Central Banks Are Buying: Three Straight 1,000-Tonne Years Are Rewiring Reserve Portfolios Away From the Dollar

The largest buyer in the gold market is no longer a person, a fund, or a jeweler: it is the official sector. Central banks net-purchased 1,082 tonnes of gold in 2022, 1,037 tonnes in 2023, and roughly 1,045 tonnes in 2024 [VERIFY: World Gold Council, Gold Demand Trends], three consecutive record-class years running at roughly double the prior decade's pace [VERIFY: WGC annual series]. The bid is not cyclical portfolio tinkering. It is a structural re-rating of gold's role as a store of value that no government can freeze, triggered by the February 2022 immobilization of roughly $300 billion in Russian reserves [VERIFY: G7/EU official statements] and sustained by a slow, measurable drift of reserve portfolios away from the dollar.

The scale of the bid

The magnitude is easiest to see against history. Across 2010-2019, central banks added an average of roughly 500 tonnes a year [VERIFY: WGC annual demand series]; since 2022 the run rate has doubled, and official buying has absorbed roughly one-fifth of total annual gold demand [VERIFY: WGC demand-by-sector tables]. That share matters because the official sector is the least price-sensitive buyer in the market: it accumulates for balance-sheet reasons, not for return, and it reports its holdings as part of foreign-exchange reserves rather than marking a P&L. A fifth of demand arriving from buyers who do not care about the price is the kind of structural floor that changes how a market trades.

The direction of travel flipped 16 years ago. In 2010 the official sector became a net buyer for the first time in roughly two decades [VERIFY: WGC/IMF IFS], ending an era in which European central banks sold 400-500 tonnes a year under the coordinated Central Bank Gold Agreements [VERIFY: ECB/CBGA records]. The 2022 sanctions shock then kinked the trend line: annual net purchases jumped from 450 tonnes in 2021 to a record 1,082 in 2022 [VERIFY: WGC] and have not returned to the old range since. The price has followed the flows, setting a record above $3,500 per ounce in April 2025 [VERIFY: LBMA/COMEX price history] and holding near $3,400 in mid-2026 [VERIFY: LBMA spot data].

The forward indicator is intent, and intent is loud. In the World Gold Council's 2025 reserve-manager survey, 95% of responding central banks expected global official gold holdings to keep rising over the next 12 months, and 43% planned to increase their own [VERIFY: WGC Central Bank Gold Reserves Survey 2025], the highest reading in the survey's history. No respondent planned to reduce holdings. For a cohort whose defining professional trait is caution, that is close to a unanimous verdict on the asset's role in the next decade.

Who is buying

China is the anchor buyer. The People's Bank of China reported 18 consecutive monthly additions from November 2022 through April 2024, roughly 316 tonnes [VERIFY: PBoC/SAFE disclosures], paused for six months, then resumed reported buying in November 2024 [VERIFY: PBoC]. Reported holdings sit near 2,280 tonnes [VERIFY: IMF IFS/PBoC], yet that is only about 6% of China's roughly $3.2 trillion reserve stack [VERIFY: WGC, SAFE]. The context is a currency that remains hemmed in by capital controls: a state that cannot yet offer the world a convertible reserve currency can still convert its own reserves into the one asset that requires no one's permission to hold.

Poland is the conviction buyer. The National Bank of Poland bought roughly 130 tonnes in 2023, the largest single-bank purchase that year, added about 90 tonnes in 2024 and roughly 62 more in 2025 [VERIFY: NBP/WGC], lifting holdings to roughly 448 tonnes, around a fifth of its reserves [VERIFY: NBP disclosures]. Governor Adam Glapinski has made the strategy explicit, stating a target of 20% of reserves in gold [VERIFY: NBP official statements] and framing bullion as the asset that survives infrastructure-level shocks to the financial system. A NATO frontline state buying gold at the fastest pace in Europe is not a yield decision; it is a security decision.

The rest of the leaderboard is emerging-market breadth. Turkey bought a record 148 tonnes in 2022, dumped much of it into domestic retail demand during the 2023 lira squeeze (net additions collapsed to roughly 11 tonnes), then resumed with roughly 75 tonnes in 2024 [VERIFY: WGC/CBRT]. India's Reserve Bank added roughly 73 tonnes in 2024, its fastest pace in years [VERIFY: RBI/WGC], and separately moved about 100 tonnes of existing holdings home from Bank of England vaults [VERIFY: RBI annual report], a reminder that custody geography is now part of reserve strategy. Gulf authorities led by Qatar and the UAE added a combined 138 tonnes across 2022-2025 [VERIFY: WGC/IMF IFS], steadily recycling energy surpluses into metal.

Gold will retain its value even when someone cuts off the power to the global financial system.

— Adam Glapinski, President, National Bank of Poland [VERIFY: NBP press statements, 2021]

The dollar question

The accumulation cycle is inseparable from the slow erosion of the dollar's reserve dominance. The dollar's share of allocated global FX reserves sits near 58%, down from roughly 71% in 1999 [VERIFY: IMF COFER], a two-decade drift rather than a collapse. What the COFER data shows is diversification at the margin: the lost share has gone not to the euro or the yen but to gold and a long tail of smaller currencies. Reserve currency status is not being transferred to a challenger; it is being partially unbundled, with gold absorbing the piece of the reserve function that is about trust rather than transactions. Ezana Echo's five-century empire-rankings study reached the same fork: the fragmentation scenario, with reserves splintering across dollar, euro, gold, and smaller units, is the one that record official gold buying already gestures toward.

The proximate catalyst has a date. In February 2022, G7 measures immobilized roughly $300 billion of Russia's central-bank assets [VERIFY: G7/EU official statements], demonstrating overnight that a reserve portfolio held in another sovereign's securities is, in the limit, a claim that sovereign can suspend. For the roughly 100 central banks outside the sanctioning coalition, the lesson generalized far beyond Russia: exposure to secondary sanctions risk is now a line item in reserve management. Gold vaulted at home carries no such clause. It is the only reserve asset of scale that is simultaneously nobody's liability and nobody's jurisdiction.

The result is a quiet milestone. Valued at market prices, gold now represents roughly 20% of global international reserves, and in 2024 it passed the euro, at roughly 16%, to become the world's second-largest reserve asset [VERIFY: ECB, The International Role of the Euro, 2025 edition]. That inversion happened without a single summit or treaty. It is the aggregate of dozens of independent balance-sheet decisions, which is precisely why it is unlikely to reverse quickly: there is no coordinating body whose policy change could undo it.

Share versus pace: reading the map

The most useful way to read the landscape is two variables at once: how much of a country's reserves are already in gold, and how fast it is buying. The legacy holders, the United States at roughly 72% and Germany at roughly 71% of reserves in gold [VERIFY: WGC/IMF reserve-share data], are not buying at all; their allocations are artifacts of Bretton Woods. The active buyers are mostly low-share, high-pace: China's 6% share against a roughly $3.2 trillion reserve base means each single percentage point shifted into gold is roughly $32 billion, or about 290 tonnes at $3,400 an ounce [VERIFY: computed from SAFE and LBMA data], nearly a third of a record global year from one policy decision. That asymmetry between where shares are and where buying is happening is the strongest structural argument that the cycle has years to run.

One caveat belongs in every reading of this data: the reported numbers are a floor, not a census. The World Gold Council's net-purchase estimates, built with trade and vault data, run persistently above the flows central banks report to the IMF [VERIFY: WGC/Metals Focus methodology notes], implying a meaningful block of unreported official buying. Analysts widely estimate China's true holdings exceed the published figure [VERIFY: named analyst estimates]. In the language of market data, the IMF series is a lagging indicator of a process whose leading edge is deliberately opaque. The honest conclusion is directional: actual official demand is at least what the headline says, and probably more.

How to trade the official bid

For investors who want the same exposure the reserve managers are buying, the instruments sort into three tiers. Bullion-backed ETFs are the direct route: GLD is the liquidity benchmark at a 0.40% expense ratio, while IAU tracks the same metal at 0.25% [VERIFY: SPDR/iShares fund documents], with both settling against the London OTC market and COMEX-linked pricing. Miners add operating leverage: GDX bundles the sector, and its largest holdings, Newmont (NEM), Agnico Eagle (AEM), and Barrick (GOLD) [VERIFY: VanEck factsheet], earn the spread between a roughly $3,400 spot price and all-in sustaining costs averaging near $1,500 an ounce [VERIFY: company reports], margins that a sustained official bid effectively underwrites. The third tier, royalty and streaming names Wheaton (WPM) and Franco-Nevada (FNV), buys gold-linked cash flow without direct mine cost exposure [VERIFY: WPM/FNV filings], historically the highest-margin corner of a commodity upcycle.

The risks are specific and worth naming. Official demand can normalize: a fourth 1,000-tonne year is the base case, not a law of nature, and a de-escalation of great-power frictions would soften the security premium. Individual buyers can flip to sellers overnight, as Turkey did in 2023 when domestic crisis demand made the central bank a supplier [VERIFY: WGC/CBRT]. And gold pays nothing: a sharp rise in real rates restores the opportunity cost that the 2022-2025 regime suppressed. Position sizing, not conviction, is the correct response to a thesis whose largest actors publish their intentions in surveys but not their execution.

The outlook

The base case is continuation: 900 to 1,100 tonnes of official demand in 2026 [VERIFY: WGC outlook/sell-side consensus], a dollar COFER share drifting lower by fractions of a point per year, and gold's reserve share compounding as both flows and price work in the same direction. The bear case requires either a real-rate shock or a credible restoration of trust in sanctioned-asset protections, and nothing on the diplomatic calendar currently supplies the latter. In the arc of the long-term debt cycle, reserve managers are behaving exactly as the playbook predicts late in the sequence: quietly swapping claims on the incumbent for the asset that predates every incumbent. Ezana users can track the two numbers that adjudicate this thesis, quarterly COFER shares and WGC demand prints, from the platform's macro dashboards, and stress-test bullion and miner baskets against rate scenarios in the For The Quants workbench. The dataset is public, quarterly, and free; the discipline is in watching it.