The rare-earth bottleneck is refining
Western dependency sits in the few plants that separate the elements, and China holds 91% of that refining.
The bottleneck of Western dependency sits above ground, in a handful of buildings where crushed rock is turned into separated oxide — and it is that step, not the mine, that China dominates.
According to the International Energy Agency, China held 91% of world refined output of magnetic rare earths — neodymium, praseodymium, dysprosium and terbium — in 2024, against 60% of mining output for those same elements and 94% of sintered permanent magnets. In rare earth refining as a whole, the Chinese share fell from more than 90% in 2023 to 85% in 2025. The agency does the arithmetic that matters: even if every planned refining project worldwide comes online on schedule, the Chinese share only declines to 70-73% by 2035.
CSIS reaches consistent figures through a different method: about 90% of global processing and 99% of heavy rare earth processing. Care is required with sources here, because the USGS — the official American reference — does not publish global refining shares: it publishes net import reliance for the United States (67% in 2025) and the origin of those imports (China 71%, Malaysia 13%, Japan 5%, Estonia 5%). The global figure comes from the IEA. Attributing a statement about world refining to the USGS is the kind of imprecision that circulates as if it were precision.
Why separation is the bottleneck
The name misleads: they are seventeen chemically similar elements that occur together and must be separated one by one through solvent extraction, in chains of hundreds of stages. It is precision chemistry, with thorium and uranium residues that require specific environmental licensing — the reason the only licensed conventional uranium mill in the United States, Energy Fuels’ White Mesa Mill, became a strategic asset.
The IEA describes the state of the art outside China in a sentence that needs no adjectives: “only a few industrial-scale facilities operating today — in Malaysia, the United States and Estonia”. And it gives the sector’s average timeline: about eight years for a new project to mature.
Who is building, and where
The short list is known and worth reading without optimism. Lynas in Malaysia is the only proven commercial separation chain for both light and heavy elements outside China, and in 2025 it produced the first commercial dysprosium and terbium oxides outside the country in decades. MP Materials at Mountain Pass sold 1,994 tonnes of NdPr oxide in 2025 and received from the US Defense Department a package combining $400 million in equity, up to $350 million more, a $150 million loan for heavy separation, a ten-year price floor and a magnet offtake.
Neo Performance Materials runs Silmet in Sillamäe, Estonia — a Soviet-era plant, the only industrial-scale one in Europe — and is building a magnet factory in Narva with 2,000 tonnes of annual capacity. Energy Fuels has produced NdPr at White Mesa since 2024 and presented 99.9% dysprosium at pilot scale in August 2025, but its second phase only expects regulatory approval in mid-2027 and commissioning in the first quarter of 2029. That schedule is the sector’s most eloquent data point: it is the real ramp-up time, with public money already committed.
In Europe, Solvay opened an NdPr line at La Rochelle in April 2025 without disclosing capacity, and Caremag raised €216 million to operate in Lacq — a project that will process 2,000 tonnes of recycled magnets a year and refine 5,000 tonnes of concentrate, with commissioning expected by late 2026. In Australia, Iluka’s Eneabba refinery was financed with A$1.25 billion from the Australian government in 2022 and is expected to commission in 2027.
China’s answer is not commercial
In December 2023 China banned the export of rare earth extraction and separation technologies — not the product, the knowledge. In October 2025 it went further and created an extraterritorial mechanism: goods manufactured abroad with 0.1% or more Chinese-origin rare earth inputs now require a licence from Beijing. That control has been suspended for a year, valid until November 2026 — that is, in force as of this edition.
The mine is geography. Refining is knowledge, licensing and chemistry — which is why China can lose mining share without losing any power at all.
What Western money buys
The West answered with money, and the numbers are large. The White House speaks of 160 critical minerals deals since January 2025, more than $40 billion, including a strategic stockpile backed by a $10 billion Exim Bank loan. The European Commission adopted the €3 billion RESourceEU plan in December 2025, and its critical raw materials regulation sets 2030 benchmarks — 10% extraction, 40% processing, 25% recycling and no single third country above 65% of consumption.
Benchmarks and stockpiles buy time. They do not buy capability: none of those targets changes the fact that separation demands years of licensing, calibrated chemistry and trained people who today work overwhelmingly in one country. The question that decides the next decade is not how many mines the West will open — it is how many separation plants can operate without a Chinese engineer, and in what year.