The numbers, and every constant they rest on
Mine shares are from World Mining Data 2024 (South Africa’s platinum share, Russia’s palladium share). The reserve concentration (~90% in South Africa), the mine split by metal (Pt/Pd/Rh), the autocatalyst demand share (~68%) and the recycled share (~27%) are from USGS Mineral Commodity Summaries and the Johnson Matthey PGM Market Report; each carries the low–high bracket. The rhodium price peak (2021) is Johnson Matthey.
Input: production.json → pgm.json (build_pgm.py). PGMs are co-mined, so “output” here is by contained metal mass; individual-metal shares vary by deposit.
The co-production lock
Platinum, palladium and rhodium are not mined separately — they come out of the same ore in this rough mass ratio. You cannot buy “more rhodium”; you can only mine more of the whole basket.
Why you can’t out-build it
A refining chokepoint — gallium’s furnace, cobalt’s Chinese refineries — is a capability: hard to reproduce, but reproducible, with capital and years. A geological chokepoint is not. The platinum-group metals are concentrated because the ore that carries them in economic grade sits, overwhelmingly, in one complex in South Africa and one district in Russia. There is no second Bushveld to build; you cannot fund a deposit into existence. This is why the atlas keeps only five of its 58 chains in the geological class — the mechanism where the entire toolkit of “stand up an alternative supplier” simply does not apply. What is left is narrower and slower: substitution within the basket (which moves nothing — Pt, Pd and Rh trade for each other but all come from the same two countries, see the substitution networks), recycling spent autocatalysts (~a quarter of supply, the one lever that adds metal without new geology), and stockpiling against a shock. The 2021 rhodium spike — above $29,000 an ounce, richer than gold by an order of magnitude — is what the lock looks like when demand for one co-product outruns the geology of the others. Producers are not wholly rigid — they can tilt the mine plan a little, draw down inventory and lift recycling — but only slowly, and only within the ratio the ore sets: a co-product constraint with limited flexibility, not a switch.
Why one basket, done deep, matters
The atlas’s thesis is that the chokepoint is rarely the mine — and it means it, which is why the exceptions carry weight. The platinum-group metals are the clean counter-case: here the chokepoint is the mine, and geology says why. Set beside gallium (a by-product you can’t scale) and cobalt (mined in one country, refined in another), the PGMs complete the picture — the three mechanisms that put a material out of reach are can’t scale it, can’t see past the refiner, and can’t out-build the rock. Knowing which one you face is the whole of the decision. The mechanism map →