An alert about Beijing is not a magnet. The thing worth watching is a named product, leaving a named plant, on a date the company has already printed. These three names are three stages of that test. None of them is a bid.
China's rare earth dominance is real. It is also the wrong size in most headlines. The country mines most of the world's rare earth elements. It separates a larger share than it mines. It makes almost all of the permanent magnets those elements become. An investor who stops at the mine has stopped before the leverage. The alert is about a licence desk in Beijing. The metal the factory needs is a powder, and then a magnet, and the magnet is mostly made in China.
Here is the only idea. A rare earth stock is worth watching when you can name the product, the plant, and the date that product has to leave that plant. An alert is not that date. A map of a deposit is not that date. A minister's comment that may or may not have changed a written rule is not that date. Below, three companies sit at three different points on the same road. One has started to ship magnets. One already separates oxides, including a thin stream of the heavy rare earth elements Beijing licences. One has finished the engineering on a plant it has not financed and has not built. Watching them means checking the next page against the date they already printed. It does not mean buying them because China is large.
This is not a recommendation to buy, sell, or hold Neo Performance Materials, Lynas Rare Earths, Aclara Resources, or any other security. It is not investment advice and not a solicitation. Rare earth stocks move on politics, on plant delays, and on prices the companies do not set. You can lose money. The phrase rare earth investment opportunities does not get an answer on this page. It gets a test, and the test is a calendar.
Two shares, and they are not the same fact
Start with the mine, because that is the number people memorize, and then leave it. On figures carried from the U.S. Geological Survey's 2026 summaries, the world mined about 390,000 tonnes of rare earth oxide in 2025. China mined about 270,000 of those tonnes. That is about 69 percent. The United States was about 13 percent. Australia was about 7 percent. A 69 percent share is a lead. It is not a monopoly of the hole in the ground. Other countries dig this rock up.
The magnet rare earths are a tighter set, and the International Energy Agency counts them apart from the broad mine total. For neodymium, praseodymium, dysprosium, and terbium together, the agency's 2024 picture had China at about 60 percent of mining, about 91 percent of refining, and about 94 percent of sintered permanent magnet output. Do not average 69 and 60 into one "China share." One number is all rare earth oxide from the mine in 2025. The other is the four elements that go into the magnets, at three stages, in 2024. The direction is the point. The lead gets wider after the mine.
A later agency note, posted on October 12, 2026, said rare earths were the exception in a world where mineral refining was growing more concentrated. The top supplier's share of rare earth refining fell from 90 percent in 2023 to 85 percent in 2025. That is a different series again. It says the grip loosened by a few points. It does not say the grip is gone. Eighty-five percent of refining is still a desk you cannot route around in a quarter. A rare earth market outlook that quotes only the mine share is describing the easy part.
Ore can cross an ocean. A separated oxide is a product. A sintered magnet is a part that fits a motor. China sits on the last two steps. For years, rock mined in California or Western Australia was cracked, leached, and split in plants that were not next to the mine. Opening another pit does not open a separation hall. The critical minerals supply chain for these elements fails at the hall, and again at the magnet press. Critical minerals stocks that own only the pit own the step the world is least short of.
Two Chinese rules, and a date the headlines do not share
The alert this month is a deadline. The deadline is not one sentence, and the sources do not agree on whether it moved. Treat that disagreement as the fact. An investor who picks the date that flatters a trade has already left the page.
There are two tracks. The first has been in force since April 4, 2025. China's Ministry of Commerce Announcement No. 18 requires a licence to export seven medium and heavy rare earth elements: samarium, gadolinium, terbium, dysprosium, lutetium, scandium, and yttrium. Trade lawyers who follow the file say this track was never suspended. Dysprosium and terbium are the elements that let a permanent magnet keep its strength when it gets hot. They sit inside motors, weapons, and some wind machines. A rule on those two is not a press release. It is the licence on the scarce additive.
The second track is the package announced in October 2025. It went wider than the April list. The notices that were later paused added further elements, including holmium, and they covered rare earth equipment, separation technology, and magnet-making know-how. Those extra elements sit inside the pause. They are not on the live April list. The same package carried a rule that would reach some goods made outside China. That piece is often called a 0.1 percent rule. In plain words, a foreign-made part could need a Chinese licence if Chinese-origin controlled rare earths were 0.1 percent or more of the part's value. That is a small share. It is why a motor plant in another country can still be inside a Chinese form. This package was suspended in November 2025 for one year, under what filings and lawyers call Announcement No. 70. The written end of that pause, in the announcement, is November 10, 2026.
Did the pause get extended? As of the week of October 12, the public record was split, and the split is narrower than a headline makes it. On September 28, 2026, a commerce ministry text said the Kuala Lumpur trade arrangement with the United States would run to January 10, 2027. The sentence that has been quoted does not name Announcement No. 70, and it does not name rare earths. Some trade notes still treated that truce date as an extension of the rare earth suspension. Other notes, including lawyers reading the ministry site as late as October 11, said no new announcement had changed the November 10 date printed in Announcement No. 70. January 10 is a truce date. It is not yet a new expiry printed on the rare earth pause. On October 8 and 9, China's commerce minister and the European trade commissioner said both sides would work to keep licence approvals moving for rare earths, permanent magnets, and dual-use goods. A promise to facilitate a form is not a repeal of the form.
What is not in dispute is the live list. The April rule on the seven heavy and medium elements has kept running. Magnet exports that contain terbium or dysprosium have kept needing a licence. Three large Chinese magnet makers, JL MAG, Beijing Zhong Ke San Huan, and Yunsheng, have been reported to hold broader civilian licences. Defence and aerospace buyers have not been in that easier lane. China rare earth exports of the heavy oxides people worry about, including yttrium, dysprosium, and terbium, were reported around half of pre-restriction levels in customs tallies cited this autumn. The pause, if it covers only the October package, never turned those taps fully back on.
So the alert has a clock, and the clock has two hands. One hand says November 10, and that hand is the date printed in Announcement No. 70. One hand says January 10, and that hand is a trade-truce sentence that does not name the rare earth notices. Between them sit talks, a European pledge, and a midterm election calendar in the United States that has nothing to do with solvent extraction. Rare earth processing outside China does not get built in the gap between those two dates. A stock that reprices because a headline picked one hand is repricing a rumor about a form. The plant either separates dysprosium or it does not.
The tonnes fell. The money did not.
Customs data carried by Reuters this week put a number on the squeeze. China exported 4,269.8 tonnes of rare earths in September 2026. That was up 6.7 percent from a year earlier and down 9.8 percent from August's 4,735 tonnes. The first nine months totaled 43,711 tonnes, down 9.6 percent from the year before. A market note, not that customs story, put September about 22 percent under a May peak near 5,490 tonnes. These are shipment tonnes. They are not a reserve, and they are not a price.
The value went the other way, and that line is a market note, not the Reuters customs story. For January through August, the note put the value of those exports up about 53.5 percent from a year earlier, even as the tonnes were down. Less rock and powder left the country. More dollars were attached to what left, if the note is right. That is what a licence does when it binds. It does not have to ban the trade. It can thin it, slow it, and lift the price of what gets through. A rare earth elements market that looks only at tonnes will call this a surplus of worry. A buyer paying the invoice will not.
Magnets, not ore, are the part the motor plant notices. Chinese magnet shipments to the United States were reported at 512 tonnes in August 2026, down 20 percent from July and 13 percent from a year earlier. Five hundred tonnes is not the world's magnet supply. It is one corridor, one month, already smaller. Permanent magnets are where the 94 percent share becomes a delivery. A mine in another country does not replace a missed box of magnets. A separation plant might, if it is running, and if someone presses the oxide into a magnet after that.
One market note this month also said China accounts for about 84 percent of neodymium-praseodymium oxide supply and of separation capacity, against about 61 percent of mine output for that stream, and that supply outside China could fall short of 2026 demand by roughly 32,000 tonnes of that oxide. Treat the 32,000 as that note's figure, not as a census. The shape is the same shape as the agency's 60, 91, and 94. The shortfall people talk about is a shortfall of separated product, not a shortfall of dirt.
What "to watch" is allowed to mean
Rare earth stocks to watch is a search phrase that usually means a shopping list. It should mean a short list of claims you can falsify. A claim is falsifiable if it has a product, a place, and a date. "Exposed to China's rare earth dominance" is not a claim. Every motor maker is exposed. The question is whether the company you are reading sells the thing the licence is about, or hopes to, or merely owns ground that contains it.
The three names below were chosen because each one answers a different version of that question, and because each one printed fresh paper in 2026. Neo Performance Materials, listed in Toronto, has started commercial shipments of magnets from a plant in Estonia. Lynas Rare Earths, listed in Australia, is the large separated-oxide producer outside China, and its heavy output is still small next to its light output. Aclara Resources, also listed in Toronto, completed basic engineering this week for a separation plant in Louisiana that does not exist yet. Canadian rare earth stocks, in this piece, means the two Toronto names. The Australian name is here because a watchlist that stays inside one exchange will miss the plant that already runs. TSX rare earth stocks are not a sector with one cost sheet. They are two files in a drawer that also holds a third file from Perth.
None of these is the only file. A mine-to-magnet chain in California and Texas has also started to move metal into magnets for an automaker, in small volumes, with a larger plant aimed at the end of 2028. That chain is not one of the three, because this page is already long, and because adding a fourth name would start to look like a basket. A basket is how a slogan replaces a date. If your own card has a different plant on it, use the same test on that plant. Do not import this list as a set.
Neo: a magnet has left the building
Neo Performance Materials said on September 14, 2026, that it had begun commercial shipments of sintered rare earth magnets from its plant in Narva, Estonia. The customer is a Tier 1 maker of electric-vehicle traction motors. The company did not name the customer. It did not disclose how many tonnes have left. A first shipment is a change of status. It is not a run-rate. Rare earth mining stocks get the headlines. This one is not a mine. It is a magnet plant next to an older separation plant, owned by a Canadian company, selling into Europe's auto supply chain.
The plant opened in September 2025. Phase 1A has a nameplate of about 2,000 tonnes of magnets a year. The company has said it is designing a next phase, often called Phase 1B, that would take nameplate capacity to about 5,000 tonnes. Detailed engineering, equipment orders, and the layout are underway. They are not a finished hall. A nameplate is the sign on the door. It is not last month's output. One Estonian report said the current plant was expected to reach full capacity at the end of 2028. That is a local report of a ramp, not a tonne you can put in a model today. Neo has not, in the notices used here, published a utilization rate.
The commercial start sits inside a longer list of programs. Neo has said it holds multiple magnet programs from three Tier 1 motor makers, and that two or three more programs should enter commercial production before the end of 2026. Qualification samples have gone out. Traction-motor magnets are the hard ones. They spin, they heat, and a car company does not switch them for a slogan. Auto programs are often awarded for the life of a vehicle platform. That can mean years of volume once a program is real. "Awarded" is not "shipped." The page to watch before December 31 is whether those two or three programs do enter production, and whether the company prints a tonne figure next to the word commercial. A rare earth supply chain is a sequence of those pages. It is not the opening ceremony.
Behind the magnet plant, at Sillamäe, Neo runs a light rare earth separation plant. Reports this autumn said a line to separate heavy rare earth elements, including dysprosium and terbium, was commissioned there in April 2026. If that line is running, Neo is one of the few companies outside China that can talk about both ends: a heavy split, and a magnet press. This page has not re-read a tonne of dysprosium from that line. Commissioned is not "sold 50 tonnes." The watch item is the first sales number for those heavy oxides, in a filing, with a buyer who is not a slide.
Further out, Neo has said it wants about 20,000 tonnes a year of magnets through plants in more than one country. The company estimates that could be 10 to 15 percent of projected rare earth permanent magnet demand outside China. Read the verb. Estimates. Projected. Outside China. Twenty thousand tonnes is four times the Phase 1B drawing, and ten times the hall that just made its first commercial shipment. It is a direction. It is not capacity. China's 94 percent of sintered magnets, on the 2024 agency figure, is not moved by a 2,000-tonne nameplate. It would not be moved by 5,000. It might be nicked by 20,000, years from now, if the oxide shows up and the customers stay. A nick is not independence. Call the ambition an ambition.
The risk that sits next to the good news is feed and price. A magnet plant consumes neodymium-praseodymium, and often a little dysprosium or terbium. If those oxides are scarce or expensive, the plant's margin is the magnet price minus a metal price Neo does not control. If the oxides are suddenly plentiful because a licence wave releases Chinese material, the premium a European buyer will pay for a non-Chinese magnet can shrink. Neo's watchers should want two numbers the first shipment did not include: tonnes shipped, and what the metal cost inside them. Without those, "commercial" is a true word and an incomplete sentence.
Lynas: the separator that already runs
Lynas Rare Earths is the contrast. It does not need a groundbreaking photo. In the year ended June 30, 2026, the group produced about 13,089 tonnes of rare earth oxides, including about 7,260 tonnes of neodymium-praseodymium, on figures compiled from its reports. In the June quarter alone it produced about 3,481 tonnes of rare earth oxides and about 1,857 tonnes of neodymium-praseodymium. Revenue for the year was reported around 978 million Australian dollars, and profit after tax around 222 million. Those are operating numbers. They are not a forecast, and they are not a reason to own the shares. They are the proof that separation outside China is not a theory.
The heavy stream is the part that matches the alert, and it is much smaller. In the June quarter, dysprosium and terbium output rose to about 19 tonnes. Nineteen. Not nineteen hundred. A year of quarters like that is still a small pile next to the heavy oxides China ships, licences, or withholds. Lynas first separated dysprosium and terbium in Malaysia in 2025. It first produced samarium oxide in March 2026. Gadolinium, on the company's own timing in that quarterly, is expected in early fiscal 2028. A heavy rare earth separation expansion at the Malaysian plant, aimed at as much as 5,000 tonnes a year of heavy feedstock, was underway as of July 2026. Feedstock capacity is not oxide in a drum. The watch item is whether those heavy tonnes keep rising in the next quarterly, and whether gadolinium arrives when the company said it would.
The geography is the whole argument in one company. Mt Weld, in Western Australia, is the mine. Kalgoorlie cracks and leaches and makes a mixed carbonate. Malaysia separates and finishes. A mixed carbonate is not a magnet. It is the middle step, and Lynas owns the middle step outside China at a scale nobody else in the West really matches. That is why the company is on this page. It is also why a problem in Malaysia or a bottleneck at Kalgoorlie is not a side note. In the June quarter the company itself pointed to a water-recycling plant and to concentrate quality at Mt Weld, and to work at Kalgoorlie to get precipitation up to design. The tonnes still came. The tonnes are not automatic. Rare earth processing is a chemical plant that eats acid, power, water, and time. A dominance story that assumes the outside-China plant runs at nameplate is a story that has not read a quarterly.
Lynas has also tied a large share of its future sales to Japanese buyers through a long arrangement. Reports of that arrangement describe a firm 5,000 tonnes a year of neodymium-praseodymium, with a floor near 110 U.S. dollars a kilogram, and room to make more available, up to about 7,200 tonnes, out toward 2038. On the heavy side, the same reports describe a firm half of heavy oxide output, with room up to about three quarters, and a March 2026 widening beyond dysprosium and terbium to include samarium, yttrium, lutetium, and gadolinium. This page has not re-read the contract. If the floor and the percentages are right, they matter more than a spot price on a volatile day. A floor is a buyer who has already signed. It is not a guarantee the plant hits the tonnes. Half of a small heavy output is still a small number. Half of a larger one is the point of the expansion.
Set Lynas next to the Chinese export tape and the scale problem is honest. China moved about 4,300 tonnes of rare earth products in one September. Lynas's June quarter of all oxides was about 3,500 tonnes, and its heavy oxides were 19. The company can be the most important separator outside China and still be a fraction of the trade the licence desk touches. Investors who want a "counter to China" in one ticker are asking one plant to do a country's job. The quarterly is the antidote. Read the 19. Then decide what the word counter is allowed to mean.
Aclara: a drawing with a price on it
Aclara Resources is the third stage, and it is the one a headline will most abuse. On October 13, 2026, the Toronto-listed company said it had finished basic engineering for Project Dynamo, a planned separation plant in Louisiana. Hatch did the engineering. The construction estimate is about 494 million U.S. dollars, and the company calls it a Class 3 estimate. In the cost-estimating scale used on big jobs, Class 3 is tighter than a sketch and looser than a finished bid. It is not money spent. It is not a plant. Site work is targeted for the first quarter of 2027, subject to financing and the usual conditions. Construction is targeted to finish in the second half of 2028. Both dates are targets. The financing condition is the sentence under them.
The capex moved because the plant got bigger, not only because prices drifted. An earlier Class 5 figure, disclosed in April, was about 277 million dollars. Engineering News-Record put the new estimate about 78 percent above that sketch. The company said dysprosium and terbium separation capacity rose about 42 percent so the plant could take feed from Chile as well as Brazil, and that new circuits were added for yttrium, gadolinium, and samarium. More metal, more tanks. A higher number is not, by itself, a failure. A higher number that is not funded is a wish with better stationery.
The design output, at the life-of-project average of the two upstream projects if they run at full capacity, is about 200 tonnes a year of dysprosium oxide, 35 tonnes of terbium oxide, 1,350 tonnes of yttrium oxide, 180 tonnes of gadolinium oxide, 190 tonnes of samarium oxide, and 1,300 tonnes of neodymium-praseodymium oxide. Read the "if." The feed is supposed to come from the Carina project in Brazil and the Penco module in Chile, as a mixed rare earth carbonate the company calls super-pure. Those are ionic-clay projects. They are not operating mines. Penco, the company has said, has an environmental approval and still needs financing before construction. If the mines slip, the 200 tonnes of dysprosium slip with them. A separator without feed is a building in Louisiana.
The oxides on that list are not random. Dysprosium, terbium, yttrium, gadolinium, and samarium are on, or next to, the April 2025 Chinese licence list. Neodymium-praseodymium is the volume metal in the magnet. Aclara is trying to build the exact split the alert is about. That is why it is on a watchlist. It is also why a press release about engineering is not supply. Two hundred tonnes of dysprosium oxide a year would matter, if it existed, because Lynas's latest quarter of dysprosium plus terbium was 19 tonnes and because Chinese exports of those heavies have been thin. "Would matter" is the right tense. The plant's own calendar says the earliest construction finish is the back half of 2028, and that date is conditional.
There is a small plant that does exist. At a demonstration line in Blacksburg, Virginia, Aclara has said it produced neodymium-praseodymium oxide to its target spec in May 2026. Dysprosium and terbium oxide to spec are targeted for this quarter, the fourth quarter of 2026. The company also said the yttrium circuit was being installed, and that gadolinium and samarium equipment was expected later in the same quarter. Today is October 15. The quarter is not over. A target for this quarter is a date you can check before Christmas. It is the cleanest near-term item on the Aclara page. If the drums are late, the Louisiana flow sheet is a model that missed its first public test. If the drums are on spec, the company has shown a chemistry at demonstration scale. Demonstration scale is not 200 tonnes.
The company also published economics, and they need a fence around them. Operating cost inside the separation plant was cut, on its estimate, from about 38 dollars a kilogram of oxide to about 26, a drop of about a third. That 26 dollars excludes the cost of the carbonate feed. Feed is not a rounding error. You cannot subtract 26 from a sale price and call the difference a margin. Using prices from Argus, the company projects an after-tax internal rate of return of about 28 percent and a payback of about 2.6 years. That is the company's arithmetic on the company's case. It is not a return you have been promised. It is not a return on the shares. It moves if the capex moves, if the feed is late, if the Argus price is wrong, or if the recovery is worse than the study. Recoveries in the engineering note are not one number. Dysprosium is about 97.9 percent, terbium about 97.7, samarium about 96.3, and neodymium-praseodymium about 98.9. Yttrium is about 85.7 percent. Gadolinium is about 86.7 percent. Purity targets are not one number either. Dysprosium, samarium, and neodymium-praseodymium are aimed at 99.5 percent. Terbium, yttrium, and gadolinium are aimed at 99.9 percent. Those figures are design aims in a study. They are not a plant log.
Two more sentences from the company belong on the card, because they are easy to upgrade by accident. The company says its carbonate has tested as non-radioactive for shipment, with support from SGS Lakefield, and that this is true in the places it operates, with room to spare. A shipping classification is not a mine permit and not a sales contract. And the company is in preliminary talks with possible buyers of yttrium and other oxides. It has no definitive offtake. Lynas can point at a Japanese floor. Neo can point at a motor maker who took magnets. Aclara can point at a conversation. Preliminary is the honest word. Leave it there.
Yttrium is the element Aclara wants you to notice, and the notice is fair as geology and unfinished as business. The company says about half the rare earth content at Penco and about 30 percent at Carina is yttrium, and that the two projects could, at full rates, put about 1,400 tonnes of yttrium a year into carbonate. Yttrium is on the April licence list. Western buyers have been short of it. A resource share is not a shipment. The Louisiana circuit that would separate it was added in the October redesign. It is a line on a drawing of a plant that still needs a cheque.
Three stages, not three peers
Put the three on one table and refuse the ranking. Neo has shipped a magnet and has not told you the tonnes. Lynas has shipped oxides for years and has told you the tonnes, including a heavy tonne-count that is still small. Aclara has shipped neither oxide nor magnet from the plant in the headline, because that plant is an estimate. A table that sorts them by "upside" has invented a price target. A table that sorts them by stage is just a description.
The cash question is different at each stage. Neo's magnet hall has a customer and a nameplate, and its margin depends on a metal price and a ramp the company has not fully disclosed. Lynas already has revenue, profit, and a quarterly, so the argument is about the mix, the heavy expansion, and the plants it must keep running. Aclara's next hard fact is still a closed financing. On September 4, 2026, the U.S. Export-Import Bank issued a letter of interest for up to 750 million dollars, with a term of up to 15 years. The letter is not a loan. It is non-binding, and it still has to clear due diligence, underwriting, and other approvals. Louisiana has also put an estimated 46.4 million dollar package of incentives and support on the table, still subject to program rules. Neither paper spends the 494 million. Four hundred and ninety-four million dollars is the plant, before the mines that have to feed it. Equity, debt, a government cheque that has actually closed, or a customer prepay are different kinds of dilution and delay. Until one of them is signed, the internal rate of return is a spreadsheet. Rare earth investment opportunities that skip the cheque are opportunities in the grammatical sense only.
The political question is also different. A November or January change in Beijing changes the premium a buyer will pay for a non-Chinese magnet or a non-Chinese heavy oxide. Neo and Lynas can feel that premium now, in opposite directions: a tighter Chinese licence can help their price, and a surprise loosening can hurt it. Aclara feels it mostly as a story, because its product date is 2028 if everything holds. A story can reprice a stock in a week. It cannot deliver a kilogram. Investors who buy the story and think they bought the kilogram have mixed the stages.
Light metal, heavy metal, and the licence in between
Permanent magnets made of neodymium, iron, and boron are the volume product. Neodymium and praseodymium are the light rare earths in that recipe. Dysprosium and terbium are added so the magnet does not quit when the motor is hot. Most of the mines people announce outside China are better at the light fraction than the heavy one. Ionic clays, of the kind Aclara is chasing in South America, are interesting because they can carry the heavies without the radioactive baggage of some hard-rock ores. Interesting is not produced.
China's April rule bites the heavies and a few special metals, not the entire light stream in the same way. A company that sells only neodymium-praseodymium is in the magnet boom. It is not, by that fact, the answer to a dysprosium licence. Lynas sells a lot of the light oxide and a little of the heavy. Neo wants to put both into a magnet in Estonia. Aclara's drawing is heavy-first, with a large yttrium line and a neodymium line beside it. The rare earth elements market is several markets that share a name. A stock screen called "rare earth" dumps them in one cell. The licence does not.
That is also why a critical minerals supply chain slogan fails. Copper, lithium, and rare earths are not short in the same place. Rare earths are short, outside China, in separation and in magnet pressing, and shortest of all in the heavy split. A policy that funds another light-rare-earth pit and calls it resilience has funded the step the agency chart says is the least concentrated. The chart is not an order to buy the companies that own the later steps. It is an order to stop pretending the pit is the later step.
The magnet is the product the factory buys
Follow one kilogram. Rock leaves a pit as ore. A concentrator lifts the grade. A cracking plant breaks the mineral. A separation plant, full of mixer-settlers and solvent, splits one element from the next. That split is slow, dirty, and skilled, which is why so few plants outside China do it. An oxide then becomes a metal, then an alloy, then a pressed and sintered magnet, then a coated part that a motor maker will qualify for a year or more. Each arrow is a plant, a permit, a power bill, and a customer who can say no.
China's rare earth dominance is the fact that most of those arrows still land inside one country. The October 2025 package, paused in writing until November 10, tried to follow the arrows abroad, including the know-how and the 0.1 percent value test. The April 2025 rule does not need to follow them abroad. It sits on the oxide as it leaves. A Western magnet plant that still buys Chinese dysprosium is still on the April form. Neo's heavy line, if it sells, is an attempt to step off that form. Lynas's Malaysian heavies are the same attempt, already in tonnes, small tonnes. Aclara's Louisiana plant is the attempt on paper.
Qualification is the dull step investors skip. A car maker does not take a magnet because the oxide was non-Chinese. It takes a magnet that survives heat, vibration, and a production line. Neo's shift from samples to a commercial traction-motor program is that dull step, completed once. Two or three more programs, if they land this year, are the same step again. Lynas's Japanese arrangement is a different dull step: a buyer of oxide, with a floor, over many years. Aclara's preliminary talks are the step before the dull step. The rare earth supply chain is mostly waiting. The waiting has a name. Call it qualification, or financing, or a licence. Do not call all waiting "patience" and then call the stock cheap.
What an investor can actually write down
You can write the Chinese dates on a card and refuse to merge them. April 4, 2025: seven elements, licence in force. November 10, 2026: the written end of the pause on the wider package, including the extra elements such as holmium, unless a gazette you can read has moved it. January 10, 2027: the end date written for the Kuala Lumpur trade arrangement, which does not, in the sentence that was published, rename Announcement No. 70. October 8 and 9, 2026: a European meeting that promised easier approvals and did not repeal the rule. If a headline uses one date, ask which instrument it means. China's rare earth dominance does not require you to guess. It requires you to know which form is alive.
You can write the shares without blending them. About 69 percent of 2025 mine output, all rare earth oxide. About 60 percent of 2024 magnet-rare-earth mining, about 91 percent of that refining, about 94 percent of sintered magnets. About 85 percent as the top refiner's share in 2025, down from 90 in 2023. Four sentences. Four sources of the number. A rare earth market outlook built on a single percentage has chosen a stage and hidden the choice.
You can write three company dates. Neo: first commercial magnet shipment announced September 14, 2026; two or three more programs aimed at before year-end; nameplate 2,000 tonnes now, a design for 5,000, an ambition of 20,000. Lynas: June quarter, about 1,857 tonnes of neodymium-praseodymium and about 19 tonnes of dysprosium plus terbium; gadolinium aimed at early fiscal 2028; a heavy expansion underway. Aclara: engineering done October 13; demonstration dysprosium and terbium aimed at this quarter; site work aimed at the first quarter of 2027 if financed; construction finish aimed at the second half of 2028; capex about 494 million dollars; a non-binding export-bank letter, not a loan; no signed offtake. Rare earth stocks to watch are stocks whose next document can hit or miss those lines.
You can refuse the flattering move. Do not call a 2,000-tonne magnet hall a break in a 94 percent share. Do not call 19 tonnes of heavy oxide a flood. Do not call a Class 3 estimate a plant, or a company return case a shareholder return, or a feedstock-excluded cost a margin. Do not call a preliminary customer talk a contract. Do not call ionic clay "clean" because a test said a carbonate was non-radioactive to ship. Do not call every Toronto ticker a producer. Two of the names on this page are Canadian listings. Only one of them has sold the product in the headline, and that product is a magnet made in Estonia, not a rock mined in Canada.
What would make this reading wrong
The reading is wrong if the mine share is not about 270,000 tonnes out of about 390,000 in 2025, or if the agency's magnet-rare-earth split was not about 60, 91, and 94 for 2024. Those are different years and different baskets. If a revision merges them, do not keep a sentence that treats them as one number. The reading is also wrong if the October 12, 2026 agency comment did not put the top refiner's share at 85 percent in 2025, down from 90 in 2023. A loosening is a fact. It is not a small fact. It still leaves a dominant refiner.
The reading is wrong if Announcement No. 18 was suspended, or if the seven elements are not the ones named here. It is wrong if a ministry gazette you can read has already moved the October package from November 10, 2026, to January 10, 2027, or further. In that case, use the gazette and throw out the comment-versus-comment section. The split in this piece is a description of the record in mid-October, not a claim that Beijing cannot write a new page tomorrow. The September and nine-month tonnes are wrong if customs did not show 4,269.8 tonnes and 43,711 tonnes. The value rise of about 53.5 percent through August, the May peak near 5,490 tonnes, and the August magnet shipments of about 512 tonnes to the United States are market notes. They are wrong if those notes were wrong. They were never the customs release itself. A licence story that cannot survive a customs revision is a mood.
The reading is wrong if Neo did not announce commercial magnet shipments on September 14, 2026, from Narva, to a traction-motor supplier, or if Phase 1A is not about 2,000 tonnes of nameplate and the next design is not about 5,000. It is wrong if the company has since printed a utilization figure this piece missed. Use the printed tonnes. The April heavy-line commissioning is from later reports of the company's facilities, not from a tonne table re-read here. If that line was not commissioned, say so, and do not give Neo credit for a heavy split it does not have. The 20,000-tonne, 10-to-15-percent line is the company's estimate of a future market. It is not demand.
The reading is wrong if Lynas did not produce about 1,857 tonnes of neodymium-praseodymium and about 19 tonnes of dysprosium and terbium in the June 2026 quarter, or if full-year oxide output was not about 13,089 tonnes including about 7,260 of neodymium-praseodymium. The Japanese volumes and the 110-dollar floor are taken from a compilation of company disclosures. If the contract does not say that, the floor leaves the page. Gadolinium in early fiscal 2028 is the company's timing in that quarterly. A slip is not a scandal. It is the date moving, which is the thing you were supposed to watch.
The reading is wrong if Aclara's October 13, 2026 notice did not put Class 3 capex near 494 million dollars, or if the design rates are not about 200 tonnes of dysprosium oxide, 35 of terbium, 1,350 of yttrium, 180 of gadolinium, 190 of samarium, and 1,300 of neodymium-praseodymium, or if those rates are not conditional on the two mines at full capacity. The 26-dollar operating cost excludes feed. The 28 percent return and the 2.6-year payback are the company's case on an Argus deck. If you treat them as a promise, the reading has failed even if every digit is right. Recoveries are not a single band from 96 to 99 percent. Yttrium is about 85.7 percent and gadolinium about 86.7. Purity is 99.5 percent for dysprosium, samarium, and neodymium-praseodymium, and 99.9 percent for terbium, yttrium, and gadolinium. The Export-Import Bank letter of September 4, 2026, is a non-binding interest in up to 750 million dollars. It is not a closed loan. Louisiana's estimated 46.4 million dollar package is not a closed cheque either. If a later filing converts either one into signed money, the stage changes. Until then, the return case is still a spreadsheet. The absence of a definitive offtake is deliberate. A later contract would change the stage. It would not, by itself, build the plant.
The reading is wrong if you use it as a list of rare earth stocks to buy because a deadline is near. A deadline can pass, get extended, or snap back, and the plants will still be where they are. Neo's first magnets do not become a bad product if Beijing loosens. Aclara's drawing does not become a plant if Beijing tightens. The politics change the price of the gap. They do not fill the gap. Filling the gap is tonnes, and tonnes have a quarterly.
The idea, once
China mines most rare earth elements and, after the mine, separates and magnetizes far more than it mines. One licence list, on seven metals including dysprosium and terbium, has been in force since April 2025. A wider package, which also added elements including holmium plus equipment and technology rules, and which includes a rule that could reach some foreign-made parts, is either about to hit a written deadline on November 10 or is being talked about, in a trade-truce sentence that does not name those notices, as if it ran to January 10. Export tonnes are down this year. Export value, on a market note, is up. That is the alert. It is a true alert. It is not a magnet, and it is not a mine plan.
The three names are three ways to answer it. Neo has started to ship permanent magnets from Estonia and still owes the market a tonne count. Lynas already separates oxides at a real scale, and its heavy oxides are still a thin stream next to the licence they are supposed to ease. Aclara has a 494 million dollar Class 3 drawing of the heavy split, a demonstration line with a test due this quarter, a non-binding export-bank letter that is not a loan, no signed buyer, and no financed mine. Watch the date each of them printed. Do not watch the adjective. Could any of this benefit a holder? Only if the next document matches the product, the plant, and the calendar, and only if that holder can say which of the three stages they actually own. The alert already happened. It was not the plant.