The dollar's portion of allocated official reserves was 57.13 per cent in the first quarter of 2026, up from 56.42 the quarter before, and about half of that rise came from the exchange rate rather than from anyone buying. The euro fell to 20.03. The renminbi reached 1.99 after fifteen years of state effort. Every forecast that had the system ending by now was wrong.
The movement that is real is not in currencies at all. Central banks bought a record 289 tonnes of gold in the second quarter of 2026, sixty-two per cent more than a year earlier. Gold now stands at 27 per cent of official reserves against 22 per cent for US Treasuries and 15 per cent for the euro, leading for the first time since 1996. Reserve managers are not changing currency. They are stepping partly outside currency, into the one asset no ministry can freeze by decision.
Nobody moves into a metal that pays nothing because they expect a better return. They do it because they have priced a risk that used to be unthinkable.
The risk has a date. Immobilising between 275 and 320 billion euros of Russian reserves in 2022 was lawful, was proportionate to an invasion, and was right. It was also a demonstration, and it was watched. Poland bought 64 tonnes of gold this year, more than any other central bank, and Poland sits inside the European Union and inside NATO, in no danger of sanction from anyone. What is being hedged is not punishment. It is the discovery that a rule can turn out to have been a decision.
The repair is not to put the weapon down. Sanctions are what a lawful order has instead of force, and the alternative to freezing Russia's reserves was not a gentler world. The repair is to make their use predictable: thresholds set in advance, agreed multilaterally, published, with a process a neutral state can read and rely on. A sanction that follows a stated rule deters. A sanction that looks like a decision taken by whoever holds the clearing system teaches forty governments to hold something else.
The dollar's portion of allocated official reserves was 57.13 per cent in the first quarter of 2026, up from 56.42 the quarter before. About half of that rise is the exchange rate rather than anyone buying. The euro fell to 20.03. The renminbi reached 1.99 after fifteen years of swap lines, clearing banks and a payment system built for the purpose. These are the numbers, and they are not the numbers the argument has been conducted on.
The genuine long-run fact is a drift, not a collapse. The share was around 70 per cent at the turn of the century and has come down across twenty-five years. What it lost did not go anywhere in particular: it dispersed into Canadian and Australian dollars, the Korean won, the Swedish krona, the Swiss franc. That is a long tail, and nobody proposes any of it as a centre. Diversification away from a currency is not the same event as succession to it, and treating the two as one has produced a decade of bad forecasting.
This year supplied the test. The strait closed on 4 March 2026, more than ten million barrels a day were out of the market by the twelfth, and Brent finished the month at 118.35 dollars. If the reserve system were fragile, that was the shock to break it. The share went up instead, for a reason so plain it is embarrassing: a shortage in a commodity priced in dollars makes importers need more dollars.
So the thing that is moving is not a currency. Central banks bought 289 tonnes of gold in the second quarter of 2026, a record, sixty-two per cent above the year before. Gold now stands at 27 per cent of official reserves against 22 per cent for US Treasuries and 15 per cent for the euro. It has not led since 1996.
Nobody moves into a metal that pays nothing because they expect a better return. They do it because they have priced a risk that used to be unthinkable.
Look at who is buying. Poland took 64 tonnes this year, ahead of Uzbekistan at 33, China at 25 and Kazakhstan at 20. Poland is the interesting entry, because Poland is inside the European Union, inside NATO, and at no risk whatever of having its reserves touched. A country facing no sanction is buying insurance against sanctions. That is not a judgement about Washington's intentions. It is a judgement about how far intentions are the relevant variable.
Honesty first about our own instrument. The freeze weapon is older than the dollar and it was not invented in Washington. When Iran nationalised its oil in 1951 it was Britain that blocked Tehran's sterling accounts, boycotted the crude worldwide and withdrew the technicians from Abadan. The economic weapon is a wartime instrument that the twentieth century repurposed for peacetime, and it unsettles the system for exactly that reason. This is a property of the job of issuing the world's money, not an American vice, and the left does itself no favours by pretending otherwise.
What 2022 established was scale. Between 275 and 320 billion euros of Russian reserves were immobilised, around 210 billion of it inside the European Union and between 185 and 190 billion at a single Belgian depository. Every one of those decisions was defensible on its own terms. Together they changed the question a reserve manager asks, from whether their government might invade a neighbour to whether their government might one day find itself on the wrong side of a coalition.
That is the ground the anti-imperialist reading occupies, and it takes it too far. The claim that the dollar is held under duress cannot explain who holds it. The largest official holders are Japan, Switzerland and the euro area: allies with open capital accounts, free to diversify tomorrow at no political cost, who do not. Six freezes across half a century, against a system used daily by nearly two hundred states, is not a regime of coercion. It is a tool used rarely and, so far, on defensible occasions.
The reading from the other direction says the reserve role hollowed out American industry, and it makes an error about causation. The current account balance is investment minus saving; capital flows in first and goods follow. That was the case a Federal Reserve chairman made in 2005 about a global surplus of savings looking for a home, and the accounting has not changed since. The clearest evidence is recent. In the first half of 2026 the American trade deficit fell by 189.3 billion dollars, or 33.8 per cent, against the same period of 2025, with exports up 11.7 per cent and imports up 0.4. A third of the deficit went in six months and the reserve role did not move at all.
The part of that argument which is true should be conceded without hedging. Manufacturing employment is around 12.6 million, about 8 per cent of the workforce against 13 per cent in 2000, and between four and a half and five and a half million of those jobs have gone since. Trade exposure did a great deal of it, whatever the productivity accounts say, and the towns it happened to were told for twenty years that it had not. But the remedy is adjustment policy, wage floors and public investment where the plants closed. Ending the reserve role would not return one job.
And the ledger usually quoted is half a ledger. In June 2026 the goods balance was minus 102.1 billion dollars and the services balance plus 28.8 billion, for minus 73.3 billion together. The services surplus is not small and it is not incidental, and it is left out of nearly every account of what the arrangement is supposed to cost.
The warning worth taking seriously came from inside the building. Jacob Lew, as Treasury Secretary, told an audience in 2016 that sanctions have to be multilateral, tied to a stated objective and lifted when they succeed, or “financial transactions may begin to move outside of the United States entirely”. That was not a critic of the instrument. That was the office responsible for it, describing the mechanism by which it gets spent.
So the programme is not disarmament. It is procedure: published thresholds, a stated objective for each designation, a sunset that bites unless renewed, multilateral agreement before reserves are touched, and a route by which a state can discover what it must do to be released. Every one of those makes a sanction more credible rather than less, because a punishment nobody can predict is not a deterrent. It is only a risk, and risks get hedged.
The place this gets decided is the 140-billion-euro reparations loan now proposed against the Russian holdings, drafted so that Russia keeps ownership and the word confiscation is never used. Do it through law, with a finding behind it and a court that could have said no, and the rule survives its own exception. Do it in a communique because the money happened to be there, and the next 289 tonnes are already bought, while the argument about whether the dollar is being replaced goes on missing what is actually happening.
