Brett Lambert said the first tranche could be several hundred million dollars, maybe a billion, for more than a dozen firms he will not name. The cargo that arrived is cathode from a mine a Chinese company operates. Those are not the same reserve.
This is not a recommendation to buy or sell a miner, a trader, a manufacturer, or any fund. It is not a price target. It is not a view that a government warehouse makes a supply chain safe. A loan can be announced and not drawn. A cargo can land and not be the metal a factory asked for. You can lose money on a reserve that exists mostly as a credit line. A credit line is not a pile. A pile of the wrong shape is not a shield.
Here is the only idea. The facility is not the metal. Project Vault is a structure: a loan ceiling at the Export-Import Bank, private capital beside it, and a plan to buy what American manufacturers say they need. This week the structure produced two headlines that do not add. VaultCo’s executive chairman, Brett Lambert, told Bloomberg the firm is likely to announce several hundred million dollars of loans to manufacturers, maybe a billion, in a first group of more than a dozen companies, with another dozen to follow. In the same week Mercuria said it had delivered 12,000 tonnes of copper from the Democratic Republic of Congo into the vault. One number is a loan that has not been named. The other is a cargo that has. If you add them and call the sum a stockpile, you own a press release.
What he said he would announce
Lambert’s words, as carried from an October 14 interview, are careful. “What’s likely to happen is we’ll announce several hundred million dollars, maybe a billion dollars, of loans to OEMs who have requested use of the Vault facility.” Likely. Maybe. Requested. Then: “In the first tranche, we’ll have more than a dozen firms, followed by more than a dozen shortly after.” He said VaultCo is in talks with more than 70 companies on both sides of the chain, the firms that use the metal and the firms that can get it. He said the recipients will not be named at this stage. The announcement, he said, should disclose the types of minerals they asked for.
Read the verbs. Announce. Request. Types. None of those verbs is “delivered,” except for the copper a trader described on a conference stage. A loan to an original equipment manufacturer is a right to finance a purchase. It becomes metal only when someone buys a specific form, stores it, and can hand it over when a normal supplier fails. Lambert himself put the test in plain language, as the trade press rendered it. The first loans have to turn highly specific requests into inventories you can rely on. Processed materials are harder to source than raw commodities. That sentence is the whole risk. It is also an admission that the risk has not been retired.
He added a line that should sit on the front of the model. “There isn’t any one mineral that stands out, but the specificity of the requests have surprised me.” No hero metal. A pile of particular asks. Firms want raw metal, and they also want magnets, powders, alloys, and other processed forms cut to a production line. A magnet is not a drum of concentrate. A powder has a particle size, a coating, a certification. An alloy has a recipe. The surprise, if you believe him, is that the requests were that fussy. A fussy request is the point of a demand-driven vault. It is also the part a trader can fail. Copper cathode is a standard. A magnet grade is a argument.
A ceiling is not a warehouse
The big number is older, and it is a ceiling. On February 2, 2026, the Export-Import Bank’s board approved a direct loan of up to $10 billion to Project Vault. The White House event that day called it a strategic critical-minerals reserve, an independently governed public-private partnership that would store essential raw materials at sites across the United States. EXIM’s chairman, John Jovanovic, later called the deal a landmark and said it was expected to produce a positive return for taxpayers. Expected is not received. Up to $10 billion is not $10 billion drawn, and it is not $10 billion of rock.
The design beside that loan is about $2 billion of private capital. In September, Glencore and Mercuria each committed $500 million, the first commodity traders to join on those terms. One billion from two houses is half of the private figure in the original sketch, not the whole sketch, and not a pile you can walk through. Reports of the September step also say the traders are in line to use EXIM financing to buy the material. A commitment, a loan, and a delivery are three different objects. Mercuria’s $500 million commitment in September and Mercuria’s 12,000 tonnes in October are allowed to be related. They are not the same fact. Do not multiply $500 million by a copper price and pretend the product is the vault.
EXIM’s February list of early manufacturer interest is also not this week’s loan list. The bank said initial indications of participation included Clarios, GE Vernova, Western Digital, Boeing, and others. Suppliers named then included Hartree Partners, Mercuria Americas, and Traxys. “Initial indications” is a raised hand. Lambert now says the firms that get the first loans will not be identified. A raised hand in February is not a drawdown in October. GE Vernova’s chief executive welcomed access to minerals for power equipment. Boeing’s government lead welcomed supply-chain resilience. Those are statements of support. They are not receipts for neodymium, cobalt, or tungsten sitting in a cage with their name on it.
There is a second stockpile, and it is not this one. The National Defense Stockpile is run by the Defense Logistics Agency for materials the military must have. Project Vault is aimed at manufacturers, and it is supposed to buy what those manufacturers specify rather than a list written in advance by a planner. The distinction matters. A defense pile can sit for years and still have done its job if the metal is there on the day a war starts. An industrial pile that bought the wrong magnet has failed on the day the line stops, even if the building is full. Lambert’s surprise at how specific the asks were is the sound of that difference.
The metal you can count
On October 15, at the FT Metals and Mining Summit in London, Mercuria’s global head of metals, Kostas Bintas, said copper from the Congo had gone into the vault that week. The amount reported is 12,000 metric tons. The price and the exact dock were not disclosed. Twelve thousand tons is a real shipment. It is also a small one next to the slogans. At a few thousand dollars a ton, the cargo is tens of millions of dollars of metal, not a billion, and not ten billion. The facility and the pile are now visible in the same week, and they are different sizes by an order of magnitude or two.
Follow the rock back. In January, Congo’s state miner, Gécamines, said it would ship 100,000 tons of copper in 2026 from its 20 percent share of production at Tenke Fungurume, with Mercuria providing finance and logistics. Tenke is operated by CMOC, the Chinese miner that owns the other 80 percent. The 12,000 tons Bintas described are part of that wider Congolese program, not a new American mine. If the January figure still holds, this week’s cargo is a slice of a state share of a Chinese-run pit. The vault’s first public copper did not come from a deposit Washington controls. It came from a trading chain that starts in a mine Beijing’s commercial world already sits inside.
That is not a scandal by itself. A state share is a real title. Metal from it can sit on American soil and be useful in a crunch. It is a correction to the slogan. “Cut China reliance” cannot mean “the first cathode came from a joint venture a Chinese firm operates” and also mean the reliance is over. It can mean the title to that slice is Congolese, the trader is Swiss, and the warehouse is American. Those are three jurisdictions. None of them is a processing plant that turns the cathode into the powder a buyer actually requested.
Copper is also the easy test. It is liquid. It is weighed. It has a price every day. Lambert said no single mineral stood out in the requests, and that the specificity surprised him. A surprise made of magnets and alloys is not answered by the most standard industrial metal on earth. The copper proves the door opens. It does not prove the vault can hold the thing the line will stop for. If the next delivery is another cathode parcel, the facility is a copper book with a critical-minerals letterhead. If the next delivery is a named magnet grade, with a cert and a customer, the letterhead starts to match the shelf.
The other Congo copper is not the vault
Do not mix this cargo with the July trade data. The United States imported a record 53,290 tonnes of refined copper from the Congo in July 2026. That one month was more than the country imported from Congo in all of 2024, a year that came in under 32,000 tonnes. Congo’s share of American copper imports that month was about 23.9 percent. Total U.S. copper imports in July cleared 220,000 tonnes for the first time. Traders were pulling metal forward in front of a possible tariff. A Commerce Department process had recommended a 15 percent duty on refined copper from January 2027 and 30 percent from 2028. As of late summer, the decision was still open. Congolese brands are not, in the main, approved for delivery against the U.S. futures contract. Factories can still buy them. The futures market cannot treat them as a substitute for a listed brand.
That July metal met ordinary demand, or a rush to beat a tax. It was not Project Vault. The people writing up this week’s news have been careful on this point, and then they have put the two facts in adjacent sentences. Adjacent is how a reader adds them. Fifty-three thousand tonnes in July plus 12,000 tonnes in October is not 65,000 tonnes in the vault. One number is a customs total. The other is a trader saying a parcel entered a program. A country can import a record and still have an empty reserve. A reserve can hold 12,000 tonnes and still be a rounding error next to a month of imports.
The July rush also shows why a stockpile and a tariff fight are different clocks. If buyers are already stocking copper because a duty might arrive in 2027, they do not need a government vault to do it. They need a warehouse and a price. The vault’s job, in its own description, starts when normal channels break. A tariff is not a broken channel. It is a tax. Metal that races a tax is commercial inventory. Metal that sits for a disruption is a reserve. Paying for both and calling them one policy is how a strategic reserve becomes a second buyer in a tight market, bidding against the factories it claims to protect.
What a demand-driven pile actually requires
The design, as EXIM and VaultCo have described it, flips the old stockpile. A planner does not fill the shelf with a standard list. A manufacturer names the material, agrees to buy it at a set price, and the vault uses financing to procure and store it until the manufacturer draws it. In a calm month the manufacturer is the customer of its own reserve. In a crisis the reserve is supposed to be the supplier of last resort for that same specification. That can work. It can also fail in ways a classic pile does not.
The price rule is the first crack. Fastmarkets, writing when the two traders’ commitments became public, said market sources could not see a clear pricing or price-support mechanism. If the manufacturer has agreed a price, someone has to wear the gap when the market moves. If the vault wears it, the taxpayer’s “positive return” is a bet on that gap. If the manufacturer wears it, the loan is just forced inventory, and the firm may not draw it. If nobody has published the rule, the billion-dollar announcement is a credit decision without a market. You cannot underwrite a reserve whose price is a private contract you have not read.
The form of the metal is the second crack. Raw copper can be bought this week from a trader who already has a Congolese book. A sintered magnet, a rare-earth powder, or a superalloy billet cannot. China dominates the processing steps that turn a lot of those ores into the object a motor plant or a turbine plant actually bolts on. A loan that can only be spent on what traders already move will fill the vault with the liquid things. The specific things Lambert said surprised him will remain requests. The test he named is exactly this: can the first loans become inventories of the processed forms, not just more cathode.
The third crack is secrecy. Not naming the borrowers protects commercial plans. It also means a shareholder in Boeing, or GE Vernova, or any other firm that raised a hand in February cannot tell whether that firm is in the first tranche. Support is not a draw. A draw that is secret is not an asset you can put in a model. Until the types of minerals are published, the only public inventory is the copper Bintas described. Trade that. Do not trade the other seventy conversations.
Who this is actually for
Sort the people who might think they gained something this week.
A manufacturer that gets a loan gains time, if the loan is real and the metal matches the line. It does not gain a mine. It gains a financed right to a stored input. If the input is a standard cathode, it could have bought that input itself, as the July import numbers show. If the input is a magnet grade that is scarce, the loan is interesting, and it is also the part not yet demonstrated.
A trader that commits capital and delivers metal gains a customer with a government loan behind it. Glencore and Mercuria are doing the job traders do: finding metal, moving it, getting paid. The Congo copper is their kind of job. A bespoke alloy may not be. Do not assume the same desk can do both just because both sit under one program name.
A miner in the United States, or in a friendly country, gains a possible buyer only if the vault’s shopping list reaches its product. A demand-driven reserve buys what factories request. Factories often request a processed form, not a concentrate from a new pit. A junior with a drill program is not, on these facts, a supplier to the vault. A producer of a qualified cathode, powder, or magnet might be. The press release does not promote the drill.
A taxpayer gains a claim, not a dividend. EXIM says the structure should return more than it costs. That claim depends on the loans being repaid, the metal being sold at the agreed prices, and the gap between those prices and the market not eating the fee. None of that has a public track record, because the reserve is new. A positive return is a design. The design is eight months old. The first cargo you can point at is from this week.
A reader who wanted a strategic copper reserve already has two stories and must not weld them. One is commercial: America bought a record amount of Congolese cathode in July because it was available, cheaper than some listed brands, and a tariff might be coming. The other is institutional: a program with a $10 billion ceiling took in 12,000 tonnes in October. The first story is a market. The second is a building. The building does not explain the market. The market does not fill the building.
What you can check
The next announcement has a short list of facts that would change the weight of this week. How many dollars were actually committed, not “up to” or “maybe.” How many firms, and whether “more than a dozen” became a number. Which forms of material, in words a plant manager would use: cathode, oxide, metal powder, sintered magnet, alloy billet. A list that says “critical minerals” is the same fog as the headline. A list that says a magnet grade, a rare-earth oxide, or a copper shape is a shopping list you can fail.
Then ask where the material sits, and who can take it out. EXIM said the reserve would use facilities across the United States. A facility across the country is not a pile in one room. If the copper is in a commercial warehouse under a trader’s control, the vault is an account. If it is in a site the program controls, with a rule for who draws it in a crunch, the vault is a reserve. Accounts get rehypothecated. Reserves are supposed to stay. The public record this week does not say which one the 12,000 tonnes are. Until it does, do not picture a cavern.
Ask what the manufacturer gave up. In the design described so far, the firm names the material and agrees to buy it. That agreement is the asset that makes a loan look safe. It is also a obligation on the firm’s own cash. A company that must buy its own stockpile, with government money in the middle, has not been given metal. It has been given terms. If those terms are better than a private warehouse, say how. If they are not published, the benefit is a slogan.
Ask whether the processed request survived contact with the trader. Lambert said the specificity surprised him. Traders are built for parcels the market already understands. A 12,000-tonne copper lot is that parcel. A powder with a certification for a named motor is not. If the first loans fund more copper, more cobalt metal, or more bulk nickel, the program bought what was offered. If they fund a magnet or an alloy that had no deep market, the program bought what was asked. The difference is the difference between a stockpile and a convenience.
Keep the Congo book separate while you watch. Gécamines talked about 100,000 tonnes in 2026 from its Tenke share. Twelve thousand have been tied, by one executive on one stage, to the vault. The rest can go to ordinary U.S. buyers, to China, or nowhere. A later parcel into the vault would mean the door is a habit. A later parcel into the commercial import tables would mean July’s record had a sequel. Do not give the vault credit for tons that clear customs for someone else.
What would make this reading wrong
The reading is wrong if the loans are announced, drawn, and matched within a quarter by inventories of the specific magnets, powders, and alloys Lambert said the firms requested. Then the facility has started to become the metal. You will see it in a list of forms, not in a round number of dollars. Wait for the forms.
The reading is wrong if the 12,000 tonnes are the start of a repeated Congo program that the vault, not the commercial market, is absorbing, and if that metal is what the unnamed manufacturers actually asked to store. Copper can be a real request. Lambert said no one mineral stood out. He did not say copper was absent. A series of deliveries against named offtake would be a pile. One conference sentence is a start.
The reading is wrong if you treat Tenke’s Chinese operator as proof the cargo is worthless. The title Gécamines sold is a state share. A share can be real metal. The point is narrower. The vault did not, with this parcel, step outside the existing Congo-China production system. It stepped into it and took a slice. That can still be rational. It is not independence.
The reading is wrong if the National Defense Stockpile and Project Vault are used as one pile in a model. They have different owners, different customers, and different ideas of what “enough” means. Adding their tons, if you ever get tons for both, will invent a reserve nobody administers.
The reading is wrong if your minerals view is a ceiling. Up to $10 billion, maybe $1 billion of loans, nearly $2 billion of private capital, $500 million from each of two traders, 12,000 tonnes of copper. Those figures do not sum. They describe a credit line, a capital plan, a commitment, and a cargo. Only the cargo has weight. The rest have conditions. Conditions are not metal.
The idea, once
Project Vault is a facility. On February 2, 2026, EXIM approved a direct loan of up to $10 billion. The design next to it is about $2 billion of private money. In September Glencore and Mercuria each committed $500 million. Early manufacturer interest included Clarios, GE Vernova, Western Digital, and Boeing. Interest is not a draw. This week Brett Lambert said VaultCo is likely to announce several hundred million dollars of loans, maybe a billion, to more than a dozen unnamed manufacturers, then another dozen. He is talking to more than 70 firms. He will not name the borrowers. He said no single mineral stood out, and that the specificity of the requests surprised him. The requests include magnets, powders, and alloys, not only ore.
The metal you can count is smaller and clearer. Mercuria’s Kostas Bintas said 12,000 tonnes of Congolese copper entered the vault the week of October 15. That parcel sits inside a Gécamines plan to move 100,000 tonnes in 2026 from the state’s 20 percent of Tenke Fungurume, a mine CMOC operates. It is not the 53,290 tonnes of Congolese cathode the United States imported in July. Those July tons were commercial, a record, and a different pipe. A customs month is not a stockpile. A stockpile door that opens onto cathode is not yet a door that opens onto the magnet a factory said it needed.
The facility is the ceiling, the loan, and the list of asks. The pile is what has been bought, stored, and matched to a line. This week the public pile is one copper cargo. Everything else is still a verb in the future tense. If you cannot say which one you are trading, you are trading the name of the building. The building does not, by itself, keep a line running.