Every quarter the IMF publishes the composition of the world's official reserves, and every quarter two camps read the same table and report different worlds. The readings: the dollar's share of allocated reserves stood near 57% through 2025, 56.9% in the third quarter1, down from about 66% a decade ago and above 70% at the century's turn. The dedollarization camp sees an empire in decline; the status-quo camp notes, correctly, that most of any given quarter's move is exchange-rate arithmetic, valuation effects explaining 92% of the Q2 2025 dip1, and that no rival currency is gaining what the dollar sheds. Both camps hold real cards, which is what makes the series a variable worth a standing watch rather than a verdict.
The move that is not arithmetic sits in a different asset class. Central banks bought more than 1,000 tonnes of gold in 2022, 2023, and 2024, 1,045 tonnes in 2024 alone, roughly double the pace of the 2010s, easing to a still-elevated 863 tonnes in 20252. The metal's price answered: through record after record to $4,000 an ounce in October 2025, up about 50% in a year in which the dollar index fell 10%3. Reserve managers are not rotating into euros or renminbi. They are rotating into the one reserve asset that is nobody's liability and sanctions-proof by construction, and they began in earnest in 2022, which is no coincidence.
The dollar's share of allocated global reserves in 2025, from over 70% at the century's start. Central banks' answer: three straight years of 1,000-tonne gold buying.
What actually moved, and when
The erosion's timeline tells its causes. The slide from 71% in 1999 to the mid-60s was globalisation's bookkeeping: reserve growth in countries that diversified at the margin, the euro's arrival, nothing strategic. The step down since 2022 carries a different signature, because it began the week the dollar system's core promise was revised: the freezing of Russia's reserves showed every treasury that dollar assets are, in extremis, conditional assets. The response was not flight, the share moved points, not tens of points, but it was universal in one respect: the gold bid broadened from the usual suspects to central banks on every continent, including American allies2. Insurance, not exit, is the accurate word, and insurance demand is information. Nor is the gold bid proof of coordination: central banks buy gold for reasons that run from sanctions insurance to routine diversification targets, and the same tonnes are consistent with several different intentions.
The rivals' table explains why exit is not on it. The euro holds about a fifth of reserves and cannot supply the world's safe asset while its bond market remains fragmented across nineteen treasuries. The renminbi, for all the summitry, holds around 2%, capped by capital controls its issuer shows no intention of lifting, because an open capital account is precisely what Beijing's model forgoes. The yen and sterling are stable garnish. What grows instead are the non-traditional currencies, Australian, Canadian, Korean, and gold, which is to say: the alternative to the dollar is not a successor. It is diversification itself, a reserve portfolio that looks less like a pyramid and more like a shelf.
Metric | Reading |
|---|---|
Dollar share of allocated reserves | 56.9% in Q3 2025 |
A decade ago / at 2000 | About 66% / above 70% |
Euro share | About one fifth |
Renminbi share | Around 2% |
Central bank gold, 2022-24 | Over 1,000 tonnes each year; 1,045t in 2024 |
2025 gold buying | 863 tonnes, below the streak |
Gold price | Past $4,000/oz, October 2025 |
History sets the tempo expectations. Sterling's reserve primacy outlived British industrial primacy by half a century and two world wars, sustained by incumbency, network habit, and the absence of a ready successor, the same three forces now working the dollar's side; when the handover came, it took a generation even with a willing heir standing by. The dollar's current erosion, a point every year or two with no heir at all, is running slower than the pessimists' timeline and faster than the complacent one, tracking sterling's long glide more than any collapse scenario. Reserve currencies do not fall. They fade, and fades are measurable, which is the premise of this page.
The fiscal backdrop is what gives the watch its edge this decade. The dollar's share is eroding into, not away from, an unprecedented supply of the asset itself: trillion-plus deficits as standing policy, debt service crossing the defence budget, and periodic political experiments with the central bank's independence and the debt ceiling's seriousness. None of that has yet cost the dollar its bid; the 2025 gold record and the dollar index's 10% slide3 suggest it is being priced at the margin. A reserve asset's privilege is exactly as durable as the issuer's restraint looks, and restraint is currently the scarcer commodity. The watch, in other words, is only half about the world's confidence in the dollar. The other half is about Washington's interest in deserving it, and both halves print in the same table.
Why the pace is the whole question
Reserve status is not a trophy; it is a subsidy with an invoice attached. The dollar's share underwrites America's cheaper borrowing, its deficits' financing, and the sanctions reach that has become its favourite instrument, the same instrument whose 2022 use started the insurance buying. A share drifting from 57 toward 50 over a decade would be absorbed in bond yields a few basis points at a time; a lurch driven by a fiscal scare, a politicised central bank, or a confiscation precedent would reprice the entire stack at once. The level is comfortable. The derivative is the risk, which is why this page watches the quarterly change and its decomposition, not the headline share, and treats any quarter where the non-valuation move exceeds a point as news1.
The parallel plumbing needs the same watch, because shares can hold while dependence quietly falls. Bilateral trade settled in local currencies, China's CIPS clearing alongside SWIFT, Gulf oil invoiced experimentally in renminbi, central bank digital currency corridors: none of it dents COFER today, all of it builds the optionality that would let a future crisis reroute around the dollar faster than the reserve statistics imply. Dedollarization, where it is real, is not a portfolio decision. It is infrastructure construction.
For the developing world, the variable is not academic. Reserve composition sets the terms on which every developing country borrows, holds its buffers, and weathers the Fed's cycles; commodity exporters price cobalt, phosphate, and oil in dollars and absorb the volatility of a currency they do not issue; and remittance flows cross borders on rails the sanctions architecture can reach. A world drifting toward the diversified shelf would be, for most of them, a world of marginally more expensive plumbing and marginally more monetary sovereignty, a trade each treasury prices differently, and several, judging by the gold in their vaults, have already begun paying for quietly2.
A note on what this page deliberately excludes: cryptocurrency, despite its dedollarization marketing, appears nowhere in the reserve data at scale, and stablecoins, the crypto instrument with genuine cross-border volume, are dollar liabilities, extending the currency's reach into exactly the informal corridors the reserve statistics never counted. If anything in the digital-asset world belongs on a dedollarization watchlist, it is the day a major stablecoin redenominates, and that day is on no visible horizon. The variable lives in the vaults and the COFER table, not the tickers.
What would change our reading
The gold streak's 2025 easing, 863 tonnes after three years above a thousand2, is itself a reading worth interpreting carefully rather than banking. Part is price: reserve managers are buyers with budgets, and a metal up 50% in a year3 buys fewer tonnes per billion. Part is arithmetic success: the vaults' gold share has risen enough, through purchase and revaluation together, that some banks' diversification targets are simply being met. Neither reading implies the insurance demand ended; the survey evidence, central banks polled about future purchases, remains near record intent. The series to watch is tonnes, but the signal is the share of fresh reserves going to the shelf instead of the pyramid.
In the section's format, the pre-committed markers. Erosion-as-usual: the dollar share drifting a fraction of a point per year, gold buying at hundreds of tonnes, no single quarter's structural move above a point; that is the base case and it is compatible with decades more dollar primacy. Acceleration: two consecutive quarters of non-valuation decline above a point, a return to 1,000-tonne gold years, or a major ally's reserve rebalancing announced rather than leaked; any of those and this page upgrades the variable from watched to urgent. Reversal: a share recovering past 60% on genuine inflows, which the status-quo camp should want as its falsifiable claim. The quarterly table arrives with a lag and no drama, and it is the closest thing the international system has to a slow-motion referendum on its own foundations. Empires have historically learned their monetary standing changed some years after it had; the table's readers get to learn it quarterly, with decimals. We will read each instalment here1, and file it beside the gold ledger, where the same referendum is conducted in tonnes, one central bank vault inventory at a time.
IMF, Currency Composition of Official Foreign Exchange Reserves (COFER), and IMF blog, Dollar's share of reserves held steady in second quarter when adjusted for FX moves: 56.32% in Q2 2025 with 92% of the decline from valuation effects; 56.9% reported for Q3 2025.
World Gold Council, Gold Demand Trends: central bank net purchases above 1,000 tonnes in 2022, 2023, and 2024, with 1,045 tonnes in 2024 and 863 tonnes in 2025.
CNBC, Gold price reaches $4,000 an ounce for the first time ever (7 October 2025): futures closing above $4,004; up about 50% in a year as the dollar index fell 10%.



