How Is the Gold Price Formed? Four Layers Behind Every Move

Who actually quotes the price on your screen

Most people assume the gold price comes from some official body, or from the jewellery shop window. Neither is true.

The international price is formed in two places: the London over-the-counter spot market and COMEX gold futures in New York. Together they account for roughly 85% of global turnover and carry the job of price discovery. In London, large dealing banks quote each other bids and offers five days a week, around the clock, and institutional clients trade on those quotes. That is where the spot price comes from.

Sitting on top is a published benchmark: the LBMA Gold Price, set by electronic auction twice each business day at 10:30 and 15:00 London time. A great deal of physical contracting, ETF valuation and central bank settlement references it. Think of it as an official settlement print twice a day, not the live market.

COMEX is a different animal. Gold futures routinely trade more than 200,000 lots a day at 100 troy ounces each — notionally over 600 tonnes changing hands daily, more than a tenth of a full year of mine output, in a single session. Almost none of it goes to delivery; it is opinion and hedging. Arbitrage pins spot to the front month, usually within a few dollars (on how the two differ, see spot gold versus gold futures).

Which leads somewhere important for beginners: almost none of the money moving the gold price intends to take delivery. It is like a cinema selling a few hundred tickets into a room with a handful of chairs — perfectly fine in practice, because nearly everyone bought the ticket to resell it and never meant to sit down. So when gold drops 30 dollars in an hour, it is almost never someone selling tens of tonnes of bars. It is a book of derivative positions changing direction.

Gold files no earnings, so what are its fundamentals?

To analyse a stock you open the accounts: revenue, profit, cash flow. Gold has none of that. Most introductions skip straight past the problem to safe haven and inflation hedge, and the reader never gets a handle they can actually use.

Gold's fundamentals are a different thing entirely, simpler than a company's and far more counterintuitive: almost all the gold ever mined still exists. It is not burned, eaten or used up. Jewellery is melted and recast, bars sit for two centuries. The World Gold Council puts the above-ground stock at roughly 216,265 tonnes at the end of 2024; 2025 mine production was about 3,672 tonnes, a record. A full year of new supply is about 1.7% of the stock.

Those two numbers are worth turning into something you can see. Melt every ounce ever mined into a single block and you get a cube roughly 22 metres on a side — about a seven-storey building. A full year of work by every mine on earth is the small block beside it, under six metres.

So even a 10% jump in global mine output would move total supply by 0.17%. That is why a major new discovery barely registers in the gold price, while the same headline about lithium, nickel or natural gas moves those markets hard. Oil is burned and gone, wheat is eaten and gone — those are flow markets. Gold almost never leaves; it is a stock.

Which gives the real conclusion: the price is not set by how much is produced this year, but by whether the people already holding gold still want to hold it at this level. Each day's price is the number required to persuade the marginal holder not to sell, or the marginal buyer to step up.

That also fixes the role of physical demand: it is a cushion, not an engine. When the price falls, Asian jewellery and bar buying comes in and absorbs part of it; when the price rises, the same buyers purchase less. What actually pushes the price around within a few days is never the retail counter — it is institutional positions in London and New York.

The closest thing to company analysis is a gold miner

If you do want to apply company analysis to gold, the object is a mining company. Profit is roughly the gold price minus unit cost, which gives it natural leverage to the metal.

But a miner is not a proxy for gold

The worked example above shows the amplification: gold rises 10% and the margin rises 40%. Industry forecasts for 2026 put operating margins near 2,800 dollars an ounce, a record, for exactly this reason.

The same arithmetic shows where the risk lives. The numerator moves fast, but the denominator moves too. Higher energy and labour costs, falling ore grades, production misses — all of them run the amplifier in reverse. And that margin is not net income, still less a share price.

On top of which you have bought a set of risks that have nothing to do with gold: strikes, power shortages, a change of tax regime in the host country, management overpaying for an acquisition. A gold miner is a business connected to gold, not a bar that automatically multiplies its return. If your view is only about the metal, gold ETFs versus gold futures express it more cleanly.

Four pricing layers, and every move belongs to one of them

If the price is set by willingness to hold, the question becomes: what changes that willingness? The answer sorts into four layers that differ in speed, in what you watch, and in how they fail.

Layer 1 · Opportunity cost. Gold pays no interest, no dividend and no rent, so the real cost of holding it is the risk-free yield you gave up. What matters is not the nominal rate but the real rate — nominal minus expected inflation. The dollar belongs here too: gold is quoted in dollars, so a stronger dollar makes the same bar more expensive in every other currency and price-sensitive physical demand steps back.

Layer 2 · Reserve and allocation demand. Central banks were net sellers of gold before 2008 and net buyers since; from 2022 they bought more than 1,000 tonnes a year for several years running, and still took 244 tonnes in the first quarter of 2026, above the five-year average. They are buying the safety of the reserve mix, not a yield, so they are relatively price-insensitive — they buy when it rises and buy more when it falls. This layer does not create short-term volatility; it lifts the whole level.

Layer 3 · Flows and positioning. Gold ETFs and COMEX futures positioning show which way leveraged money is leaning and how crowded it has become, and they are the main source of short-term direction. Crowded positioning makes the market fragile: a modest piece of contrary news can be amplified into a rout by concentrated unwinding (on reading this data, see how to read the COT report).

Layer 4 · Events and microstructure. Data releases, central bank remarks, geopolitics, stop-loss cascades, thin holiday liquidity. The moves can be large, but the direction often has nothing to do with fundamentals. The value of understanding this layer is not prediction — it is not mistaking it for a signal from the other three.

Layer 3 comes with a data trap worth clearing up first

Shares in a physically backed gold ETF are, most of the time, traded between investors. You buy a share, someone else sells one, and not a gram of the fund's metal moves. Only when an authorised participant creates or redeems a basket does the fund's physical holding actually change.

That distinction decides which number you should be reading. Tonnes held reflects the real change and is the actual measure of investment demand. Assets under management equals tonnes times price, so it is inflated by the price itself. Net flows strip the price effect out and are the most direct read on whether new money is arriving.

Global gold ETF holdings set a record of 4,176 tonnes on 27 February 2026 with assets peaking around 701 billion dollars, then eased to 4,068 tonnes and 530 billion by July 2026. May 2026 alone saw two billion dollars of net outflows (Asia −1.2bn, North America −1.1bn, Europe +334mn) — and that was precisely when the price was falling fastest. Holdings and flows turning together is the real signal.

The horizon decides which layer is speaking

Ask what gold does today and ask what it does next year, and you are not looking at the same variables. This is not a difference in depth of analysis — the dominant variable itself is different.

It also explains something that confuses a lot of readers: a weekly forecast can say bearish and an annual one bullish, and both can be right. They are not answering the same question. The weekly one says opportunity cost and positioning are against gold for a few days; the annual one says the floor is still rising. Treating them as contradictions is the most common mistake people make reading forecasts.

Real rates, in plain language

Put an ounce of gold in a safe and take it out a century later and you have an ounce of gold. It pays no dividend, no interest, no rent. That is the most basic thing to understand about it.

Two hens make it easier to hold in your head. One is the Treasury hen, and how many eggs it lays each day is the interest rate. The other is the gold hen, feathers of solid gold, which never lays anything. When rates are high the Treasury hen lays five eggs a day, everyone feeds it, and nobody has time for a bird that lays nothing. But once prices rise faster than the eggs accumulate — same number of eggs, less they will buy — people remember something: the gold hen may lay nothing, but her feathers do not rot.

In numbers: on the same 10,000 dollars, 3% pays 300 a year and 5% pays 500. That extra 200 is what makes a yielding asset attractive, and choosing gold means giving it up. That is opportunity cost — gold did not charge you 200 dollars, it cost you 200 dollars of foregone income.

One technical note: to watch real rates the market uses inflation-protected government bond yields. Do not take a ten-year nominal yield and subtract last month's CPI — the maturities do not match, and realised inflation is not expected inflation.

Why inflation can lift gold and crush it

The same inflation print starts two paths running in opposite directions. One is preservation: purchasing power is eroding, so I want to hold some gold. That adds demand. The other is the rate path: inflation this high means the central bank keeps rates higher for longer. That raises the cost of holding gold.

Which way the price goes depends on which force is stronger. Inflation is not a switch for gold; it is a variable acting on both the numerator and the denominator at once. In 2022 the run of monthly declines showed the rate pressure, while a slightly positive full year showed other demand offsetting it.

The price is in dollars; your return is not

Everything above is the dollar price. Unless you earn and spend in dollars, your actual result passes through two more gates: the exchange rate and the spread you pay to get in and out.

Ignoring costs, the conversion is simply local price per gram ≈ dollar price per ounce × the USD rate against your currency ÷ 31.1035. That 31.1035 is the number of grams in a troy ounce — gold uses troy ounces, not the avoirdupois ounce of about 28.35 grams used for everyday goods. The two differ by roughly 10%, and mixing them up produces a nonsense number.

Use your own pair, not the dollar index

In the example above the dollar price falls 5% while the dollar gains 3% against your currency, so what you hold becomes 95% × 103% = 97.85% of its former value — a fall of about 2.15%. The exchange rate absorbed more than half the decline. It works the other way too: gold up 5% while your currency strengthens 3% leaves you with roughly +1.85%. You never hold an abstract international gold return. You hold gold priced in one currency, in one product form.

A common error here is reaching for the dollar index. DXY is a basket weighted heavily toward the euro; for most readers it is only loosely related to their own currency, and for some currencies it contains no exposure at all. Use the actual pair.

The second gate is the spread, and it varies enormously by product. A futures or ETF position is the tightest and is priced off the wholesale market. Bank passbook or vaulted gold posts a buy and a sell price with a fixed gap, which suits steady accumulation but punishes short round trips. Coins and small bars carry a wider spread and a fabrication premium. Jewellery is the most expensive of all, because you are also paying for craft and design, and you get almost none of that back when you sell — jewellery is a consumer good that happens to contain gold, not an efficient way to own it.

There is a third gate in some markets: local premiums. Import duties, quotas and shipping costs mean that onshore prices in India or China can sit meaningfully above or below the international price converted at spot, and those gaps move on their own schedule. In practice, the channel you choose decides how far the price has to move before you break even — a separate question from where the international price is going, and both decide whether you make money.

What technical analysis is actually good for in gold

The conclusion first: technical analysis is not a tool for predicting the future. It is a map of where other participants will change their minds.

Gold has a peculiarity here. A stock has earnings and cash flow, so even a rough valuation gives you an anchor. Gold has no internal cash flow, so every participant ends up referring to the price itself — prior highs and lows, round numbers, moving averages. The more people watch the same set of levels, the more those levels become real behavioural boundaries. That is not mysticism; it is a self-fulfilling coordination mechanism.

Which is why 2,000, 2,500 and 3,000 are never merely memorable numbers. They are trenches where money is dug in: stop orders pile up at the same prices, large option strikes cluster on round hundreds, and dealers holding those options must hedge into the market as price approaches, amplifying the swings right there.

In practice the technicals answer three questions. Where are the stops sitting — a mass of participants place them just under the same obvious levels, so touching one triggers them in bulk and the selling itself pushes price further. When does liquidity thin out — early Asian hours, Western holidays, the minutes before a release, when the people willing to quote step back and the same order size moves price much further, widening slippage. Is the break real — if volume and follow-through are missing after a level gives way, price often snaps back, which is a false breakout.

To judge whether a level still holds, the question is not whether you drew the line correctly but three behavioural ones: is anyone still buying here, is the selling thinning, and can the bounce continue? Once the bid disappears, the most textbook support fails (for the full test, see why support and resistance levels fail).

The limit is equally clear: technicals cannot stand against repricing in layers 1 and 2. When expected real rates are rewritten within an hour, no line on a chart matters. Technical analysis describes behavioural inertia while other things are held constant — and a macro release is precisely the moment those other things all change at once.

Why war does not reliably lift gold

Gold is a safe-haven asset, but safe haven does not mean every piece of bad news lifts it. Triaging a shock with three questions beats memorising conclusions.

One: does this threaten money or the financial system itself? Bank crises, sovereign default risk, reserves being frozen — these strike directly at the reliability of paper claims, and gold's entire proposition is that it depends on nobody's credit. This channel is strongly bullish and can run for years.

Two: does this change the inflation and rate path? Many geopolitical events push energy prices up, energy pushes inflation up, inflation forces central banks to hold or raise rates, and higher real rates suppress gold. The direction here is ambiguous and frequently negative. This is where most intuitive readings of geopolitics go wrong.

Three: does it only affect physical supply or demand somewhere? A strike in a producing country, an import tax, a shift in Indian wedding season demand — real effects, but tiny against the stock. Physical being hard to get somewhere and gold being worth more everywhere are different questions.

Triage gives you direction; duration is a separate question, and it is where most people who buy the headline get hurt. A shock usually runs in three stages. In the first hours to days, haven money sweeps the offer with market orders and the price goes vertical. Over the following days to weeks, the world fails to end, the economy keeps running, the risk premium bleeds out, early positions take profit, and much of the move is handed back. After that it depends on whether the shock converts — a contained incident returns to the old path; one that lifts energy prices and rewrites the rate path turns against gold; and only one that undermines the assumed safety of reserve assets sinks into layer 2 and lasts for years.

One more case deserves separating out: when a crisis gets bad enough that the question changes from what should I hold to how do I raise cash today, the logic inverts.

Picture the desk that day. Equities are limit down with no bid. The bonds are quoted at levels nobody wants to look at. The broker is already on the phone asking for margin by the close, and behind that sits a queue of clients wanting their money back. Look around the book, and the only thing that will still sell, in size, near the price on the screen, is gold.

So the gold gets sold. The people selling it are not bearish — many of them want it back the next morning — they simply have to survive until then. People buy gold when they fear the future; they sell it when they must pay for today. On how risk premium itself gets priced, read the geopolitical risk premium.

Eight episodes, run through the same framework

Take the framework to real markets. Each case asks only two things: which layer was in charge? and which popular intuition did it break?

January 1980. Inflation out of control, revolution in Iran, Soviet troops in Afghanistan, and gold reaching 850 dollars an ounce with almost everyone convinced it would keep going. Then Volcker pushed rates to historic highs — the US prime rate touched 21% — inflation expectations were broken, real rates swung from deeply negative to sharply positive, and gold was below 400 dollars within two years, starting a twenty-year bear market. Layer 1 was in charge. What ended the bull was rates outrunning inflation.

Autumn 2008. Gold was near 1,011 dollars when Bear Stearns went down in March. After Lehman failed on 15 September it did not rally — it fell to 692.50 by October and November, about 30% off the high. The cause sat in layers 3 and 4: institutions facing margin calls and redemptions at the same time had to sell what could be sold, and gold was among the most liquid things they owned. Once the regime changed — quantitative easing, real rates negative — layer 1 took over and gold ran to 1,921 by September 2011. When gold and equities fall together, check first whether it is a liquidity event rather than hunting for a bearish story about gold.

2013. On 15 April gold fell 140.30 dollars in a session, about 9%, the worst day in three decades, and more than 200 dollars across two. The triggers were modest: a rumour that Cyprus might sell reserves, a Goldman recommendation to short, and expectations that QE would be tapered. What magnified it into that scale was layer 4 — stop cascades and forced ETF redemption. But the most instructive thing that year is in the annual data: global consumer demand rose 21% to 3,864 tonnes and bar and coin demand hit an all-time high, while gold ETFs shed 881 tonnes — and the price fell about 28% on the year. Chinese retail buyers queuing at jewellery counters made headlines worldwide, and the price ground down to around 1,050 dollars afterwards; anyone who bought into that queue waited years to break even. Split demand into jewellery, bars, central banks and funds — a photograph of a queue is not the market.

2020. Gold opened the year at 1,575, replayed the 2008 script in the March liquidity shock, then, as unprecedented stimulus drove real rates deeply negative, reached 2,072.50 in August — a record at the time and a gain of about 32% in eight months. One crisis, traded in stages; you cannot run the first week's logic all year.

2022. US CPI hit 9.1% in June, a forty-year high. Gold did not simply climb: it fell for six consecutive months and dipped near 1,671.8 at the end of September. The Federal Reserve's own review puts the cumulative 2022 tightening at 425 basis points. On the USD LBMA PM benchmark the World Gold Council uses, gold finished the year up about 0.4%. Layer 1 — rates and the dollar — supplied the pressure, and central bank and retail demand offset it. The lesson: high inflation does not mean a one-way climb, and a full-year return does not describe what holding it felt like.

2022 to 2025. Real rates sat at decade highs and gold kept climbing anyway, and the difference came from layer 2. After some countries' reserves were frozen in a geopolitical conflict, reserve managers reassessed what counts as a safe asset; central bank buying ran above 1,000 tonnes a year, which amounts to laying a new floor underneath the old price structure. The 2024 demand mix shows the change of cast most clearly: the LBMA afternoon benchmark set 40 record highs that year, investment demand rose 25% to 1,180 tonnes, a four-year high, while jewellery volume fell 11% to 1,877 tonnes — and jewellery spend rose 9% to 144 billion dollars. Prices were high enough that ordinary buyers bought less metal and paid more money. By 2025 gold was up about 60% on the year, its largest annual gain since 1979.

2026: a war that pushed gold down

The most recent case is also the most instructive. On 29 January gold set a record at 5,595 dollars an ounce. On 28 February the US and Iran went to war — which, on the rule that conflict lifts gold, should have started another leg up.

What happened instead: the conflict pushed oil above 90 dollars a barrel and briefly to 112, energy fed into inflation, and expectations for the Fed shifted from when do they cut to do they need to hike. Layer 1 deteriorated fast. On 5 June, May payrolls came in at 172,000 against a consensus of 80,000, rate-cut hopes were extinguished, and gold fell 3.27% in a single session, erasing the year's gain. By 11 June it traded at 4,023.95 intraday, about 26% below the January high, with the dollar index through 100 over the same stretch.

Layer 3 amplified the decline: gold ETFs saw two billion dollars of net outflows in May 2026. But layer 2 barely moved — official sector buying was 244 tonnes in the first quarter, still above the five-year average. In August, US data softened, hike expectations gave way to cut expectations, and gold rebounded more than 10% from 4,038 to close the month at 4,449. In September, energy-driven inflation pressure returned (August CPI 3.4% year on year, core 2.4%) and the market repriced toward a 25 basis point hike to 3.75%–4.00%, which would be the first increase since 2023.

This time the war pushed gold down, precisely by lifting oil, lifting inflation and lifting rate expectations. Remember only buy gold when war breaks out and the first half of 2026 is an inexplicable loss; triage it with the three questions and you find it ran down the second channel, where the direction was always in doubt.

Four popular intuitions worth discarding

Inflation rises, so gold must rise. Real rates and other sources of demand both matter. In 2022 inflation hit a forty-year high and gold fell for six straight months, yet finished about 0.4% higher on the USD LBMA PM benchmark. Inflation alone cannot explain that path.

There is a war, so buy gold. It depends which channel the event runs down. Threaten the monetary system and it is bullish; lift energy prices and therefore rate expectations and it can be bearish. The first half of 2026 is the textbook counterexample.

A crisis is here, so gold protects me. In the first phase of a liquidity squeeze, gold is often sold first precisely because it is the easiest thing to sell. October 2008 and March 2020 both worked that way. The haven property shows up later in a crisis, not on day one.

The dollar falls, so gold rises. A statistical tendency, not an identity. The rolling correlation between them shifts noticeably across regimes and can even turn positive in a panic. Treating a tendency as a law is the classic correlation misreading.

How to read a gold forecast, including ours

First, establish which horizon it is answering. A daily forecast discussing central bank buying trends, or an annual outlook agonising over this week's payrolls, is a mismatch of layer and horizon. Mismatched analysis is not necessarily wrong, but it is useless to your decision.

Second, reduce its argument to a layer. A weaker dollar supports gold is layer 1. Central banks keep buying is layer 2. ETFs have taken inflows three weeks running is layer 3. A key level gave way and triggered selling is layer 4. Once reduced, you can see immediately whether it is telling a story that lasts days or years.

Third, find its invalidation. A forecast worth reading states what would make the call wrong. Analysis that gives a direction and a target but no invalidation is an unfalsifiable claim — it can never be wrong, and therefore carries no information (more on this in scenario plans versus market predictions).

Back to three questions

Analysing gold comes back to three questions every time: whose expectations changed, how does that change buying and selling, and on which horizon does it act?

Answer those and rates, the dollar, central banks, ETFs and chart patterns stop being a pile of contradictory terms and become a map you can actually use.