Duncan Wanblad said the interest is deep and the deals are unsigned. Amazon’s one physical copper contract is a two-year Nuton pact, and it is small. The BHP item is a certificate, not a shipment. Apple’s rare earth page is a 2025 purchase pledge with MP Materials. A knock is not a signature.
This is not a recommendation to buy or sell any rare earth stock, any miner, any magnet maker, or any critical minerals stock. It is a reading of what technology companies have said, what they have signed, and what five rare earth mining companies still have to prove. Interest can fade. A pilot can stay a pilot. A purchase commitment can be delayed. You can lose money if a tech giant does take a stake, and you can lose money if it never does.
Here is the only idea. A knock is not a signature. On October 15, 2026, the chief executive of Anglo American said technology companies have shown deep interest in buying equity in miners, or in signing long offtake, so they can secure metal for data centres. No stake has been taken. The metal in that conversation is the metal data centres eat in bulk, above all copper. Rare earth stocks are a different aisle. A rare earth investment opportunity appears when a named buyer signs for separated oxide or finished magnets, at a price and a date you can read. A speech about scarcity does not do that work. The five names below are worth a look only as a way to sort the signature from the knock.
What Wanblad actually said
Duncan Wanblad spoke at the Financial Times Metals and Mining Summit in London. He said the discussions are absolutely being had. The question from the technology side is whether to lock in long-term offtake or to take an equity stake in future supply. He called the interest deep. He also said no stakes have been taken. “Given the scarcity issue that we’ve been speaking about, a lot of people are looking for ways to secure supply.” That is the whole news. It is a sentence about a conversation.
The companies were not named in the public account. “Tech giants” is a headline compression of “technology companies.” Compression is how a copper talk becomes a rare earth rally. If the buyer will not put its name on the page, you do not get to write the name in yourself. Microsoft has been mentioned, in other reporting, among firms looking at direct access to raw materials as data-centre needs rise. Mentioned is not the same as signed. A critical minerals supply chain is a set of contracts. A list of interested parties is a rumour with a better suit.
Anglo is not a rare earth company. It is a diversified miner leaning harder into copper, including through a proposed merger with Teck that the parties still hope to close by March 2027, pending Chinese clearance. The scarcity Wanblad was describing is the scarcity of the metals that go into power lines, motors, and the shells of server halls. Treating that sentence as a bid for every rare earth mining stock on a Canadian screen is a category error. The error is profitable for a day. It is not a method.
How tech companies are actually securing metal
How tech companies are securing critical mineral supplies, so far, is narrower than the headline. One contract moves metal. One pilot moves a credit. Neither moves a rare earth.
In January 2026, Rio Tinto signed a two-year supply pact with Amazon Web Services. The copper is cathode from the restarted Johnson Camp mine in Arizona, made with Rio’s Nuton bioleaching process, which uses bacteria and acid on ore that had not paid. The cathode is routed to companies that make parts for Amazon’s U.S. data centres. Rio has guided about 14,000 metric tons from that project over four years. The Wall Street Journal noted that this would not fill one of the largest halls. Amazon is also giving Rio cloud tools to help the process. That is a real signature. It is a copper signature. It is a sliver. It does not reserve a rare earth oxide, and it does not make Rio the owner of the mine.
In September 2026, BHP and Amazon announced a different kind of pilot, and the difference matters. It covers Environmental Attribute Certificates tied to concentrate and cathode at Escondida in Chile. Escondida runs on renewable power and desalinated water. Amazon can retire the certificates against copper it uses in power gear, cooling, and computing parts. BHP was clear that this is in-setting, not off-setting: the credit stays inside the copper chain. It is also book-and-claim. The certificate can move without the same tonne of Escondida metal moving into an Amazon device. BHP’s chief financial officer, Vandita Pant, speaking at the same London summit, said the company is in talks with hyperscalers and is working with Amazon on lower-carbon copper, and that equity is a different issue. A credit is not a shipment. A shipment is not a seat on a rare earth board.
BHP has put a number on the hunger, and the number is copper. The company has said that every extra $200 billion of annual data-centre investment needs about as much copper as a new mine producing 150,000 tonnes a year. In its last full year, copper was more than half of underlying EBITDA, on production of about 2 million tonnes. EBITDA is not profit. Those figures are still a copper budget. They are not a dysprosium budget. Data centres do use rare earth magnets, in motors, drives, and cooling gear. The tonnage that moves this miner’s result, on BHP’s own telling, is the red metal.
McKinsey’s 2026 materials work put data centres at about half a percent of mining-industry revenue in 2025, on the order of $12 billion, with a larger share expected later. The same outlook sketched global data-centre power capacity rising from 84 gigawatts in 2025 to nearly 600 gigawatts by 2040. An EXANTE note the same week quoted a Goldman Sachs figure of about $1 trillion in AI-related investment in 2026, including a large U.S. share, and argued that the boom is being priced as software while it runs on copper, power, and water. Read that as a warning about the bill, not as a shopping list that starts with a junior rare earth deposit.
This is the pattern. Where a technology company has taken metal, it has taken a small amount of copper, for two years, from a process trial in Arizona. Where it has taken a pilot with BHP, it has taken a carbon claim, not a cargo. Where it has only talked, the talk is equity or offtake in a large house, and the talk is unsigned. Critical minerals investment, in the form a data-centre builder will actually write, shows up as a named metal, a named mine or process, and a term. Everything else is the knock.
The rare earth signature is older, and it is not this knock
There is one place a technology company did put a number on rare earth magnets. It is not this week’s Anglo meeting.
In July 2025, Apple announced a $500 million commitment to buy magnets from MP Materials, tied to recycled content, and the two companies described a California recycling effort to pull rare earths out of used material. Shipments under that commitment have been described as a later event, not as a wire that landed on October 15, 2026. Volumes and the exact length of the deal were not laid out as a simple mine contract. It is still the cleanest public example of a tech buyer stepping onto a rare earth page. It is a purchase commitment. It is not an equity stake in the mine. It does not spread, by osmosis, to every other rare earth stock.
The larger signature on that same company is not a tech giant at all. On July 10, 2025, MP announced a package with the U.S. Department of Defense. The department agreed to buy $400 million of a new convertible preferred stock and received a warrant. Together, on an as-converted and as-exercised basis, that was about 15 percent of the common stock at the time, and it was built to make the department the largest shareholder. Largest is not control. Fifteen percent does not run the company. The package set a ten-year price floor of $110 a kilogram on neodymium-praseodymium products. For ten years after the 10X plant is built, the department agreed to ensure that 100 percent of that plant’s magnets are bought by defense or commercial customers. A $150 million loan was for heavy rare earth separation at Mountain Pass. By 2026 the company was calling that department the Department of War. The 2025 release said Defense. Use the name on the contract you are reading. Do not pretend the two labels are two different cheques.
General Motors, an automaker, has been the foundational magnet customer since a December 2021 agreement. Trade press has described small commercial shipments on that domestic chain as starting in the fourth quarter of 2026. Treat that date as a reported schedule, not as a clause you have read in the GM contract. The state and the car industry signed. A phone company signed a purchase pledge. The data-centre giants, in the news of this week, did not.
That split is the whole lesson for rare earth stocks. The buyers who have moved are the ones who cannot build a missile, a motor, or a flagship device without a magnet. The buyers who are knocking are the ones who cannot build a server hall without power and copper. Both can be true. They are not the same trade.
Why rare earths are the wrong metal for this headline
A data centre is a copper, power, and water problem that happens to contain some magnets. The magnets matter. They are not why Anglo’s phone is ringing.
Rare earth mining companies sell a mixed product that is useless until it is cracked, separated, reduced to metal, alloyed, and pressed into a magnet. China still does most of that middle and most of the last step. Public estimates of the share vary with the year and the stage. One October 2026 account put China’s hold above 90 percent for rare earths, in a list that also cited very high shares of processed gallium, graphite, cobalt, and tungsten. The sentence did not isolate “processed rare earths” as its own audited figure. You do not need the precise percent to see the shape. The scarce step is not “a tech giant might buy equity.” The scarce step is a plant that splits the elements and a customer who takes the magnet when the Chinese price tries to undercut it.
Equity in a miner does not, by itself, give a technology company separated dysprosium. It gives a claim on whatever the miner produces, in proportion to the stake, after the other shareholders, the state, and the offtake that was signed first. A small stake in a copper house does not move a magnet line. A 100 percent offtake of a junior’s future concentrate does not move it either, if the concentrate still has to be separated in the country the buyer was trying to avoid. Critical minerals stocks that blur these steps will use this week’s headline. The steps remain.
There is a second blur. “Critical minerals” is a policy basket. It holds copper, nickel, lithium, cobalt, graphite, gallium, and the rare earths. A basket is a convenient word for a summit. It is a bad word for a portfolio. Which critical minerals investment opportunities exist in North America depends on which mineral you mean. For copper, North America has mines, smelters, and a tariff politics of its own, and the one tech signature that moves metal already points at Arizona. For rare earths, North America has one operating integrated name, a handful of separation plans, and a long list of deposits. The opportunities are not interchangeable. An investor who buys the basket because Wanblad said scarcity will own a story. An investor who buys the mineral the contract names will own a position.
MP Materials: the signature, and its limits
MP Materials, NYSE: MP, is the name the tech headline will be stapled to, and the staple is only partly fair. Mountain Pass in California is an operating rare earth mine. The company has been separating neodymium-praseodymium oxide there, not only shipping concentrate. It reported 2,599 metric tons of that oxide in 2025, more than double the year before. In the first quarter of 2026 it reported 917 tons. In the second quarter, a maintenance quarter, it reported 840 tons. The half is 1,757 tons. That is about two-thirds of the prior full year, not almost all of it. The ramp is real. The adjective “almost” is not.
The downstream piece is the one Apple and the Pentagon are actually paying for. Magnets are made, and will be made, in Texas. Independence, in Fort Worth, is the earlier plant. Northlake is the larger one, the 10X facility, with operations aimed at 2028. Published descriptions of the Northlake spend run above $1.25 billion, and some accounts say about $1.3 billion. A commitment letter from JPMorgan and Goldman Sachs, at the time of the 2025 package, was for $1 billion toward building it, subject to the usual conditions. In October 2026 the company launched Project Swarm: a way for drone and motor makers to reserve magnet capacity, described by the chief executive as up to 2 million drones a year, without first signing a binding purchase. Reservations that are not purchases are a way to gather demand. They are not revenue. The plant they point at is a 2028 plant.
Watch three things, and ignore the Anglo noise. First, does the Apple commitment turn into shipments of recycled-content magnets, or does it stay a press release with a large number. Second, does the price floor and the 10X backstop survive a change in administration, a change in price, or a change in what “ensure the output is bought” means in practice. A floor of $110 a kilogram is a contract with a state, and states rewrite contracts more politely than they cancel them, but they do rewrite them. Third, does NdPr oxide keep rising without the magnet plants becoming a warehouse for powder no buyer has called for yet. The reported GM start in late 2026 is the public test of that third point. A tech giant eyeing Anglo does not change any of the three.
The limit is concentration. One mine. One country. A state holder that is the largest, not the boss. One big car customer. One phone commitment. That is a supply chain. It is also a short list of ways to be disappointed, with a New York listing on top. Rare earth mining stocks do not get safer because the government bought preferred stock. They get more political. Politics can be a floor. Politics can also be the reason a multiple never matches the tonne.
Lynas: the plant that does not need this knock
Lynas Rare Earths is the other operating fact outside China. The mine is Mt Weld in Western Australia. The cracking plant was built nearer that ore, at Kalgoorlie. Separation has run for years in Malaysia. The company has also been pulled into U.S. heavy-rare-earth plans. It sells oxide because it already separates, not because a data-centre landlord called a London summit. The tonnes are Mt Weld’s. Kalgoorlie cracks them. Do not move either number onto a Toronto story.
That is why Lynas belongs on a watch list for this headline, and why the headline does not belong on it. If technology companies decide they want non-Chinese neodymium-praseodymium, Lynas is one of the few places that can ship it without first building a plant. If they decide they want equity in a copper miner, Lynas is irrelevant to the decision. The rare earth market will try to trade it as if those were the same decision. They are not.
What to watch is feedstock, permits, and the Malaysian plant’s political weather, not a Silicon Valley stake that has not been offered. A processor with a real circuit is exposed to the oxide price, to the host government’s patience, and to the cost of more cracking capacity. It is not exposed, in any direct way, to whether Amazon’s Escondida certificates become a ten-year metal contract. Canadian rare earth stocks sometimes move with Lynas because the screen is one colour. The assets are not one asset.
Neo Performance Materials: the processor that already changed shape
Neo Performance Materials, TSX: NEO, is the Canadian-listed company that already does the step the juniors advertise. It separates, makes alloys, and makes magnet materials. Silmet, in Sillamäe, Estonia, is one of the few commercial separation plants in the West. The company is not a moose pasture with a rare earth label. It is also not the company it was when every slide still showed a Chinese separation circuit.
In May 2025 Neo said it had sold its majority stakes in two Chinese separation plants. It still lists manufacturing in China, Estonia, Thailand, Canada, Germany, and the United Kingdom, and it is still an integrated magnet-materials business. Separation in China is the piece it left. On October 13, 2026, the European Commission named Silmet 2.0, the plan to modernize the Estonian plant, a strategic project under the EU Critical Raw Materials Act. The company says the program is meant to triple capacity, widen heavy rare earth separation, and take new feedstock. The same release said the plant is ending its current feedstock deal. Estonian reporting identified that supply as Russian, from Solikamsk, and described lower output plus temporary shutdowns of the existing light rare earth circuit while the switch is made. A processor that is between feeds is not a quiet compounder. It is a plant with a hole in the schedule.
The rumour of this week still does not fit. Tech companies, on the evidence, are trying to secure copper, or a carbon claim on copper, for halls they are building now. Neo sells specialty materials into motors, autos, and electronics on a book of customers it already has. A new data-centre motor program could, over time, show up as magnet demand. It will not show up because Wanblad said the talks are deep. If you own Neo as a critical minerals investment, judge it on whether Silmet keeps running through the feedstock change, on whether 2.0 is funded, and on the price it gets versus the Chinese quote. Do not judge it on an unsigned equity conversation at a miner that does not make magnets.
Neo is also the cleanest answer to the North American question that is still honest. The question was which critical minerals investment opportunities sit in North America. Neo’s listing is Canadian. Its headquarters are in Toronto. The separation plant this news actually concerns is in Estonia. A listing is not a mine, and a mine in Estonia is not a corridor in Nevada. North American rare earth opportunity, if the phrase means rock in the ground plus a plant on the same continent, is a shorter list than the phrase suggests. Neo is a way to own a real circuit without pretending the circuit is in Ontario.
Aclara: a finished study, an unfinished cheque
Aclara Resources, TSX: ARA, is the shape the headline will be abused on. The plants that would feed a U.S. separator are ionic-clay projects: Penco in Chile, which has an environmental approval, and Carina in Brazil, the one the company calls its flagship. Peru is still in the portfolio as concessions to explore. It is not a third mine shipping carbonate to Louisiana. Treating three countries as three projects is how a land package becomes a reserve in a paragraph.
The Louisiana piece is Project Dynamo, at the Port of Vinton. On October 6, 2026, the company said it is widening the plant. Neodymium-praseodymium, dysprosium, and terbium were already in the design. Yttrium, gadolinium, and samarium were added. Mixer-settlers go from about 340 to about 478. A state air permit is in hand. On October 6, basic engineering was in its final stage. By October 13, Hatch had finished it. The Class 3 construction figure is about $494 million, roughly 78 percent above a $277 million early estimate. Site preparation is aimed at the first quarter of 2027, subject to financing and other conditions. Construction completion is aimed at the second half of 2028. The U.S. Export-Import Bank has issued a nonbinding letter of interest for up to $750 million. Nonbinding means the letter can sit in a drawer. The company has also said it is in early customer talks and has no finished offtake.
Subject to financing is still the sentence. No technology company has signed the financing. No October 15 conversation at the FT summit named Aclara. A data-centre builder that wanted magnets in 2028 might someday want a Louisiana circuit. Wanting is the knock. Aclara’s job is to turn the knock into a term sheet before it turns dirt. Until that page exists, the stock is a financing story wearing a tech costume. The engineering being done does not retire the condition. It prices it. A study that comes in far above the old sketch is a reason to read the new number, not a reason to celebrate that the study ended.
The clays are the other half, and they are not in North America. Ionic clay is a plausible heavy-rare-earth feed, which is the feed magnet makers actually run short of. It is also a permitting and community story. Penco’s environmental approval is a step, not a pour. A Louisiana plant with no feed is a building. Feed with no plant is a bag of dirt. The company has to build both, in order, with money it has not finished raising. That is a real rare earth investment opportunity or a real way to lose money. It is not a derivative of a copper certificate in Chile.
Ucore: a machine with a company clock
Ucore Rare Metals, TSXV: UCU, sells a separation method, RapidSX, and a plan to run it. The Strategic Metals Complex is in Alexandria, Louisiana, at England Airpark. Ground was broken in May 2025. Bokan-Dotson Ridge in Alaska is the heavy rare earth deposit that sits beside the story and is not the same thing as the plant. An engineering update cut the plan from four production lines to three and kept a first small unit, Machine A, in front of them. The company has said installation, testing, and commissioning of Machine A are targeted for the first half of 2027. It has not given firm dates for the later lines. One market note in October 2026 talked about late-2027 commercial output and revenue in 2028, and about U.S. government and Sumitomo support. That note is a commentator. It is not the company’s clock. Use the clock the company printed.
That capital family still matters, and it is not Amazon’s. Public support and an industrial partner are the kind of money that has actually moved into non-Chinese separation. A technology company looking at equity in Anglo is a different family, shopping for a different metal. If Ucore works, it will work because Machine A separates a real feedstock at a recovery a customer accepts, on the date the company already put in the release. A summit and a deep expression of interest do not commission a line. A Seeking Alpha sentence does not either.
Watch the feedstock contract and the first-half 2027 gate. Do not watch the share price on the morning the Anglo story is recycled. A pre-revenue separation story can double on a headline and still be early. Early and right is a fine outcome for a patient holder. Early and funded by a misread headline is how Canadian rare earth stocks give the gain back.
The five, sorted by distance from a signature
Put them in a line, from closest to a named buyer to furthest.
MP Materials has the signatures: a government floor, a government preferred share, a carmaker, and a $500 million Apple commitment. The limit is that the big magnet tonnes are a 2028 plant, and the Apple page is a purchase pledge, not a data-centre offtake signed this week. The state holder is the largest. It is not in charge.
Lynas has the tonnes, at Mt Weld, cracked at Kalgoorlie and separated in Malaysia. It does not have this week’s tech meeting. It does not need it in order to be a producer. It can still disappoint on price, on Malaysia, or on the pace of heavy rare earth separation.
Neo has a real circuit in Estonia, a Canadian listing, and a feedstock change underway. It left Chinese separation in 2025. It does not have a new tech equity story. Judge the plant through the shutdowns. Ignore the costume.
Aclara has a finished basic engineering package, a $494 million figure, a Q1 2027 site date, and a financing condition. No tech name is on the condition. Peru is not the third mine.
Ucore has a method, a Louisiana site where work has started, an Alaskan deposit, and a company gate in the first half of 2027 for Machine A. The gate is not a customer. A commentator’s 2028 revenue line is not the gate.
None of the five is Anglo. None of the five is the Nuton contract. None of the five is an Escondida certificate. None of the five got a phone call in the story Wanblad told. If a rare earth stock moves because that story ran, the move is the third kind of trade: the multiple. Multiples expire. Signatures do not, until they are broken in writing.
What a real tech deal would look like
If a technology company does cross from copper into rare earths, the page will not look like a summit quote. It will look like one of four instruments.
A purchase order for magnets, with a volume, a price or a collar, and a first shipment date. Apple’s 2025 commitment is the sketch. A sketch without a volume is still a sketch. Ask for the tonnes.
An offtake for oxide, longer than a pilot, from a plant that already separates or has financed the circuit. A two-year copper trial is not the template. A ten-year magnet backstop is closer, and that one was signed by a defense department, not by a cloud company.
Equity that is closed, with a percent, a board seat or a contractual supply right, and a use of proceeds. Wanblad was explicit that this has not happened. Pant said equity is a different issue from a lower-carbon copper conversation. When a stake happens, it will be a filing, not a colour story. Until the filing, “deep expressions of interest” are a mood.
A recycling venture, because technology companies have scrap and rare earth plants need feed that is not a new pit. Apple’s recycling language with MP is the hint. A hint is not a second Mountain Pass. Urban mining is slow. Collection rates are the bottleneck, not the press release.
Any of those four can be a rare earth investment opportunity. All of them can also be a bad price. A signature tells you the customer exists. It does not tell you the equity is cheap. Pay for the gap between the signed price and the all-in cost of the plant. Do not pay for the logo of the customer.
North America, without the brochure
Which critical minerals investment opportunities in North America are actually in front of you this month?
For the metal the tech companies are securing, look at copper that can actually arrive. The Rio contract is the proof of physical demand, and the proof is small: on the order of 14,000 tonnes over four years, against halls that can take tens of thousands of tonnes each. The Escondida pilot is proof that Amazon wants a carbon claim it can retire. It is not proof of a new cargo. Neither item is a rare earth stock. Mixing them into a rare earth basket because both sit under “critical” is how the brochure gets written.
For rare earths, the North American list that is more than a claim is short. Mountain Pass is producing oxide. Texas is producing, or about to produce, a small number of magnets, with a larger plant dated 2028. Louisiana holds two unfinished plants, Aclara’s at Vinton and Ucore’s at Alexandria. Alaska has a deposit that is not a plant. Canada has a processor listed in Toronto whose separation news is in Estonia, plus a long file of juniors whose economic studies assume a separation step they do not own. That file can still make money. It does not make money because a hyperscaler might want copper.
The honest North American screen has three columns. Column one: who separates today. Column two: who has a named buyer for magnets or oxide. Column three: who has cash or a committed loan that covers the next plant, not the next drill season. MP sits in all three, with the political risk that comes from a state shareholder who is large and not in control. Lynas sits in column one and part of column two, outside the continent. Neo sits in column one, off the continent, with a Canadian ticker and a feed switch in progress. Aclara and Ucore are trying to enter column one and do not yet sit in column two. A junior with none of the columns is a mineral right. It is allowed to be a mineral right. It is not allowed to borrow Apple’s $500 million.
How the headline will be misused
The misuse has a script. A newsletter will put “tech giants” in the first line and a rare earth ticker in the third. The middle line will be Wanblad’s word, scarcity. The reader will feel late. Late is the product.
Check the script against five questions. Did the technology company get named? Did the metal get named? Did anyone sign? Is the signed thing copper metal, or only a certificate about copper? Is the rare earth company you are being sold a party to any of it? If the answers are no, copper, no, and no, you are being sold the knock. Put the article down. The rare earth thesis, if you want one, has to stand on oxide, magnets, and a customer of its own.
There is a softer misuse, and it sounds smarter. It says data centres need everything, so a bid for copper is a bid for the whole critical list, so rare earth mining stocks should re-rate with the copper houses. The word everything is doing too much work. A server hall’s copper bill is measured in thousands of tonnes. Its rare earth bill is measured in a way that does not move BHP’s EBITDA. A small, non-substitutable input can still be strategic. Strategic is not the same size as the trade the summit was about. Size the position to the input, not to the adjective.
What would make this reading wrong
The reading is wrong if a named technology company, in the next filing season, buys equity in a rare earth miner or signs a magnet offtake that this week’s coverage simply did not have. Wanblad said the interest is deep and the stakes are unclosed. Deep can become a contract. If the contract is in copper, this reading stands. If the contract is in NdPr or finished magnets, the rare earth aisle just got a new buyer and the five names have to be re-sorted by who can deliver into that contract. Re-sort then. Do not re-sort on the rumour.
The reading is wrong if the Escondida pilot quietly becomes a physical offtake and the coverage you read still calls it a certificate. Instruments change. A retired credit is not a cargo. A cargo would be news. Until BHP or Amazon says the tonnes are theirs to take, do not upgrade the pilot to keep a prettier paragraph.
The reading is wrong if Project Swarm, or the Apple commitment, fills the 10X plant with real purchase orders before 2028. In that case MP is not only a 2028 story. It is a demand story happening now, and the limit section above is too harsh. The test is purchase orders, not reserved capacity. Reserved capacity was designed so the buyer would not have to commit. Do not treat the design as the opposite of itself.
The reading is wrong if Aclara or Ucore announces a technology company as financier or offtaker. Either announcement would move that name from the schedule column into the signature column. A nonbinding export-credit letter does not count. A government grant does not count. A grant is public capital. This piece is about private technology capital. Keep the two cheques in two envelopes.
The reading is wrong if the copper pact stays small, the equity talks die, and rare earth prices rise for their own reasons: a Chinese export rule, a defense order, a magnet shortage. Then rare earth stocks can go up while this entire tech story goes nowhere. That outcome does not make the tech story true. It makes it unnecessary. The best rare earth positions will be the ones that did not need the unnecessary story.
The idea, once
Tech giants are eyeing mining investments. Duncan Wanblad said so on October 15, 2026. The interest is in equity or long offtake. No stake has closed. The signature that moves metal is copper, and it is small: Amazon and Rio Tinto’s Nuton cathode from Johnson Camp, a two-year pact, about 14,000 tonnes over four years. The BHP item with Amazon is a certificate on Escondida copper, not a shipment of it. Pant said equity is a different question. BHP’s own arithmetic is still a copper mine for every large new block of data-centre spending, and copper was more than half of underlying EBITDA, not a rare earth line. That is how tech companies are securing critical mineral supplies today. They are securing a little of the metal in the cable. They are accounting for the carbon on a larger flow. They are talking about the rest.
Rare earth stocks are the rest. MP Materials is the exception with a real page: a 2025 Pentagon package, since described by the company under the War Department name, a $110 floor, GM’s magnets, and Apple’s $500 million commitment from July 2025. First-half oxide was 1,757 tons, about two-thirds of 2025, not a match for the full year. Lynas already separates ore from Mt Weld. Neo already separates in Estonia, is changing feed, left Chinese separation in 2025, and the Canadian listing does not move the plant to Canada. Aclara’s basic engineering is done, the bill is about $494 million, and groundbreaking still waits on money. Ucore’s company gate for Machine A is the first half of 2027, not a commentator’s revenue year. Which critical minerals investment opportunities in North America are more than a brochure? The ones in column one, two, or three. A knock on Anglo’s door is none of the columns.
A knock is not a signature. Watch the name on the page. If the name is a copper miner, ask whether the page is a tonne or a credit. If the name is a rare earth company, ask who signed, for which product, and when the first tonne ships. The summit does not answer that. The filing will.
A week is a long time to misread a metal
The next five sessions will try to glue the summit to the rare earth tape. Let them try. The glue fails a simple test. Anglo does not separate neodymium. Amazon’s January deal does not mention a magnet, and it does not fill a hall. BHP’s September pilot does not deliver a pound of Escondida copper to a server. Apple’s number is from the summer of 2025, and it sits on one company. A tape that lifts five rare earth names by the same percent has not read five filings. It has read one headline.
Use the week for inventory, not for prophecy. Write the buyer, the metal, the instrument, and the date next to each of the five. MP: state floor, largest but not controlling holder, carmaker, Apple purchase pledge, Texas magnets dated 2028, half-year oxide at about two-thirds of last year. Lynas: operating separation of Mt Weld ore, no new tech equity. Neo: operating separation in Estonia, Chinese separation sold, feedstock switch and shutdowns in progress, Canadian ticker. Aclara: engineering finished, about $494 million, site work in early 2027 only if the money arrives, Peru not the third mine. Ucore: Machine A targeted for commissioning in the first half of 2027, later lines undated. Four of those lines do not change if Wanblad’s talks die on Monday. One of them, MP’s, does not change unless Apple or the Pentagon changes it. That is the definition of a story you do not need.
Critical minerals stocks will keep arriving in the same envelope as rare earth mining stocks. Open the envelope. Copper, a carbon credit, and rare earth magnets can share a sentence and still not share a cash flow. The hall is the customer’s building. The cash flow is the miner’s. Do not pay the building’s multiple for the miner’s unfinished plant. And do not sell a finished plant because the building’s buyer has gone quiet. The knock can stop. The circuit, if it is real, keeps running.