This group makes the raw stuff behind permanent magnets — the small but essential magnets that turn electricity into motion (and motion into electricity). They come from rare-earth elements (REEs): a family of 17 metals, of which a handful really matter for magnets — neodymium, praseodymium and dysprosium (the scan file names these three; terbium is a fourth magnet REE commonly cited in industry sources) est.. The companies dig the ore out of the ground, then run it through a chemically complex separation and refining step (splitting apart 17 chemically near-identical elements) to turn it into purified rare-earth oxides (the powder form sold by the tonne), and a few go one step further and press those into finished NdFeB magnets (neodymium-iron-boron, the strongest commercial magnet type). These magnets are inside the motor of almost everything that needs to be small, strong and efficient: data-center cooling fans, electric-vehicle and humanoid-robot motors, hard-drive actuators, wind turbines, and precision defense systems. The catch is geographic, not geological: rare earths are not actually rare in the ground, but China controls roughly 60% of mining and 85-90% of the world's refining — the bottleneck step (Source: scan file, sector 05).
In plain terms: the scan file describes future demand for magnet rare earths as outrunning the supply that exists outside China, with the binding limit being not ore in the ground but refining/separation capacity, which takes years and a lot of capital to build and requires hard-won chemical know-how. Because there are few Western producers, the leading names have at times traded at high multiples of current revenue — on the order of roughly $10 to $30+ of market value per $1 of this-year sales est. (not-live-verified) — meaning a buyer is paying today for capacity and tonnes that mostly do not exist yet. These are facts and arithmetic for the reader to weigh, not a recommendation.
The physical product is sold in three escalating forms, and money is made at each step:
How the cash actually arrives: a miner sells tonnes of oxide (or magnets) at a market price, and its profit is (price per tonne) minus (cost to mine and refine per tonne). The leverage in this business is large because mining/refining is high-capex (capex = capital expenditure, the upfront money to build mines and plants) and high fixed cost — once the plant is built, extra tonnes are cheap to make, so when the oxide price rises the profit tends to rise faster, and when it falls the company can burn cash. Some contracts are structured as offtake or price-floor deals (offtake = a buyer agrees in advance to purchase a set volume; a price floor = a buyer, or the US government, guarantees a minimum price) — this converts a volatile commodity into something closer to a contracted cash stream. contracted revenue under such floors is more predictable than spot-price (today's open-market price) revenue.
Source: scan file sectors 05 & 13; product-form breakdown from general industry knowledge.
Today, the dominant buyers of magnet rare earths are EV motor makers, wind-turbine makers, consumer electronics/hard drives, industrial automation, and defense. Global demand for the magnet "heavy lifters" (NdPr oxide) is on the order of tens of thousands of tonnes per year est., and the whole rare-earth-oxide-plus-magnet value chain is a roughly $10-15B/yr est. market. These are order-of-magnitude figures, not live-verified.
The forward driver (AGI lens): the scan file states that the AI build-out is a physical build-out. Reasoning from the premise that AGI is arriving — more compute and more physical automation both need more magnet motors — rare-earth magnets sit on multiple demand curves at once:
Hard vs forecast: Today's demand (tens of thousands of tonnes, ~$10-15B market) is approximate and not live-verified. The step-change from humanoid robots is a forecast — it depends on robots actually shipping in volume, which has not happened yet. The direction (more compute and more physical automation both need more magnet motors) follows from the AGI premise; the magnitude and timing remain open questions.
Supply has two separate layers, and they are constrained very differently:
Why the bottleneck is sticky: the limit is not a single physical input — it is a stack of capital + lead time + process know-how + environmental permitting (separation generates difficult waste streams). The scan file also notes China has used this position as leverage: it embargoed exports to Japan in 2010 and imposed fresh export controls across 2023-2025. That history is the scan's stated reason the US government is now subsidizing and price-supporting Western capacity — the scan flags this as a "national security bottleneck."
Market-share structure (who controls supply):
Source: China shares, single-US-mine fact, 5-10yr build time, export-control history — scan file sector 05. Lynas/USAR positioning and global magnet shares — general industry knowledge (not-live-verified).
Putting the two sides together: per the scan file, the product is short outside China and roughly balanced-to-long inside China. That distinction is the whole game. The world is not running out of rare-earth atoms; it is short of non-Chinese refined supply, and that shortage is what every Western producer and Western government is trying to fix.
| Signal | What it shows | Direction |
|---|---|---|
| NdPr oxide price (spot) | Volatile; spiked in 2021-2022, fell back, with renewed strength as export controls bit est. | Up over multi-yr, choppy |
| Non-China refining capacity | Small vs demand; takes 5-10 yrs to add (scan file) | Structurally short |
| Government floor prices / offtake | US has stepped in with price floors & offtake to guarantee Western volumes contracted est. | Signals shortage |
| New demand layer (robots) | Humanoid-robot magnet demand forecast, 2027-2030 window (scan file) | Widens the gap |
| China supply lever | Can raise output (pressure price down) or restrict (pressure price up) | Two-way risk |
When could it flip to oversupply? Two distinct ways. (1) Inside China, the scan describes it as already structurally long — Beijing can raise output and lower the global price, which is the largest single risk to a Western producer's economics. (2) Outside China, a true Western glut would require the announced mines and separation plants to all finish and ramp at once — given 5-10 year build times and frequent slippage, a non-Chinese oversupply is years away at the earliest and is a forecast scenario, not a current fact. The near-term question is therefore less "will demand exceed Western supply" and more "will the Western producer's selling price stay above its cost when China can move the global price." That is the stated reason government price floors exist, and floors change the risk profile of the cash flows. Facts and arithmetic; the reader judges.
| Company (ticker) | What it makes | Exposure to magnet rare earths | Rough size | Position |
|---|---|---|---|---|
| MP Materials (MP) | Mines ore at Mountain Pass; building US separation + US NdFeB magnet plant | Rare-earth-focused (~95-100% of revenue) est. | Large-cap of the group est. | Only operating US mine; moving up the chain to magnets; US gov't support (scan file) |
| Lynas (LYSDY) | Mines in Australia, refines separated oxides in Malaysia; new Texas plant | Rare-earth-focused (~95-100%) est. | Large-cap; described as biggest ex-China oxide producer est. | General sources describe it as the at-scale non-Chinese separator already running est. |
| USA Rare Earth (USAR) | Building NdFeB magnet plant (Oklahoma); developing Round Top deposit | Rare-earth-focused, pre-scale forecast | Small/early-stage est. | Magnet-first US strategy; mostly future capacity forecast |
| Energy Fuels (UUUU) | Uranium + rare-earth separation at White Mesa | Diversified — RE is a slice; uranium is core est. | Small/mid est. | Has separation capability; RE is a smaller line alongside uranium (scan file sector 13) |
| Texas Mineral Res. (TMRC) | Round Top deposit (development stage) | Rare-earth-focused, development-stage forecast | Micro-cap est. | Resource at development stage, no production (scan file) |
| TMC the metals co. (TMC) | Seabed nodules (nickel, cobalt, copper, manganese) | Adjacent, not RE — battery/critical metals, pre-revenue forecast | Small est. | "Critical minerals" exposure but not a rare-earth magnet producer; deep-sea regulatory risk |
Source: MP, UUUU, TMRC roles & the single-US-mine fact — scan file sectors 05 & 13. LYSDY, USAR, TMC and all size/percent figures — general knowledge, not-live-verified.
Translate it to money-in/money-out. This is a high-capex group: the leaders are spending billions today (mines, separation plants, magnet factories) to produce tonnes that arrive later, so several of them generate little or no owner cash (free cash flow — cash left over after running and reinvesting in the business) right now — and the pure development-stage names (TMRC, much of USAR) generate none yet.
Arithmetic summary: a buyer today pays a high multiple of present sales (~$10-30+ per $1 of revenue est., not-live-verified) for exposure to a non-Chinese supply gap the scan describes as real, while the selling price — and therefore the profit — is sensitive to a global price that one large producing country can move, except where a government price floor applies. No verdict implied; the reader weighs the numbers.
Source: capex-heavy structure & gov't-support point — scan file sectors 05 & 13; revenue-multiple and capacity figures — general knowledge, not-live-verified.
Factual descriptions of where a company-level deep-dive would add the most information — not a ranking or recommendation:
Factually: the two already-producing rare-earth-focused names are MP and Lynas; the rest are either diversified (UUUU), pre-production (TMRC, much of USAR), or in an adjacent metals category entirely (TMC).
Confidence: HIGH on the qualitative supply structure and the bottleneck location (refining, not ore); MEDIUM on directional demand growth under an AGI ramp; LOW on precise dollar figures and per-company sizes, which are not live-verified.