Why Central Banks Buy Gold, and What It Does

Central banks buy gold because it is the only major reserve asset that is nobody else's liability. A dollar reserve is a claim on the United States. A euro reserve is a claim on the euro system. Gold in your own vault is a claim on no one, and it cannot be frozen by the country that issued it.

That is the whole thesis, and the buying data since 2022 reads as a straightforward expression of it.

What the 2026 numbers actually show

Per the World Gold Council's Gold Demand Trends for Q2 2026, central banks bought 289 tonnes in the second quarter. That is a record high for a second quarter, and roughly five times the revised 57 tonnes recorded in Q1.

The first-half total was 345 tonnes — the weakest first half since 2022, when it was 241 tonnes. Those two facts sit oddly together until you look at what happened in Q1: sales by Türkiye, Russia and Azerbaijan pulled the net figure down sharply, masking continued buying elsewhere.

The largest individual buyers in Q2:

• Poland — 51 tonnes in Q2, 82 tonnes across the first half.
• China — 33 tonnes in Q2, 40 tonnes across the first half.
• Uzbekistan — 16 tonnes.
• Kazakhstan — 15 tonnes.

Net figures hide a lot. A quarter in which one bank sells 40 tonnes and six banks buy 10 tonnes each looks like modest net buying, but it describes two entirely different behaviours.

The four reasons reserve managers give

No counterparty

Every other reserve asset is somebody's promise. Gold held domestically is not, which means it cannot be defaulted on, devalued by an issuer, or frozen by a foreign authority. After 2022 demonstrated that reserves held in another country's currency can be immobilised, this stopped being an abstract argument.

Diversification that actually diversifies

Gold's correlation with major sovereign bonds and equities is low and often turns negative in crises. For a reserve portfolio whose job is to work in bad conditions rather than to maximise return, that property is the point.

Inflation and debasement

The historical record is messier than the slogan. Gold has held purchasing power over very long horizons but has tracked measured inflation poorly over shorter ones, and it went nowhere in real terms for long stretches. Our gold vs inflation comparison charts the real return rather than the nominal one, which makes the gaps visible.

Reducing dollar dependence

Several of the most persistent buyers are countries with a strategic reason to hold fewer dollar assets. Gold is the only liquid, deep, universally accepted reserve asset available for that substitution, which is less an endorsement of gold than an absence of alternatives.

Why it matters for the price

Central bank demand affects the gold market differently from investment demand, for three reasons.

1. It is price-insensitive. A reserve manager executing a diversification mandate buys to a target allocation, not to a valuation. That demand does not withdraw when the price rises.
2. It removes supply for a long time. Reserve gold is rarely traded actively. Metal that moves into official holdings mostly leaves the float.
3. It is slow and publicly reported. Monthly IMF and national data make the flows visible, which lets other participants position around them.

The caveat is that central bank buying is a large influence, not the only one. Jewellery demand, ETF flows, mine supply, recycling and real interest rates all act on the same price. Attributing a given move to one of them is usually a story told after the fact.

How this interacts with the rest of the market

Official-sector demand lands on a supply base that barely moves, which is what gives a few hundred tonnes such leverage on the price.

Mine supply is slow by construction

A new gold mine takes the better part of a decade to move from discovery through permitting to first pour. Production therefore cannot respond to a price signal within the timeframe that price signal lasts. When demand rises, the supply curve does not swing out to meet it — the price does the adjusting instead.

Recycling responds, but imperfectly

The one genuinely price-elastic source is scrap. Higher prices pull jewellery and old holdings back into refineries, which cushions rallies. It is a real buffer, but a partial one: much of the world's gold is held for reasons that a price move does not override, so the response is smaller than the theory suggests.

Investment demand can move the other way

ETF flows are the most volatile component and can offset official buying entirely over a quarter. A period of heavy central bank accumulation alongside heavy ETF redemptions can net out to a flat price, which is one reason the headline tonnage figure so often fails to explain what the chart did.

The practical consequence: treat central bank demand as a slow floor under the market rather than as a timing signal. It explains why the long-run picture looks the way it does, not why last week moved.

What the figures do not tell you

Worth being blunt about the limits of this data, because it gets over-read.

• Reporting is voluntary and lagged. Not every purchase is disclosed promptly, and some are not disclosed at all. Estimates get revised substantially — Q1 2026 is an example.
• Tonnage is not conviction. A 50-tonne purchase by a large reserve holder may be a smaller allocation shift than a 5-tonne purchase by a small one.
• It says nothing about your time horizon. A reserve manager holds for decades with no redemption pressure. That is not most people's situation.
• It is not a signal to trade on. By the time a purchase is reported, the market has seen it too.

If the interest is in how gold has actually performed against other assets rather than in why institutions hold it, gold vs S&P 500 and gold vs Bitcoin chart twenty years and fifteen years respectively.

Where these numbers come from

Central bank holdings data originates with the institutions themselves and with IMF reporting, and is compiled quarterly by the World Gold Council. Pricing is referenced to the LBMA Gold Price, set twice each London business day by auction.