On Thursday, 30 July, the World Gold Council published its demand statistics for the second quarter of 2026. The week belonged to other stories. Amazon and Apple reported, the Federal Reserve decided on Wednesday, and the long end of the yield curve climbed to levels last seen before the financial crisis. The gold table disappeared into that noise.
That is a shame, because the table contains two numbers that, read side by side, dismantle the most frequently repeated story about the gold market. Not the story about rising prices — that one has already collapsed on its own. The far more durable story about who actually buys in this market, and why.
Two Numbers That Are Not Supposed to Fit Together
In the first quarter of 2026, gold set its all-time high. In late January the ounce printed 5,589.38 dollars (28 January, intraday) or 5,597.23 dollars (29 January), depending on which benchmark you use. It was the end point of a move that delivered roughly 65 percent during 2025 alone and took out the 3,000 and 4,000 dollar marks for the first time in history. In that very quarter, the world’s central banks bought a combined 57 tonnes of gold — a revised figure, and the weakest quarter in years.
In the second quarter of 2026, the gold price fell 14.1 percent. That was the steepest quarterly loss since the second quarter of 2013, when the metal dropped 22.7 percent. In that very quarter, the same central banks bought 289 tonnes — 62 percent more than a year earlier, and the highest figure ever recorded for a second quarter.
Five times as much metal after the price fell by a seventh. The World Gold Council itself gives two reasons for the jump in its commentary: the geopolitical backdrop, and — explicitly — softer gold prices.
That sentence contains the whole point, and it contradicts the standard line about this market. Central banks are described as the structural buyer, the patient buyer, the price-insensitive buyer. They buy for reserve policy, the argument goes, not for market opinion. The first half of 2026 says the opposite. The buyer everyone treats as indifferent to price turned out to be the most price-sensitive participant in the market this year. He skipped the record and took the discount.
What Actually Happened to the Price
The average price in the second quarter was 4,506.29 dollars an ounce. That is eight percent below the first quarter — and still 37 percent above the second quarter of 2025. The metal ended the quarter near 4,008 dollars.
On Friday, 31 July, the August futures contract opened at 4,102.40 dollars and touched 4,112.90 dollars during the morning; spot readings ran around 4,038 to 4,040 dollars. That leaves gold roughly 27 percent below its late-January record and, at the same time, about 25 percent above where it stood twelve months ago. Both numbers are correct. Which of the two an investor experiences as the real one depends entirely on when he bought.
Silver made the same move, only harder. The old record of 49.95 dollars from October 2025 was obliterated over the winter; on 29 January 2026 — the same day gold marked its high — the ounce reached 121.62 dollars. Between 1 April and 30 June it fell from 75.36 to 58.59 dollars, a decline of more than 22 percent. Anyone who wants to understand what leverage means inside a precious metals cycle does not need to go as far as the mining shares. Silver on its own is the lesson.
The Buyers, and the Sellers Nobody Quotes
The largest buyer of the quarter, by a wide margin, was the National Bank of Poland. It added 51 tonnes, bringing reserves to 632 tonnes at the end of June, and 82 tonnes for the first half. This is not opportunism but stated policy: in January 2026 the bank’s board approved a plan to acquire a further 150 tonnes and lift holdings to 700 tonnes, which would place Poland among the ten largest sovereign gold holders in the world. Governor Adam Glapiński has pushed gold to roughly 30 percent of total reserves and describes the programme openly as protection against geopolitical risk and as a way to reduce dependence on traditional reserve currencies.
Second came the People’s Bank of China with 33 tonnes, its largest quarterly addition since the fourth quarter of 2023, taking holdings to 2,346 tonnes. And here sits the sharpest single observation in the entire release. China bought 40 tonnes across the first half — 33 of them in the second quarter. In the record quarter, with the ounce running towards 5,600 dollars, the Chinese central bank took seven tonnes. After the drawdown, it took five times as much. Anyone looking for proof of price sensitivity will find it in that one pair of numbers.
Behind them, Uzbekistan added 16 tonnes, Kazakhstan 15, Jordan and the Czech Republic six each, with smaller purchases from Ghana, Singapore, the United Arab Emirates and Kyrgyzstan.
The sellers are left out of most summaries, and they are the reason the annual figure looks so weak. Russia disposed of 22 tonnes and was the quarter’s largest seller; the occasion was a federal budget deficit of more than six trillion roubles. Turkey shed four tonnes and cut its swap positions from 80 to 60 tonnes. Even the Bundesbank appears on the list with one tonne sold, though that is its routine outflow for coin minting.
Gross and net therefore diverge sharply. Across the whole first half of 2026, central banks bought only 345 tonnes on a net basis — the weakest first half since 2022, when the figure was 241 tonnes. The record headline is true for one quarter. The year so far is below average.
Four Groups of Buyers, One Price Move, Four Opposite Reactions
Total demand in the second quarter came to 1,269 tonnes, unchanged from a year earlier. That stability is an illusion. It is the sum of four groups that responded to the same decline in four different directions.
Central banks bought more, as described. Exchange-traded funds did precisely the opposite and shed 45 tonnes, outflows the World Gold Council attributes to weaker prices and a firmer dollar. That is the textbook procyclical investor: he buys while the chart points up and sells when it stops.
Jewellery demand fell to 278 tonnes, down 17 percent and the weakest quarterly volume since the pandemic. Yet spending on jewellery rose to 40 billion dollars, 14 percent more than a year earlier. Less metal, more money — exactly the pattern the luxury sector produces when pricing power replaces volume.
Bar and coin investment held flat at 307 tonnes, which the Council describes as a return to more typical levels. Recycling, the supply that comes out of household drawers, fell six percent to 326.1 tonnes: when prices decline, people stop taking old jewellery to the buyer. Supply does respond to price — but in the direction that cushions a fall rather than accelerating it. Mine production, meanwhile, rose two percent to 965.6 tonnes.
One small line at the edge of the table deserves a mention. Technology use rose two percent to 80.4 tonnes, because demand tied to artificial intelligence offset weakness in consumer electronics. Gold sits in the bond wires and contacts of every accelerator that dominated this earnings season. It is the one row where the two big stories of this summer touch.
The practical conclusion from those four reactions: there is no such thing as the gold market’s view. There are four groups of buyers with different horizons, different motives and opposite behaviour. Anyone who quotes one of them and builds a forecast on it has simply picked the group that suited the argument.
Why the Price Fell Anyway
The mechanism behind the decline is unglamorous, and that is exactly why it is reliable: gold pays no coupon. Its price depends on what the alternative pays.
The Federal Reserve under Kevin Warsh, confirmed by the Senate on 13 May 2026, left the policy rate at 3.50 to 3.75 percent on 29 July — but with three dissents in favour of a hike. At the start of the year, futures markets were still pricing cuts for 2026. By the end of May, essentially no cut was priced at all and the market had begun to debate increases. At quarter-end, the implied probability of a July hike stood at 33.7 percent, and the odds of at least one move by the September meeting at 67 percent.
At the long end the picture is starker still. On Friday, 31 July, the ten-year Treasury yielded 4.73 percent, the highest level since 15 January 2025; the thirty-year stood at 5.28 percent, a level last reached before the financial crisis. Against a guaranteed five percent, a metal that yields nothing has to make its case afresh every year.
The second engine of 2025 has also stalled. The geopolitical premium that carried gold towards 5,600 dollars over the winter shrank when the United States paused its airstrikes against Iran — the same pause that knocked nine percent off crude in late July. Both of the forces that drove gold in 2025 ran in reverse in 2026.
That answers the question of who actually sets the price. Two hundred and eighty-nine tonnes out of 1,269 tonnes of total demand is not even a quarter. Central banks set a floor, not a price. The price is set at the margin by whoever decides each morning whether to hold gold or 4.73 percent — and that is the fund and futures investor, not the reserve manager in Warsaw or Beijing.
The Equity Side: Miners Amplify in Both Directions
For investors who express precious metals through the stock market, the quarter was harder than it was for owners of bullion. The VanEck gold miners fund fell from roughly 96 dollars in early April to about 75 dollars on 30 June, a decline of 21 percent — around one and a half times the move in the metal. It has since recovered to about 78 dollars.
Over twelve months the picture inverts completely: the miners returned roughly 50 percent against 22 percent for the physically backed bullion fund. That is not an anomaly, it is the business model. A mine is a leveraged claim on the gap between the sale price and the cost of extraction, and leverage works in both directions.
The cost side is what makes the current phase uncomfortable. Industry all-in sustaining costs sit around 1,400 dollars an ounce, but they are rising: Newmont guides to roughly 1,680 dollars an ounce for 2026 against 1,358 dollars in 2025, citing lower sales volumes, higher royalties and production taxes. Rising costs against a falling price is a squeeze from two sides at once. This is also the argument for the royalty and streaming model — Franco-Nevada and Wheaton Precious Metals buy production at contractually fixed terms and are therefore structurally insulated from exactly this kind of cost inflation, which is why they usually fall less and rise less than the operators. Investors using the sector fund should also note that its largest holdings dominate it; a miner basket is far less diversified than its constituent count suggests.
There is a specifically American footnote to all of this, and it is not small. The United States is not part of the buying story at all. The Federal Reserve holds the largest official gold reserve in the world and has not added to it in any meaningful way since the 1970s. The record accumulation of the past several years is entirely a non-US phenomenon, and much of it is explicitly framed by the buyers as a reduction in dollar dependence.
The tax treatment matters just as much. In the United States, physical gold and physically backed bullion exchange-traded funds are treated as collectibles, with long-term gains taxed at up to 28 percent rather than the 20 percent that applies to equities. Mining shares are ordinary stock and are taxed as such. That single distinction argues for holding bullion exposure inside a tax-advantaged account and taking the equity leg in a taxable one — a decision worth more, over a long holding period, than most attempts to time the metal.
Risks and Counterarguments
The most important objection to any bullish reading is sitting inside the comparison itself. The second quarter of 2013 was not the end of a correction; it was the beginning of years in which gold made investors nothing. Central banks kept buying heavily through that period too. Their purchases did not rescue the price. Anyone reading the 289 tonnes as a buy signal should know that the identical configuration once meant nothing at all for a very long time.
The second objection is aimed at my own central claim. The 57 tonnes for the first quarter is a revised number, and central bank reporting arrives late and incomplete; some purchases only surface months afterwards. If the first quarter is revised upward, the contrast between 57 and 289 narrows. The direction would survive. The force of the argument would not.
Third, the category central bank is not an actor. Russia’s 22 tonnes had nothing to do with a view on gold and everything to do with a hole in the budget. And that is where the strongest challenge to the grand de-dollarisation narrative lies: the country whose roughly 300 billion dollars of frozen reserves in 2022 supplied the original argument for gold as a seizure-proof reserve asset was the largest seller of the quarter. This is not a contradiction — insurance exists to be liquidated when the emergency arrives. But it does mean the tonnage will not rise in a straight line, and that extrapolating from one good quarter is a mistake.
The survey evidence needs the same framing. Eighty-nine percent of reserve managers expect global central bank gold holdings to rise over the next twelve months; 45 percent plan to increase their own; 74 percent expect to reduce dollar holdings over five years. Those are intentions, not orders, and they have been consistently high for years while actual purchases have swung violently.
What an Investor Can Take From This
The rule that follows from this release is uncomfortable, because it cuts against a popular way of thinking: who is buying is not a forecast. The identity of the buyer tells you something about the floor and nothing about the next quarter. Central bank data arrives with weeks of delay and is then revised — by the time a private investor reads it, the market has long since had the information.
Three questions are more useful. First, what does the alternative pay? As long as the long end offers more than five percent, gold needs a reason beyond fear, and that reason has to come from monetary policy rather than headlines. Second, am I buying the metal or the leverage? Fifty percent against 22 over twelve months, and minus 21 against minus 14 in a single quarter, are the same investment in two states of matter, and choosing between them is a statement about your own tolerance rather than about gold. Third, in which currency and in which tax wrapper? For a US investor, the collectibles rate and the choice of account will change the after-tax result more reliably than a ten percent move in the price.
The most remarkable thing about Thursday’s table, in the end, is not that central banks bought a record. It is that they waited for the discount. The most patient buyer in this market also turns out to be its most disciplined one — and that is a considerably more useful piece of information than the headline it appeared under.
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.
