Central banks have been buying gold at a pace not seen in generations, adding roughly a thousand tonnes a year to official reserves for three consecutive years, while the gold price set repeated records through 2025. Retail commentary treats each new high as a surprise, usually accompanied by a debate about whether the move has gone too far. The institutions doing the largest share of the buying do not appear surprised at all, and they are not buying because of the price chart. It is worth taking seriously what they are actually responding to, because the logic scales down to a private portfolio more directly than most investors assume.
The buyers who do not flinch
Reserve manager surveys point to two motives that have strengthened rather than faded. The first is diversification away from dependence on any single reserve currency, a slow-moving concern that predates the current cycle. The second arrived abruptly in 2022, when Western governments froze a large share of Russia’s foreign reserves. Whatever one thinks of that decision politically, every reserve manager on earth absorbed the same technical lesson: assets held inside someone else’s financial system carry a counterparty and a jurisdiction, and both can be switched off. Gold held in your own vault cannot. Central banks have been converting that lesson into tonnage ever since.
For a private investor the relevant translation is not geopolitics but counterparty exposure. Most gold held privately is not gold at all in the legal sense. It is paper with gold-shaped price behaviour: exchange traded funds, unallocated accounts, contracts for difference. Each structure inserts at least one institution between the owner and the metal. That is perfectly fine for trading exposure, and often the cheapest way to get it. It is a different instrument from owning gold bars and coins outright, where the position is property rather than a claim, and the distinction only matters on the days when it matters enormously.
Paper gold and the thing itself
The gradations are worth spelling out, because they get blurred in marketing. An ETF holder owns shares in a trust that owns gold, with redemption mechanics designed for institutions rather than individuals. An unallocated account holder is, legally, an unsecured creditor of the bank that runs the account. An allocated holder or a physical owner has title to specific metal. In calm markets these produce identical price exposure and the differences look academic. In stressed markets they are different legal positions with different queues, which is precisely the scenario a gold allocation exists to address. Buying the exposure without the property right solves a portfolio construction problem while leaving the original problem intact.
The standard objection is that these differences never actually bite, and it is half true. Most investors will go a whole career without the legal distinction between a claim and a title ever costing them anything. But the episodes when it has mattered follow a pattern: they arrive during exactly the market conditions that prompted the gold purchase in the first place, and they arrive without notice. Redemption terms tighten, unallocated accounts get settled in cash rather than metal, and the fine print that was academic in the good years becomes the whole story. Insurance that fails only in the specific weather it was bought for is not obviously worth its premium.
Physical ownership does carry costs that paper conveniently hides, and pretending otherwise is how new buyers get disappointed. Bars and coins trade at a premium over the spot price that covers fabrication and distribution, and the spread between buying and selling narrows as sizes increase. Product selection matters more than most first-time buyers expect, because resale liquidity is determined at the moment of purchase: recognised refiner bars and sovereign mint coins sell easily anywhere, while obscure products do not. Storage is the other honest cost, since home insurance policies typically cap precious metals well below a meaningful allocation, which is why allocated vault storage has become the default for anything beyond a modest holding.
A few disciplines separate a deliberate allocation from an impulse purchase:
- Decide the allocation percentage before looking at the price chart, and rebalance to it rather than to headlines
- Buy recognised refiner bars or sovereign mint coins, since resale liquidity is fixed at the point of purchase
- Treat premiums, spreads and storage as the true cost of ownership rather than judging everything against spot
Reading the record highs
On the price itself, the composition of demand matters more than the level. World Gold Council data shows official sector buying running near historic highs alongside renewed investment demand, and central banks are price-insensitive buyers with multi-decade horizons. That does not make any price sustainable, and gold remains an asset that pays nothing and can go sideways for a decade. What it does mean is that a structural buyer has entered the market with motives unrelated to yield or momentum, and structural buyers change how drawdowns behave.
The sizing logic follows from the job description. Gold is not a return engine and treating it as one leads to overallocation at exactly the wrong moments. Its historical role is as the asset that behaves differently when confidence in the payers of every other asset is being questioned, which argues for a small, permanent, boring allocation rather than a conviction trade. Single digit percentages of a portfolio capture most of the diversification benefit, and anything much larger starts to be a macro bet dressed as insurance.
The market structure around private ownership has matured to meet this demand. Dealers that pair bullion sales with allocated vault storage, such as New Zealand’s Commonwealth Vault, reflect where private buying has actually gone: investors who want to own the metal outright without taking on the risks of keeping it at home. That model, dealer and custodian under one roof, is essentially a retail-scale version of what central banks have always done.
Central banks are not clairvoyant, and a thousand tonnes a year is not a price target. What they model more carefully than most private investors is the difference between exposure and ownership, and their behaviour over the past three years suggests they consider that difference worth paying for. Private investors relearn the same distinction in every crisis. The cheaper approach is to learn it from the buyers who never forgot it.