The Coming Precious Metals Shock Has Already Begun in the Vaults and It Will Decide Everything From the Price of Bread to the Price of Borrowing

Guest Post by Milan Adams



Metal markets rest on a distinction that most participants stop noticing early in their careers, which is the difference between owning a bar and owning a promise about one. Promises clear instantly, travel by wire, cost almost nothing to move and can be written in quantities far exceeding the bars behind them, and for roughly half a century that arrangement has served the wholesale bullion trade well enough to make the alternative look archaic. Bars are heavy, insured, assayed, trucked and counted. Promises are elegant, and the whole of the modern trade sits inside that elegance.

Every so often the elegance becomes the problem.


Across the autumn of 2025 and the winter that followed, the world’s central bullion market discovered what happens when the promises outnumber the bars by a margin wide enough to be felt on an ordinary Friday afternoon in October. London’s vaults emptied of metal that anybody could actually take possession of, the cost of borrowing silver briefly reached thirty-nine per cent annualised, aeroplanes were chartered to fly bullion across the Atlantic, and four months later the price of silver doubled and the price of gold went to five thousand six hundred dollars an ounce before both collapsed with a violence that erased the entire advance and rather more besides.

What makes this worth a long examination is that the collapse has been widely reported and almost entirely misunderstood. Commentary has settled on a simple story about a bubble, a hawkish central banker and an exchange raising margin, which captures the shape of events while missing the mechanism and the aftermath. Bubbles explain why a price goes too far. They explain nothing about why the market ran out of deliverable metal in the first place, why a Section 232 proclamation declined to impose tariffs and promised price floors, why the People’s Bank of China bought through the entire drawdown, why the United States Treasury carries a trillion dollars of gold on its books at eleven billion, or why a clearing system built on netted obligations has spent a year publishing figures that quietly describe a thinning floor.

Eight propositions organise the investigation that follows, and each of them runs against some part of the conventional account.

1 — The squeeze of October 2025 was a settlement event dressed up as a price event. Metal became scarce in the one city where the world clears, and the scarcity expressed itself through a borrowing cost belonging to a different era of market history. Dealers in London spent a fortnight ringing each other for bars while the one-month lease rate climbed past thirty per cent and then past thirty-five, and the men who had spent careers pricing risk found themselves pricing logistics instead.

2 — Silver’s deficit has become a structural feature of an industrial supply chain, running six consecutive years. The 2025 shortfall reached 40.3 million ounces and the cumulative drawdown since 2021 comes to roughly 762 million, close to a full year of global mine output, taken out of a stock that has gone unmeasured for a century. A metal that gets consumed by solar manufacturers and connector plants and then thrown away behaves nothing like a metal that gets buried in a vault.

3 — Official-sector demand has decoupled from price. Central banks bought a thousand tonnes a year for four years running, and the People’s Bank of China bought through twenty-four consecutive months, straight through a thirty-five per cent drawdown, lifting holdings towards 2,366 tonnes. A reserve manager adding metal at five thousand dollars an ounce is buying insurance, and insurance buyers keep paying after the fire.

4 — The January collapse was administered through margin mechanics at a moment of maximum leverage. CME Group lifted maintenance margin on a silver contract from $15,000 to $16,500 and initial from $16,500 to $18,150 in the early afternoon of the last Friday in January, effective immediately, and silver closed twenty-six per cent lower near $85 after touching a thirty-one per cent intraday loss. The exchange performed exactly as a risk manager should, and the consequence landed on ordinary holders who had done nothing wrong except borrow against a rising asset.

5 — The correction has repaired almost nothing. Debt stands higher than it did when the year opened, at $40.2 trillion with $1.05 trillion of annual net interest against a $956 billion defence budget, the free float remains thin, the deficit persists, and the policy apparatus has produced intentions where mechanisms belong.

6 — Silver has been designated a critical mineral and the government has begun treating it as infrastructure. In November 2025 the Interior Department added silver to the national list alongside copper, uranium, silicon and potash, and in January a Section 232 proclamation found that imports threatened national security before ordering negotiations towards price floors rather than tariffs. Metals policy has started to look like grid policy.

7 — The systemic variable worth watching has shifted from price to delivery. A spike is survivable and the world has absorbed several. A clearing member that fails to hand over metal is a different order of event, and silver carries the exposure with zero central bank stockpile behind it and no lender of last resort, while a fifth of it flows into solar farms and server racks that run on construction schedules.

8 — The physical floor is thinner than anybody outside the trade understands. At the worst of it, the metal immediately available in London outside the exchange-traded funds fell to perhaps a hundred and fifty-five million ounces, roughly seventeen per cent of what the vaults held and about ten weeks of world demand. Ten weeks of demand is what stood between an ordinary retail mania and a genuine default.

Here is how the pieces connect, and why several of them are unflattering to nearly everybody involved.

Where the world’s silver and gold actually change hands on nothing but promises, inside a private club under a city that publishes almost nothing, and what happened during the fortnight when every promise came due at once and the floor had run empty​


London has priced gold twice daily since 1919, when five firms met in the offices of NM Rothschild and fixed a number across a mahogany table. The ritual outlived the Depression, the war, the closure of the gold pool in 1968, the collapse of Bretton Woods and the ferocious bull market of 1979-80 that briefly turned the fix into front-page news. What finally killed it was the telephone. Administration passed to the London Bullion Market Association in 2015, pricing moved onto an electronic auction run by ICE Benchmark Administration, and the mahogany table became a server.

Settlement changed far less, and settlement is where the whole story lives.

Wholesale bullion in London clears through London Precious Metals Clearing Limited, a private company owned by the clearing banks themselves, whose members net their obligations against one another and settle the residual. Netting is what permits a market to turn over hundreds of millions of ounces each day against a vault stock in the low hundreds of millions. Netting also severs any fixed relationship between the bars and the claims written against them, and no designer ever built a market in which the efficient mechanism would one day become the fragile one.

Beneath the netting sit unallocated accounts, and the distinction between those and their allocated cousins carries the entire argument. An unallocated account amounts to a debt. The bank owes a weight of metal, the client stands as an unsecured creditor for the duration, and the bank may lend that metal onward because until somebody asks for it the metal belongs to the bank. Allocated metal behaves differently: segregated, identified by serial number, ring-fenced from the balance sheet, and carrying a storage fee for the privilege. Wholesale participants overwhelmingly use the unallocated form. Allocated custody costs money, moves slowly, and the purpose of the market is movement.

Fifty years of use have validated the convention, which works because the overwhelming majority of participants take delivery only in the abstract, because the ones who intend to take it give notice, and because the floor has always held enough metal to satisfy whoever gave notice that day. The arrangement amounts to a bank run prevented every single day since 1968.

Three properties of that structure deserve attention before the chronology resumes. Turnover bears almost no relation to inventory, which means the ratio between the two can widen for years without anybody noticing and then snap back into relevance inside a week. Legal title to the metal sits with the bank until allocation occurs, which means the client’s exposure during the interval is credit exposure to an institution, and credit exposure behaves differently from property exposure once confidence thins. And the clearing club publishes aggregate vault tonnage monthly while disclosing ownership, allocation status and free float nowhere, so the market’s own participants estimate the floor from scraps.

Autumn 2025 came within a fortnight of converting all three properties into a single problem.

Metal had been leaving the floor for months before anybody counted the gap. Through the late summer, coin shops across the American Midwest and the Mountain West sold out, and premiums on a one-ounce silver eagle climbed to levels that made the spot quote look like a rumour. Retail demand is trivial in tonnage and enormous in behaviour: it arrives all at once, from one direction, and it takes possession, which means every ounce it buys leaves the lending pool permanently. Bars go into a safe in Ohio and stop existing as far as anybody else is concerned.

Exchange-traded products had already absorbed most of the remainder. Global silver ETP holdings climbed to 1.3 billion ounces across 2025, a twenty-six per cent jump in twelve months, and reached a record 1,276.7 million by March 2026. Metal inside a fund belongs to its shareholders and the custodian holds it off the lending market.

Subtract the funds from the vault stock and the free float falls out. By late September that figure sat somewhere beneath a hundred and fifty-five million ounces, and the share of London holdings standing outside the funds had dropped to roughly seventeen per cent. Seventeen per cent of a vault stock sounds survivable until the nature of the vault stock registers: one city’s inventory, in a market where the world clears through that city, against a demand base that consumes more than a billion ounces annually.

Borrowers paid whatever lenders named. Lenders staring at the same screen discovered they could name almost any figure and get it, and thirty-nine per cent came out of that discovery.

Relief arrived in physical form, which is the part worth studying hardest. Tonnes came out of New York vaults and out of China, landed at Heathrow, and within a fortnight the lease rate had collapsed back towards one per cent. Reuters covered the easing in the flat register wire services reserve for matters that have stopped being emergencies. COMEX stocks had been built up through the preceding winter during the tariff scare, so metal existed to be moved; Swiss refiners were running normally; flights were running normally; the arbitrage closed.

Most of the market read that sequence as proof of resilience — plumbing held, arbitrage worked, doom-mongers early and therefore wrong. Each statement holds true and the conclusion drawn from them still misses, because what actually occurred is that the world’s central bullion market got rescued by air freight and by a pile of metal that happened to sit in the wrong country for reasons unrelated to supply. Had the tariff scare of 2024-25 failed to drag hundreds of millions of ounces across the Atlantic, New York would have had nothing to fly.

image-5.png


The episode also demonstrated that the mechanism resolves pressure by moving it rather than dissolving it. Somebody paid thirty-nine per cent. Somebody’s short position left on a stretcher. Bars returned to the pool, the pool looked deep again from outside, the market resumed climbing, and four months later it climbed into a wall.

Underneath the episode sat a slower accounting that had been running for years and shows every sign of continuing. The Silver Institute’s World Silver Survey, researched with Metals Focus, recorded a 40.3 million ounce shortfall for 2025 and initially projected 46.3 million for 2026 before revising towards thirty-five as high prices chewed through demand. Cumulatively since 2021 the market has consumed 762.1 million ounces, some 23,705 tonnes, beyond what it produced, which approximates a full year of global mine output drawn out of above-ground stocks that have gone unmeasured for a century.

Silver differs from gold in that most of what has been dug up has been thrown away. It went into photographic film and down a drain, into solder and into landfill, into mirrors and into the backs of photovoltaic panels where it sits in grams and becomes economic to recover only at prices the industry has never managed to hold long enough to justify building a recovery chain. The above-ground pile runs thinner than the annual flow, which inverts the ordinary relationship between stock and production and makes every deficit year bite harder than the one before it.

Roughly fifty-eight per cent of silver demand is industrial, on the Institute’s own accounting, and that single proportion governs everything downstream of it. The metal behaves like a small, furious relative of copper instead of a leveraged gold. A shortage here hurts more than a shortage anywhere else in the metals complex, because gold shortages inconvenience people while silver shortages stop lines, and lines run on schedules. A solar farm under construction and a server rack on a delivery date both need the metal in the specified quantity on the specified day, and neither keeps a warehouse of alternatives.

The industry answered the price quickly, rationally and insufficiently. Manufacturers cut the loading per cell, reformulated the pastes, thinned the fingers, redrew the busbars. The Survey’s authors compressed the problem into one dry clause in the 2026 edition, observing that high prices would render silver “somewhat the victim of its own success” through price-led thrifting and substitution in photovoltaics and elsewhere. Industrial demand was forecast down three per cent across the year, and in November the Institute cut total demand five per cent to 1.08 billion ounces with jewellery accounting for the largest single reduction.

Metallisation paste deserves the detail, because the numbers explain the constraint. A modern cell carries silver in the fingers that collect current and in the paste that bonds them, measured in milligrams per cell, and the loading has fallen steadily for fifteen years as manufacturers chased cost. High prices accelerated a curve that had been grinding along quietly. Paste formulators reduced the silver fraction, printers laid thinner lines, and designers experimented with plating and with copper at the edges of the architecture.

Thrifting works. Substitution works. Neither abolishes the requirement. A heterojunction cell needs silver in the paste and performs worse as the loading falls, and the volume of cells the world intends to build keeps rising even as the loading per cell declines. Industrial consumption drops three per cent in a year when installation schedules slip; let those schedules recover and the tonnage returns on a higher base than the thrifting removed.

The same pattern reaches into electronics, connectors, switch contacts, solder, and every piece of switchgear in every substation scheduled for replacement as the electrical economy expands. Silver carries the lowest contact resistance of any metal and the highest thermal conductivity, which is why it appears in places where a cheaper substitute degrades performance measurably. Designers can trade against it at the margins. They carry it where the margin runs thin.

Above the manufacturing layer sits a supply side that answers slowly and a recycling side that barely answers at all. Almost all the silver ever mined has been discarded, and a photovoltaic recovery sector at scale has yet to be built and would need a decade of volume to stand up. Scrap from electronics runs through informal channels in much of the world and recovers a fraction of what it might. The consequence is a market that must clear every year through price and through demand destruction, in that order, because the queue holds nothing else.

Gold’s 2025 was a different animal altogether and the difference matters. Nobody thrifts gold. Technology absorbed eighty-five tonnes in the third quarter of 2026, a rounding error beside seventeen hundred tonnes of investment and official buying. Gold amounts to a monetary asset with a jewellery habit, and the price of it through 2025 was being set by people forming a judgment about the durability of the dollar system.

The year opened near $2,634. By the eighth of October it had cleared $4,000 for the first time in history, and silver broke $50 the same week, a level unseen since 1980. The move grew violent enough that the usual explanatory machinery — rate cuts, real yields, the dollar — stopped being adequate to the task. Something larger and vaguer was happening, and Wall Street, which cannot hold a vague thing in its mouth, promptly named it.

The debasement trade entered the vocabulary that autumn and saturated it inside a fortnight. The phrase predates the trade by several centuries, naming what Nero did to the denarius and what Henry VIII did to the English shilling when both cut the precious content and pocketed the difference. Applied to 2025, it meant a bet that the issuing authority of the world’s reserve currency would keep expanding its own liabilities faster than it could credibly promise to stop.

Goldman Sachs analysts Daan Struyven and Lina Thomas carried the expression into client notes by January. Bloomberg ran a newsletter on the subject in October under the heading that it was Wall Street’s latest obsession, which roughly marks the point in any cycle where the thing has already occurred. Beneath the fashion sat accounting that refused to go away.

Fiscal 2026 closed with $1.05 trillion of net interest paid, $1.4 trillion on the Government Accountability Office’s audit of the Treasury’s books, and a defence budget of $956 billion. Total debt reached $40.2 trillion against a deficit of $2 trillion. Revenue grew three per cent, spending grew six, interest grew eight. The Congressional Budget Office projects $23.1 trillion of deficits across the coming decade, $20.1 trillion of interest through 2036, and debt at 175 per cent of output within two decades.

Interest crossing defence carries a significance beyond the ledger. Defence spending is discretionary in principle and politically sacred in practice, which makes it the benchmark against which other claims on the budget get measured, and an expenditure line that outruns the sacred one begins to crowd out everything the state would rather do. Fiscal 2026 marked the first year in which servicing past borrowing cost more than the entire military establishment, and the trajectory of the line runs steeply upward in every published baseline.

That converts a fiscal problem into a monetary one. Austerity has zero coalition. In November the Democrats took both chambers and the dollar index slid to a one-month low on the assumption that divided government spends more instead of less, which tells you what the market has learned about incentives. Both parties have spent a decade discovering that cutting anything costs votes and borrowing costs nothing until later. The line of least resistance runs through the bond market, and the bond market has started charging for it.

You can watch the strain in the foreign bid. Overseas holders shed $75.5 billion of Treasuries in September, the sharpest monthly drop in six months, led by China and France. Official institutions now hold a share of the debt that has been falling since 2008, and the buyers who replaced them are private and price-sensitive in a way reserve managers were never required to be.

That is the world in which an ounce of gold travels from $4,333 at the start of January to just beneath $5,600 four weeks later. Nobody possessed a spreadsheet demonstrating that it should. The alternative — forty trillion of paper with a trillion-dollar annual service charge and zero political mechanism for reducing either — simply began to look like the risk instead of the hedge.

Official institutions had decided long before retail noticed. That is the fact the trade discovered early and the public discovered late.

Central banks bought an average of a thousand tonnes annually for four years, against roughly half that in the preceding decade. Pace survived price. The first quarter of 2026 came in at 244 tonnes before revisions stripped it to fifty-seven as some buyers were reclassified into the over-the-counter category; the second quarter produced 289, a record for any second quarter and a fivefold rise on the revised first; the third added 321, forty-two per cent up year on year and well clear of the five-year quarterly average of 227.

China’s programme carries the monthly count. Twenty-four consecutive months of purchases lifted official holdings towards 2,366 tonnes, roughly forty per cent higher across the streak, and gold still amounts to a few per cent of Chinese reserves, a figure that reads either as a finished programme or a commenced one depending on the reader. The PBOC bought through the entire thirty-five per cent drawdown. It bought in the week silver lost a third of its value. It will almost certainly buy in December.

Reserve managers rank among the least momentum-sensitive participants in any market, and that is exactly why their bid matters out of proportion to its size. A thousand tonnes a year equals a fifth of mine supply and, more importantly, a standing bid beneath the price that remains indifferent to what the price happens to be.

The World Gold Council’s June survey put numbers on the mood. Seventy-six institutions responded, the highest participation in nine years. Forty-five per cent intended to add over the following twelve months, a record and two points above the year before. Eighty-nine per cent expected global official holdings to rise. Ninety-three per cent already held some gold, up from eighty-one a few years back. Eighty-four per cent expected metal to occupy a larger share of their reserves five years out.

Read as a group, those four figures describe an institution that has stopped treating gold as diversification and started treating it as insurance against a specific set of outcomes it declines to name in the questionnaire. The survey never asks them to name it. The outcomes are the same ones the reserve community has priced quietly since February 2022, when the G7 immobilised roughly three hundred billion dollars of Russia’s central bank reserves and demonstrated that the safest asset on earth is safe only while you remain on the right side of somebody else’s foreign ministry.

That was the break. Everything since amounts to reallocation, executed slowly enough to avoid moving the price against itself and dressed in the vocabulary of portfolio construction.

Custody adds a layer the survey leaves out. A reserve held abroad sits inside somebody else’s legal system, and the lesson of 2022 reached the vaults as well as the balance sheet. Central banks brought about 6,900 tonnes home from American and British vaults between 1972 and now, and the present wave moves faster and more visibly than earlier rounds. France holds all of its gold domestically. Germany keeps 1,236 tonnes, or 36.6 per cent of its reserve, at the New York Fed, with its Taxpayers Federation and senior CDU politicians pressing for return on the argument that bullion has stopped being safe in American vaults. The Bundesbank resists on grounds of institutional diplomacy. Italy, holding 2,452 tonnes, keeps about forty-three per cent in New York and resists similarly.

The Fed’s vault beneath 33 Liberty Street still holds roughly 6,300 tonnes of foreign-owned metal, worth well over eight hundred billion dollars at recent prices, an enormous quantity of somebody else’s property in a building in lower Manhattan, with the political weather around it changed since a president began questioning the independence of the institution above the vault.

Fort Knox poses the domestic version of the same question and every party has mishandled it. The Treasury has published zero comprehensive assays since the 1950s. A government efficiency office promised an audit in early 2025 and quietly dropped it in favour of a bill, the Gold Reserve Transparency Act, which has yet to pass. A senator visited, and the assurance that followed amounted to everything being present and accounted for, minus a bar list. The Treasury Secretary said the same to reporters while declining to travel to Kentucky for a public count.

Part of the scepticism is earned and part misses. The metal almost certainly sits where it says it sits. Exotic theories about tungsten cores and sealed deep vaults have never quite died because nobody produced the document that would kill them, and the document is a bar list, and the bar list stays in a drawer. Every commercial depository twice the size of Fort Knox performs that verification as routine.

Treasury gold carries a separate accounting and it deserves its own examination, since it is where fantasy and substance meet. The United States holds 261.5 million ounces, of which about 147.3 million rest at Fort Knox. The department carries the hoard at the statutory price of $42.22 an ounce, fixed in 1973, valuing it a little above eleven billion dollars. January’s peak made it worth north of 1.3 trillion. Even at today’s beaten price it clears nine hundred billion with room to spare.

That gap, a hundredfold discrepancy between book and market, amounts to the largest unexploited line item in Western public finance, and it has sat there, unremarked in polite company, since the Nixon administration. Every few years somebody notices and writes about it. Then everybody returns to pretending a central bank balance sheet is a technical document instead of a political one.

Revaluation would run through gold certificates. Treasury issues them to the Federal Reserve against the metal and the Fed holds them as an asset, locked to the statutory price under 31 USC 5117, which means the Fed’s balance sheet carries the gold at the same fiction as the Treasury’s. Congress would have to amend 5116 and 5117, Treasury would reissue at the new price, and the Fed would credit the difference, up to roughly $1.18 trillion at recent prices, back to Treasury.

Mechanically that amounts to a bookkeeping entry. Politically it conjures a trillion dollars of fiscal room minus issuing a bond or raising a tax, arriving precisely when the interest line has just passed the defence line and both parties are hunting money.

Scott Bessent has denied it twice. On the All-In podcast in March 2025 he said the Treasury was revaluing nothing; in July he told Fox that gold “doesn’t matter” for the value of the dollar. Behaviour since has been read by a growing number of observers as preparation: Kevin Warsh, who has written seriously about gold’s monetary role, occupies the chair; a former senior Bessent aide was reportedly bound for the office handling certificate mechanics; and the department has been more forthcoming about the physical state of the reserve than any administration in seventy years, which sets a low bar and still marks a change.

Confirmation has yet to arrive. Revaluation stands as a live idea in place of a policy. It keeps returning because it solves a political problem with an accounting solution, and accounting solutions are how systems get modified by people unable to admit they are modifying them.

Washington’s policy response proved stranger than its reserve management. In November 2025 the Interior Department published a final critical minerals list adding ten commodities, and silver appeared on it beside copper, uranium, silicon, potash and metallurgical coal. A metal that had spent its modern history as an industrial commodity with a monetary shadow became, overnight, a national security input.

Listing triggered a Section 232 investigation. In January 2026 the President issued a proclamation finding that processed critical mineral imports threatened national security and then did the thing almost nobody had prepared for: he imposed zero tariffs.

Negotiations followed instead, with a report due within 180 days on how they were progressing and a stated aim of securing price floors for domestic production. Floors, in place of tariffs or quotas or a stockpile. A floor is a far more interesting instrument than a tariff, because a tariff taxes the buyer and shields the marginal domestic producer at everybody else’s expense downstream. A floor requires a buyer of last resort at a stated price, which converts a market into a managed one and, if the commitment carries credibility, removes the downside for anyone contemplating a mine in Nevada. That is precisely the attraction and precisely the difficulty, since nobody has published a level, a funding line or a procurement body.

By the time the proclamation landed, silver had already run to a hundred and twenty-one dollars. Whatever floor eventually takes shape is being designed for a market that has since halved.

Floors entered American minerals policy through the same door as the critical list, and the pairing describes a coherent posture: designate the input as strategic, find that imports threaten security, then underwrite domestic production while sparing the downstream user the tax. Applied to a metal trading at $44 an ounce and costing perhaps $25 to produce at the primary operations still standing in the United States, a floor would guarantee a margin to a very small number of mines and a very large number of by-product credits at copper and lead operations.

Applied to copper, also on the list, the consequences run wider, because copper is the entire electrical economy. Nobody has costed it publicly. The 180-day report came due in July and delivered more process than policy.

January arrived with gold at $4,333 and left it a whisker beneath $5,600. Silver doubled inside the same window.

Both markets cleared $4,500 and $100 respectively with the kind of ease that encourages the belief that liquidity is a permanent property of the contract. Gold touched $5,595 to $5,602 intraday on the twenty-eighth of January, depending on the feed and the currency, and closed at a record the following day. Twenty-nine per cent in under four weeks, and silver up more than a hundred across the same span.

Reasons to keep buying lay everywhere. Central banks were purchasing. The deficit had entered its sixth year. Silver had been designated critical in November. A Section 232 finding had landed and, though it imposed zero tariffs, it had pointed at floors, which is the more bullish instrument. The dollar was soft. The debasement trade possessed a name, a Goldman note and a Bloomberg newsletter. Retail was arriving in volume.

Reasons to stop were equally available and equally ignored. Positioning was crowded, leverage ran high, and the exchange had already restructured its metals margining, shifting platinum, palladium, silver and gold from fixed dollar amounts per contract onto percentage-of-notional calculations, the step an exchange takes when it stops trusting the notional to stay where it is.

On the twenty-eighth came the nomination of Kevin Warsh to chair the Federal Reserve.

Warsh is a hawk with a long paper trail arguing that the Fed’s inflation-fighting credibility had been spent. Markets read the nomination as the end of the easing cycle before the easing cycle had properly started. The dollar index gained nearly a per cent. Gold’s advance stalled, and silver, running on leverage and momentum, stalled harder.

CME raised the rate on high-risk silver accounts from eleven to twelve point one per cent after the close that day. A small number doing a large amount of work: a position margined at eleven per cent holds roughly nine times its capital in exposure, and at twelve point one it holds eight, and on a contract that has doubled in three weeks the gap between nine and eight is the gap between solvent and otherwise.

Silver broke the next session. On the Friday, with the market already sliding, CME went further. Initial margin on a silver contract moved from $16,500 to $18,150; maintenance from $15,000 to $16,500; effective immediately; notice out in the early afternoon in New York.

For anybody holding silver on leverage, that announcement constituted the entire world. A position comfortably margined at eighty-five dollars became marginal at seventy. The exchange sold zero ounces, merely asking for more, and the asking did the selling. Forced liquidation ran into a market whose marginal buyer had just been liquidated, which defines an air pocket, and silver fell through it.

Thirty-one per cent intraday. Twenty-six per cent at the close, near $85. The steepest session in the modern history of the contract, and among the largest single-day drops in any commodity within living memory. Gold shed about nine per cent the same day, its worst session since 1980 on some measures.

Assessing the exchange here requires care, because the reflexive reading casts it as the villain of the piece. CME manages a clearing house with a mutualised default fund, and its obligation runs to the solvency of that fund in place of the comfort of any individual participant. Percentage-of-notional margining reflects the reality that a fixed dollar requirement becomes inadequate as the notional doubles, and the sequence of increases across January tracked a sequence of price increases with a short lag. The intervention was orthodox, defensible and, from the clearing house’s perspective, overdue.

The consequences for holders were severe and entirely outside the exchange’s mandate. Every leveraged long converted into a forced seller at the same moment, and the resulting cascade took the price far below any level the fundamentals would have suggested. That is the structural asymmetry at the centre of futures markets: the mechanism that protects the system operates through the accounts of the people least able to absorb it.

Leveraged exchange-traded products amplified the fall in a way that deserves separate treatment, because their construction makes them inherently procyclical. A fund promising a multiple of the daily move must rebalance at the close, buying after rises and selling after falls, and the selling grows as the drawdown deepens. Systematic funds tracking moving averages cut exposure on the break of the same thresholds, and they cut.

By March, United States-listed gold funds had shed eighty-five tonnes, thirteen billion dollars, the largest monthly outflow in dollar terms on record and the third largest in tonnage, with North America alone supplying all thirteen billion and ending a nine-month inflow streak inside a single month.

Silver bottomed near $57.67 at the start of August, half its January peak, then slid further through the autumn.

Gold’s mid-March low came in around $4,100, a twenty-five per cent drawdown in six weeks, and the timing carries the interest, because the Iran war opened in February.

The asset that exists to be bought when shooting starts fell straight through a war.

Explanations for that inversion run along well-worn channels. War lifted oil, oil lifted inflation expectations, expectations lifted the rate path, and the rate path crushed an asset with zero yield. Gulf energy risk made the dollar attractive again while making the Fed’s task harder, and the market solved both problems one way: sell metal, buy bills.

Something deeper operated underneath, a mechanism every desk knows and few discuss aloud. During a genuine risk event, gold ranks among the few things you can always sell. Institutions needing cash through February and March sold gold because gold carried a bid, which tells you about their liquidity instead of their convictions. Euronews reported the dynamic at the end of March in four plain words, investors rushing to raise liquidity, and the phrase is exact. Gold becomes the collateral you post once the collateral you would prefer to post stops being accepted.

Warsh took office and held rates at his first meeting in June, stripping the easing bias from the statement. Nine of nineteen participants in the projections, which Warsh had previously disdained as a crutch, pencilled in a hike before year-end. In August he delivered a speech Capital Economics called “a far clearer and hawkish message” than his press conference. September brought a unanimous increase against an intense public campaign from the White House. By November, nine of eighteen officials had signalled more.

The dollar reached a nineteen-month high of 103.97 in mid-November. Gold reached a fourteen-month low. Silver, at $44.57, had surrendered the entire move and then some, roughly sixty per cent down from January and tracking the 1980 collapse closely enough that the overlays have begun appearing on trading desks.

Money changed hands throughout, and the direction of it is worth naming. Miners and royalty companies enjoyed the best first half in the industry’s history. Scrap dealers, refiners, mints and assayers printed money. Bullion banks earned the widest spreads in living memory and, among the London clearing members, the fee income that comes with owning the only door into a market everybody suddenly needed. Coin shops made a margin on the way up and a fatter one on the way down, buying back at spreads that made January’s queues look like a rounding error.

Losses landed on the people who bought the story late and the people who live downstream of the metal. A buyer at $110 silver sits more than half under water. A buyer at $5,500 gold has shed a third. A fabricator in Vicenza, a workshop in Arezzo, a jeweller in Coimbatore, a Swiss watchmaker re-weighting cases, a manager on a solar cell line watching paste triple — none held a position and all of them paid.

Mining has lived through its best and most frustrating year, a sentence that sounds contradictory and holds together. Record costs arrived in the same quarters as record margins. Average all-in sustaining costs reached $1,785 an ounce in the first quarter and a further record $1,930 in the second, up twenty-two per cent year on year on a sixteen per cent jump in royalties and supply chain inflation from tyres to mill liners, and margins still touched $3,076 an ounce in the first quarter, roughly double the year before. By autumn, with spot near $3,700 and costs holding, that margin had compressed a fifth and the equities had de-rated by nearly a third.

Mine production set a record too: 1,071 tonnes in the third quarter, the strongest third quarter ever, with the year to date running two per cent ahead and the full year heading for an all-time high.

None of that constitutes a supply response. Almost every ounce produced in 2026 comes from a decision taken before the price moved. Gold mines need ten to fifteen years from discovery to production inside a permitting environment that has grown harder, and the operations commissioning now were financed when gold sat at $1,300 and the case rested on a lower number. What rose was intensity at existing sites: higher grades mined, cut-offs lowered, marginal pits pushed back into service.

Cost inflation measures the tightness of inputs. When all-in costs climb twenty-two per cent in a year, you are bidding for the same drills, the same diesel and the same geologists as everybody else, and prices do what prices do when demand outruns supply. Above that sits a slower decay: grades decline across the major regions, the easy ounces were found and worked in the last century, and reserves replacement has trailed production at the majors for a decade, which is why the sector consolidates in place of exploring.

Silver’s supply side carries an additional wrinkle that the gold literature tends to ignore. Roughly seventy per cent of silver emerges as a by-product of lead, zinc, copper and gold operations, which means the volume produced responds to the economics of those metals instead of the economics of silver. A silver price at $120 brings forward very little primary silver capacity, because the mines that would produce it make their decisions against copper and zinc prices and credit the silver at whatever the market pays. Primary silver operations are a minority of global output and most sit in jurisdictions with long permitting queues and thin capital markets.

Scrap and recycling form the other half of supply and behaved in a way that surprised every model this year. India’s gold recycling surged sixty per cent in the second quarter to a record ₹75,819 crore, about $7.29 billion, as households brought in bangles held for thirty years and took the cash. Scrap had already risen twenty per cent year on year in the first quarter to 31.2 tonnes.

Then it died. Third-quarter recycling fell twelve per cent to nineteen tonnes while domestic prices ran forty-five per cent above the year before. The Economic Times caught the reason in a headline that could stand as a summary of the asset class: Indians refuse to cash in on gold in their lockers, hoping for a better price.

Household behaviour sits at the centre of that pattern and repays closer attention. Indian gold ownership is a savings technology embedded in a family structure, accumulated at weddings, stored in bank lockers and jewellery boxes, drawn on for medical bills, school fees and dowries, and inherited down a female line. The metal functions as an interest-bearing asset only in the sense that it can be pawned, and the pawn decision carries a stigma a Western holder would find unrecognisable. Selling the bangles means admitting the bangles have become money.

Price moves the decision at the margin and stops moving it soon after. A forty-five per cent rise brings out a wave of light scrap, then the remaining holders re-anchor on the new level and wait for the next one. Supply that responds for a quarter and then freezes produces exactly the pattern observed: a spike in recycling, a collapse in recycling, and a market that must find its metal elsewhere.

Demand destruction has been real and landed hardest on the people least able to absorb it. India’s third-quarter demand fell twenty-nine per cent to 162.5 tonnes; jewellery volume dropped a fifth; investment demand fell forty per cent; rural buyers on a monsoon income withdrew. Jewellery’s share of Indian demand collapsed to about thirty per cent, the lowest in data reaching back to 2000, while bars and coins climbed to fifty-two per cent of local demand net of recycling, the highest since that series began in 2013. The World Gold Council expects full-year Indian demand to be the weakest in six years.

The composition shift matters more than the tonnage fall. A market that buys jewellery consumes craft labour, supports a workshop sector and produces a slow, sticky stock of held metal. A market that buys bars consumes distribution margin and produces a liquid, price-sensitive position. India is converting from the first to the second under price pressure, and the conversion will outlast the price.

China’s numbers look gentler and point the same way. Shanghai Gold Exchange withdrawals ran 126 tonnes in January, 85 in February, 87 in June, 62 in August, a twenty-seven per cent year-on-year fall in the last, before recovering to 92 in September and 94 in October as buyers returned for lower prices. The pattern shows opportunistic restocking instead of conviction. Chinese demand clears through licensed banks into a rationed import pipe, and the premium over London acts as the pressure gauge on that pipe.

Import licensing gives the Chinese market a property Western commentary handles badly. The premium over London reflects the width of a pipe fixed by administrative allocation, so a fat premium describes demand pressing against a constraint in place of demand pulling metal freely. Bars cannot be shipped east to capture it at will, because every ounce entering legally passes a licensed bank and then the exchange. The premium persists, and its persistence is the point.

Taken together, the two markets mean the physical stabiliser has been flattened by the level of the price itself. At thirty-five hundred dollars an ounce gold stops being a mass-market product in the countries that once defined its demand, and Asian consumers buy only on falls and sell only on rallies, the inverse of the cushioning behaviour the trade leaned on for forty years.

Beneath everything sits the plumbing, which few want to discuss because discussing it sounds like conspiracy and the conspiracy version is mostly wrong.

COMEX publishes vault data daily, which makes it the most transparent component of the system and therefore the most scrutinised. On the twelfth of November its vaults held 22.9 million ounces, of which 14.2 million were registered, meaning deliverable against futures, and 8.7 million eligible. Registered coverage against open interest ran near thirty-five per cent, comfortable by ordinary standards and alarming by the standards of a market fresh from what this one has endured.

London publishes less. The LBMA reports aggregate vault holdings monthly, minus ownership, allocation status or free float. Clearing runs through LPMCL, a private company owned by its own members, which nets obligations and settles the remainder. Netting is efficient. Netting is also, structurally, what lets claims outrun bars.

Two markets, one price, joined by an arbitrage called the exchange for physical. Ordinarily the EFP runs a few dollars an ounce. When it blows out, metal moves, and it moved spectacularly through the winter of 2024-25 once the threat of tariffs on bullion dragged hundreds of millions of ounces across the Atlantic into New York. The threat receded, the EFP normalised, and the metal stayed in New York because the spread that had paid for the flight home disappeared with it.

That explains why COMEX stocks remained elevated all year while London ran tight, and why delivery mechanics became a live subject at the moment the market could least afford distraction. Mid-November found the EFP at minus 0.80 per cent of spot, inverted, with the deliverable bid sitting in London instead of New York, while 1.15 million ounces were reclassified out of the COMEX deliverable pool minus leaving the vault, which is what happens when warrant holders decline to offer bars.

Reclassification deserves a word because it is routinely misread. Metal that moves from registered to eligible stays in the same building and stops being available against futures. A holder does it when the contango fails to compensate for the risk of standing for delivery, or when the holder expects a squeeze and prefers to sit on the warrant. Movement out of the deliverable pool minus movement out of the vault is a statement about the holder’s expectations in place of a statement about the state of the building.

Assessing the coverage ratio requires the same care. Thirty-five per cent of open interest in registered stock is comfortable against the historical average, and open interest is a poor proxy for delivery demand because the overwhelming majority of contracts close before notice day. Against a genuine delivery run, the ratio would compress within days.

Regulation forms the most badly reported part of this and the untangled version runs stranger than the myth. On the first of July 2025 a wave of internet commentary announced that allocated physical gold had become a Level 1 high-quality liquid asset under Basel III, and gold Twitter declared the paper market finished. The LBMA issued a correction. The World Gold Council followed with a paper titled, with audible weariness, “You asked, we answered”.

Reality is narrower and points the same way. Gold carries a zero per cent risk weight for capital purposes and can be posted as collateral at a twenty per cent haircut. It sits outside the high-quality liquid asset definition for the liquidity coverage ratio, and unallocated gold carries an eighty-five per cent required stable funding burden under the net stable funding ratio. Allocated metal held in custody sits off the bank’s balance sheet, which alters the NSFR accounting for the custodian minus converting the metal into a Level 1 asset.

image-6.png


The gap is the point. Under the NSFR an unallocated claim costs a bank eighty-five cents of stable funding per dollar of exposure; a Level 1 classification would cost nothing. That leaves a standing structural incentive to hold metal instead of claims, and it rests on a policy choice a motivated regulator could reverse in an afternoon. Gornall and Tully argued for reclassification in a Suerf brief in April 2025. Nobody has acted.

Collateral tells the same story from the far side. CME Clearing accepts COMEX gold warrants and London bars. LME Clear accepts bullion through an approved LPMCL member as custodian, tiered by haircut and limit. Movement runs one way: metal enters the collateral systems of the derivatives markets slowly, with caps attached, because clearing houses need something that carries zero sovereign liability and clearing houses can read sovereigns.

The regulatory gradient therefore points consistently toward physical form, and it does so through three separate channels that reinforce each other. Capital treatment rewards holding metal with a zero risk weight. Collateral schedules admit it with a haircut, which makes it usable in a crisis. Funding treatment penalises the unallocated claim at eighty-five per cent. Taken together, they describe a framework written for a world of coupon-bearing paper that keeps conceding ground to an asset paying zero.

Africa is where gold-backed money has actually been attempted, and the results merit examination in place of romance. Zimbabwe introduced the ZiG in 2024, backed by physical reserves and foreign currency, after a decade of collapse in its predecessor; by May 2026 the Reserve Bank held 4.48 tonnes, and the currency has held better than its forebears with considerable help from the gold price. Ghana carved five billion cedis, about $429 million, out of a revised budget to buy gold and has been building a domestic purchase programme alongside, on paper, the groundwork for a gold-backed stablecoin.

The IMF flagged Ghana’s programme in its May review as a quasi-fiscal risk to the central bank balance sheet. That objection is correct and dull: a central bank buying gold from domestic producers at a set price runs a subsidy, subsidies cost money, and the money lands on the same balance sheet meant to be backing the currency.

Backing constrains, which is the reason to want it and the reason it stays rare. A currency tied to a fixed metal quantity cannot expand when the economy needs expansion, which is why the world left the gold standard. What looks likelier everywhere is a hybrid: gold as reserve anchor and political signal, paper as circulating medium.

Tokenised bars and gold-backed stablecoins digitise the hybrid and inherit the custody problem one software layer further out. A token amounts to a claim on a bar in a vault, which makes it an unallocated account with better marketing and thinner legal precedent. Audits are attestations. Redemption runs in business hours. The failure mode is October 2025 arriving through a different door.

Three technical features of tokenised issuance deserve scrutiny, because each maps onto a weakness already visible in the wholesale market. Attestation replaces audit, and an attestation states what a custodian says in place of what an assayer counted. Redemption windows compress the bank-run problem into office hours and leave the weekend exposed. And the token’s legal character varies by jurisdiction in ways the buyer rarely reads, so the same instrument can represent property in one place and a general claim in another.

History offers completed experiments in place of forecasts, and the closest one hurts.

Silver’s 1980 collapse began from a peak near fifty dollars, driven by the Hunt brothers’ corner and a genuine physical squeeze, and it took twenty years to recover. Gold peaked at $850 in January 1980 and regained that nominal level only in 2007. A generation of buyers who arrived in 1979 spent their working lives under water, with real returns worse than nominal ones because inflation ran while the metal idled.

The 2011 top reads shorter and kinder: gold at $1,920 in September, a twenty-eight per cent fall within weeks, a four-year grind to $1,050, silver from $49 to $14.

The 2026 chart tracks 1980 closely enough that desks have begun overlaying them. Parallels are real: a parabolic run, a physical squeeze, an exchange raising margin, a collapse of more than half in the white metal. The disanalogy is real too. In 1980 the Fed took the funds rate to twenty per cent to break inflation and sovereign debt amounted to a footnote. This time the central bank raises into forty trillion of debt carrying a trillion-dollar service charge, and every hundred basis points adds roughly a hundred billion of annual spending.

Two later episodes carry more operational relevance than either top. March 2020 delivered a liquidity dislocation in which Swiss refiners shut, flights stopped and the EFP blew out because metal physically could travel nowhere; the spread resolved through central bank action and a relaxation of logistics, and it demonstrated that the arbitrage depends on aeroplanes. The 2022 nickel squeeze on the LME ended with the exchange cancelling $3.9 billion of trades to save its own clearing house, an intervention still in litigation and still doing damage to the LME’s credibility four years later.

Those two are the useful comparators, because both describe the plumbing failing in place of the price. A 1980 comparison tells you how far a bubble can retrace. A 2020 or 2022 comparison tells you what breaks first and who gets rescued.

Systemic risk here concerns settlement, and the distinction from price risk runs through every institutional arrangement described above.

A spike is survivable and the world has absorbed several. What would genuinely reshape the economy is a moment when a clearing member fails to deliver, when a sovereign asks for leased metal back and gets told to wait, when a tariff or sanction splits the market into two price zones unable to arbitrage, or when a large producer de-hedges into a market with zero float and takes the offer out.

A silver settlement failure would hurt more than either modern precedent. Silver has zero central bank stockpiles behind it and no lender of last resort, and a fifth of it flows into things built against a schedule. Solar farms, switchgear and server racks refuse to wait for a market to normalise. Metal goes missing, and the line stops.

The lender-of-last-resort gap is the sharpest difference between the two metals and the one least discussed. Gold has an official sector that buys on weakness and, in extremity, an entire framework of central bank cooperation available to it. Silver has an industrial base that must buy whatever the price happens to be and a set of fabricators with balance sheets measured in weeks of working capital. The metal with the smaller market carries the larger operational exposure.

What breaks first when a country of forty trillion has to choose between its army and its interest bill, and how the next squeeze will pick its own date while everybody watches the wrong number​


One path from here runs flat, and it holds the highest single probability. The correction grinds through another year, gold drifts towards $3,000 and chops, silver falls to thirty-five and sits, retail leaves, funds bleed, mining equities de-rate, the financial press writes the gold-is-finished pieces it writes every cycle, and the structural arguments return to the drawer they inhabit between manias. Debt keeps growing through all of it, interest keeps compounding, the PBOC keeps buying quietly, the deficits keep running, and the resolution waits for the bond market to impose one, later and worse. A slow grind writes badly, which is why this path attracts the least coverage.

The other path justifies the word shock. A delivery starts it: a sovereign, probably an unfriendly one, demanding metal held in custody abroad. The custodian, which has been lending against it, finds the gap. Buying covers the gap at the offer into a market with zero float, and the price gaps. Margin calls go out. Leveraged funds sell. The exchange raises margin again, as it did in January, except this time metal can move from nowhere because the shortage spans regions in place of postal codes.

Silver clears a hundred and fifty. Gold clears eight thousand. Indian wedding season stops. Solar construction slips a year. Two or three industrial users hedge forward at any price and pin the top. Central banks keep buying because halting would constitute an admission. Then it falls again, because it always falls again, and the buyers of the top write what the January buyers wrote.

The instruments that separate the two paths sit in plain sight, which is the best thing about them. One-month silver lease rates matter: anything sustained above five per cent warns, above twenty constitutes an event. The EFP matters, and a persistent inversion means the deliverable bid has migrated to London while New York runs short of what it must hand over. COMEX registered inventory and its coverage ratio matter. So does the LBMA’s non-ETP share of London holdings, which tripped the squeeze at seventeen per cent and has since recovered to about twenty-four. The monthly PBOC figure matters more than any of them, as the sole central bank number published with consistency. The ratio matters, historically a buy signal at eighty-three and over any short window a coin toss. All-in sustaining cost matters because a price beneath it shuts mines, and shutting mines is the only supply response available.

Time is what the market handles worst, and the metal’s calendar runs in election cycles. A mine permitted in 2026 produces in 2038. A refinery built in 2026 runs in 2029. A recycling chain for photovoltaic silver has yet to be built and would need a decade of volume to justify itself. Any shortage therefore resolves through price and then through demand destruction, in that order, because the queue holds nothing else. Price rations; demand destruction is what rationing looks like from inside; and the 2026 correction has been, in part, precisely that, a market solving a supply problem by making itself unable to afford the metal.

Three calls follow from the framework. Silver retests fifty before it retests thirty, because the deficit holds, the free float stays thin, and the industrial buyers who spent 2026 running down stock get forced back into the market; the ratio compresses from eighty-three towards sixty by the middle of 2027. Gold finds a floor between $3,100 and $3,300 and defends it through the first half of next year, since the Fed runs out of hiking room before the Treasury runs out of borrowing room and the official sector bids into every dip. A Section 232 price floor arrives in some form before the 2028 election, at a level the industry will call inadequate and every industrial user will call ruinous, and it gets sold as national security in place of industrial subsidy.

Recent months have demonstrated something narrower than either camp claims and more useful than both. The physical market is thinner than its price implies. A hundred and fifty-five million ounces of float equals ten weeks of demand, and ten weeks of demand evaporates the instant a mania arrives. Mechanisms that resolve a squeeze, air freight and lease rates and margin, function, and they function by moving pain onto whoever holds the leverage.

Official institutions have decoupled from price. Retail has yet to. Industry possesses zero choice.

Politics has begun, clumsily, treating metal as infrastructure: a critical minerals list, a Section 232 finding, a floor existing as intention and a report existing as delay. All of it says the commodity has grown too important to leave to a market that clears in a private club in the City of London.

Forty trillion of debt, a trillion a year in interest, a divided Congress, a hawk in the chair, a war in the Gulf, a metal six years into deficit, a free float you could fit in a swimming pool, and a Treasury sitting on a trillion dollars of gold it carries at eleven billion.

Something gives. Which of those goes first, and how much of the remainder it drags down with it, stays the only question worth pricing.

Bars stopped moving in October. They started again in November, the price went vertical in January, and in February it landed on everybody’s head. Nothing structural got repaired in the interval. The vaults hold more than they held, the float stays thin, and the debt has grown a trillion dollars larger since the year turned.

Metal stays indifferent through all of it, and will still be there afterwards.

Tweet

Continue reading...

[ H/T The Burning Platform ]
  • Reading time 4 min read
  • Reading time 5 min read
  • Reading time 6 min read

If

  • Reading time 5 min read
  • Reading time 7 min read

Comments

There are no comments to display
Back
Top