Critical Materials Atlas
Deep dive · one metal, followed to the bottom

Cobalt, end to end

Cobalt is barely mined for its own sake — it comes up with the copper of the DR Congo and the nickel of Indonesia. So its supply is set by someone else’s metal, its mine sits in one country and its refining in another, and — uniquely on this atlas — it is the material where our own reconciliation engine got it wrong. This page follows the one metal all the way down, and corrects the record where our tool fell short.

The mass balance, and every constant it rests on

World cobalt production (World Mining Data 2024, ~269,000 t) with the DR Congo’s mine share taken from the same source. The by-product split (~70% recovered with copper, ~28% with nickel, ~2% primary) is from USGS and the Cobalt Institute; the China refining share (~76%) from the IEA Critical Minerals Dataset and USGS MCS; the battery end-use share (~72%) from the Cobalt Institute / IEA. Each carries the low–high bracket shown on the bars.

The engine-miss is measured, not asserted: the 2020 comparison below is computed from the atlas’s own flows_2020.json against CEPII BACI (HS 2822.00), reproducible with extract_baci.py. Inputs: production.json → cobalt.json.

The mass balance — one metal, two countries

From the world’s mined cobalt down to the battery, and note the hand-off: the country that mines it is not the country that refines it. Bars to scale; the shaded bracket is the low–high bound.

Why price can’t fix it — and why a ban can

Cobalt has two chokepoints, and they are different in kind. Upstream, it is a by-product: you do not open a cobalt mine, you open a copper mine and the cobalt comes with it. So when cobalt’s price spikes, supply cannot answer on cobalt’s terms — it answers on copper’s, and copper expands for copper’s own reasons. Downstream, refining is a built capability concentrated in China (~76%): that one can be rebuilt with capital and years, but until it is, an export or licensing decision there reaches every cathode maker on earth. The mine is a governance-and-geology problem (the DRC’s copperbelt cannot be relocated); the refinery is a capacity problem. The one genuine demand-side escape is chemistry: LFP batteries use no cobalt at all, and their rise is the first thing in decades to actually loosen the squeeze — a lever the supply side never had.

The day our own tool got it wrong — and how we know

A research tool is only honest if it shows where it fails. For cobalt, the atlas’s reconciliation engine — which reconstructs trade from both partners’ reports — badly understated the DR Congo’s real export dominance during the 2020–2023 battery boom. It reconstructs a fraction of the trade, with the DRC a minor player, where BACI (and reality) show the DRC dominant:

2020, HS 2822.00 cobalt oxides/hydroxidesTotal tradeDR Congo share

Why one metal, done deep, matters

A scorecard across 32 materials shows breadth; a single chain followed to the bottom shows the mechanism — and, here, its edges. Cobalt makes three of the atlas’s abstractions concrete: by-product becomes “~98% comes up with copper or nickel, so price cannot summon it”; the chokepoint is rarely the mine becomes “mined in the DRC, refined in China — two countries, two different fixes”; and the engine understates concentration stops being a caveat and becomes a measured, corrected fact you can see. The materials the atlas is most sure about are the ones where it has also found its own limits. What you can trust →