Gallium for chargers, cobalt for batteries, copper for every wire. Most of each passes through a handful of refineries. China produced 99% of the world’s primary gallium in 2025 (USGS).
Plate I. — The critical elements and the places that refine them. Labels show the top country’s share (USGS 2026; rare earths IEA, polysilicon Bernreuter); routes are illustrative.
Pick an object. Each mineral inside it has a chain: dug from the ground, refined, purified, made into a part. Control is often not at the mine but at a refinery.
Chips get the headlines, but every watt they use arrives through copper, transformer steel and aluminum. Follow the power from the grid down to the chip, and see where each metal goes and which step is the hardest to get.
This is a real supply-chain optimization model for gallium, solved live in your browser every time you move something. It knows every mine, refinery and wafer plant it can use, what each one costs, how much carbon it emits and how much child-labor risk sits behind it. It always picks the cheapest way to meet US demand, unless you tell it to care about something else.
China's plants are enormous. They run at a tiny fraction of capacity for US orders, so only deep cuts bite. Try below 5%.
Drag the dot. At "cost" the model is a pure bargain-hunter. Pull toward carbon or child labor and it pays more to avoid them.
Read this as "the cheapest world that's possible", not "the world today". Today China recovers about 99% of primary gallium (USGS 2026). The model's least-cost answer recovers it in Australia because it only counts the costs it's given: no export licences, no politics, no years-long build times. That gap between possible and actual is the point. China's hold comes from built capacity and policy, not geology.
Wind turbines carry rare-earth magnets: about 600 kg per megawatt offshore and about 120 kg onshore (placeholder values from JRC/IEA ranges). A US grid plan that reaches 100% clean power by 2035 builds only a little more wind by 2050 (about 4%), but reaches the normal plan's 2042 level by 2035, about seven years earlier.
Squeezing the same demand into fewer years is what hurts. When that surge is pushed through a supply-chain model, more of it goes unmet, more of it is imported, and more of it routes through China.
US installed wind capacity (GW). Normal plan in blue, 100%-clean-by-2035 in red. NREL 2024 Standard Scenarios.
Magnet demand, cost, import and unmet figures come from a grid-to-minerals bridge between NLR's grid-planning model (ReEDS) and its supply-chain model (BLEECAM), published as a method demonstration. Magnet-per-MW values are placeholders, so compare the two plans, not the absolute numbers. Cost is a model figure, not a price.