Gold Council's 3,661 tonnes and the USGS's 3,300 tonnes for 2024. Different methodologies, neither authoritative over the other. Where two sources measure the same quantity and no arbiter exists, both readings should travel together. Model both, average neither.

Why tonnage will mis-rank this sector

Weight the sector by volume and you will load heavily on iron ore, because iron ore is where the mass is. Per BGS production data, iron ore's 3.3 billion tonnes generated $130.65bn of Top-100 revenue, almost identical to gold's $131.81bn from a tiny fraction of the mass.

Volume and value have almost no relationship in mining, and concentration risk tracks neither reliably. It tracks the processing step. A model weighted by tonnage produces a confident ranking that loads on iron ore and misses the copper, graphite and rare-earth chokepoints completely.

What follows from this

A mine-diversification programme that does not also secure refining capacity has bought optionality on feedstock, not on supply. It hedges one risk, a specific country expropriating or a specific deposit failing. It does nothing about the risk that actually binds, which is a processing step in a jurisdiction that can decide who gets served.

So the question to ask of a supply chain is not where the material is mined. It is where the material is refined, how many places can do that, and what happens if the largest of them stops.