On 7 September 2026 gold passed the inflation-adjusted level of its January 1980 peak. That is a genuinely rare event. The last time gold cleared that bar, most people reading this had not been born.
The explanations you will see for it are usually one line long: inflation, or fear, or the dollar. None of those survives contact with the data on its own. Gold rose through years of low inflation and fell through years of high inflation. It has rallied in risk-on markets and sold off in panics. A single-cause story is comforting and wrong.
What actually sets the gold price is five forces pulling at once, and at any given time two of them are doing most of the work. This article explains each one, how to watch it yourself, and which of them is currently dominant.
This is the change that separates the current era from the previous one.
Central banks were net sellers of gold for most of the 1990s and 2000s. They became consistent net buyers after 2010, and then the pace changed again: purchases jumped above 1,000 tonnes a year in 2022 and stayed historically elevated. In 2025 the official sector bought 863 tonnes. The World Gold Council's expectation for 2026 is around 850 tonnes.
Two features make this demand different from everything else on this list.
It is price insensitive. A reserve manager diversifying out of a single currency is not trying to buy the dip. They buy on a schedule, through the price, and they are generally not going to sell into a rally.
It removes metal from circulation more or less permanently. Central bank holdings are roughly 17% of all above-ground gold and they turn over slowly. Metal that moves into official reserves is metal the market no longer trades.
The National Bank of Poland was the largest single buyer in 2025 for the second year running, adding 102 tonnes. China and Kazakhstan stayed active in 2026, and countries including Indonesia and Malaysia returned as buyers after long absences. The full mechanics of this, and what it says about the dollar, are covered in our guide to central bank gold buying.
The textbook model of gold is one sentence: gold pays no income, so the higher the real return on a safe alternative, the less attractive gold becomes.
That model works. It explains 2013, when real yields rose and gold fell hard. It explains 2019 to 2020, when real yields went deeply negative and gold set records.
It stopped explaining things in 2022. Real yields rose sharply and gold did not collapse. It held, then advanced. The most persuasive account of why is on the previous point: a buyer who does not care about the opportunity cost of yield arrived in size and changed the marginal bid.
Do not read that as "real rates no longer matter." Read it as "real rates are one input into a market that acquired a second large one." When real yields fall from here, they push in gold's favour. They are simply no longer the only hand on the wheel.
How to watch it yourself: the 10-year Treasury Inflation-Protected Securities yield, published daily by the US Treasury, is the standard proxy. You do not need a model. You need the direction.
Where central banks are the slow structural bid, investment flows are the fast one, and they are what makes gold move in weeks rather than years.
The 2026 numbers are large. Collective gold ETF holdings reached an all-time high of 4,176 tonnes on 27 February 2026. They eased back to roughly 4,068 tonnes by late August, and global assets under management stood at about US$530 billion in July. Year to date through August, inflows totalled US$29 billion, equal to about 160 tonnes.
The important thing about this demand is that it is reversible. An ETF holder can sell tomorrow. A central bank generally will not. When you see a sharp gold drawdown in an otherwise strong year, this is usually the flow that caused it.
The cost side of ETF ownership is a separate question and a material one over long holding periods. We ran the arithmetic in Gold ETFs vs Tokenized Gold.
Gold is quoted in dollars, so mechanically a weaker dollar lifts the dollar price of gold without anything happening to gold itself. That is arithmetic, not insight.
The deeper version matters more. A reserve manager holding a large single-currency position faces two risks that have nothing to do with exchange rates: the risk that the issuing country's fiscal path erodes the value of the asset, and the risk that the asset can be frozen. Gold held domestically answers the second risk in a way no financial asset does, because it is nobody's liability. That is the argument that has been running through official-sector behaviour since 2022, and it is why the buying has been geographically broad rather than concentrated.
What to watch: the dollar index for the mechanical effect, and quarterly IMF and World Gold Council reserve data for the structural one.
Every other commodity market has a release valve: high prices bring on new supply, which caps the price. Gold's valve is rusted shut.
World mine production was an estimated 3,300 tonnes in 2025, against 3,280 tonnes in 2024. That is roughly 1% growth in a year when prices were setting records. Discovery to first pour on a large deposit routinely takes a decade or more, and permitting is slower now than it was a generation ago.
Recycling is more responsive, and high prices do pull old jewellery back into the refining stream. But recycled supply is finite and behaviourally sticky: people part with heirlooms slowly.
Meanwhile the cost of producing an ounce keeps climbing. Global average all-in sustaining costs rose to about US$1,785 an ounce in the first quarter of 2026, the latest in a long run of year-on-year increases. What that does to mining equities, which is a different trade from owning metal, is covered in Gold Mining Stocks vs Gold.
The full supply picture, including how much metal is actually left in the ground, is in How Much Gold Is Left.
Weighing the five as of September 2026:
The structural drivers are pointing one way and the reversible one is currently pointing the same way. That combination is what produces a record. It is also the reason a sharp pullback is entirely possible without the structural story changing at all: the fast money can leave while the slow money stays.
It does not tell you where the price goes next. Every driver above is observable in the past tense. Anyone extrapolating them forward with a target price is making a forecast, and forecasts about gold have an unimpressive record.
It does not tell you how much to own. That is a portfolio question, not a price question, and we treat it separately in How Much Gold Should You Own.
It does not tell you what form to hold. Bars, coins, funds and tokens give you the same price exposure with very different costs, custody arrangements and failure modes. What Is a Gold Token covers the on-chain end of that choice.
One thing is worth saying plainly, because it is the uncomfortable part of every bullish gold case: gold produces nothing. A rising price is the only return physical gold has ever offered, and storage costs run the other way. Anything that changes that arithmetic, including staking arrangements, deserves scrutiny of where the income actually comes from rather than excitement about the number. We are direct about that in Is Gold Staking Safe.
Why is gold going up in 2026? Primarily because two slow structural buyers, central banks and reserve diversifiers, met a fast cyclical one, investment funds, in a market whose supply grows about 1% a year. Central banks bought 863 tonnes in 2025 and ETF holdings set a record in February 2026.
Does gold always rise when inflation rises? No. Gold performed extraordinarily in the inflationary 1970s and then lost value in real terms for roughly two decades while inflation was moderate. The relationship is real but unreliable, and we set out the decade-by-decade record in Is Gold Actually an Inflation Hedge.
What is the single best indicator to watch? There is no single one. If forced to choose two: the 10-year real yield for the cyclical signal and quarterly central bank purchase data for the structural one.
Can the gold price fall from a record high? Yes, and it has done so repeatedly, including a multi-year decline after the 2011 peak. Records are not floors.
Does a strong dollar mean gold must fall? It creates a headwind, because gold is priced in dollars. It is not a rule. Gold has risen alongside a firm dollar when other drivers were strong enough.
Do higher mining costs push the gold price up? Not directly. Costs influence how much supply eventually arrives, which matters over years. They do not set the daily price.
Is tokenized gold exposed to the same drivers? Yes. A gold token tracks the metal price, so the five forces above apply in full. What tokenization changes is access, settlement and custody, not price exposure.
Gold is not rising for one reason, and the accounts that say otherwise are generally selling something. What has happened is that a market with effectively fixed supply acquired a large, price-insensitive, structurally motivated buyer, and then investment demand arrived on top.
The honest way to hold that view is to hold it loosely. Two of these five drivers can turn quickly. If you own gold, know which parts of the case are structural, which are cyclical, and what the position costs you to hold each year. That last figure is the one most holders have never calculated, and it is in What It Really Costs to Own Physical Gold.
This article is for informational purposes only and is not financial advice.
| Driver | Direction | Speed | How durable |
|---|
| Central bank buying | Supportive | Slow | Multi-year, policy-driven |
| Real interest rates | Mixed | Medium | Cyclical, can reverse quickly |
| ETF and investment flows | Supportive in 2026 | Fast | Reversible in days |
| Dollar and diversification | Supportive | Slow | Structural, geopolitical |
| Mine supply | Supportive | Very slow | Effectively fixed for years |
