Two months ago we wrote that gold’s two great groups of owners had traded places. Western investors, who hold the metal mostly through exchange-traded funds, were selling into the correction — $8.9 billion out in June alone, capping North America’s weakest first half since 2013 — while the world’s central banks used the same weakness to buy. We argued that the roughly 298 tonnes of ETF gold then sitting at a loss would act as a ceiling, with every rally running into holders waiting to get out at break-even. August put that idea to the test, and it did not hold. Global gold ETFs took in $18 billion, the second-largest monthly inflow on record, and their holdings rose 121 tonnes to a record 4,189. Gold climbed 13 percent, its third-best month in twenty-five years by the World Gold Council’s count. Whatever break-even selling there was, the buying swamped it.
It is worth being plain about a call that did not survive contact with the data, because what overturned it is more interesting than the call itself. The Western buyers came back — but not for the reason Western buyers usually come back to gold.
Who came back
The reversal was broad, and it was led from the West. European funds took in $7.9 billion, their strongest month on record; within that, investors in the United Kingdom added $4.4 billion, their second-largest monthly inflow ever, and those in France $1.5 billion, their largest. North American funds — the same investors who had been heading for the exits in June — added almost as much as Europe, their third-largest month on record. Asia contributed $2 billion, its best month since February, most of it in China. The total value of gold held in ETFs worldwide rose 16 percent in a single month, to $615 billion, lifted by the rising price as well as the new money. For the year to date, flows now stand at $29 billion of net buying.
The futures market told the same story at a faster tempo. Net long positions on the COMEX exchange jumped by more than 200 tonnes, to 753 tonnes, and average daily trading volumes in gold rose by about a fifth. Some of that was systematic, trend-following money of the kind that piles into whatever is moving. ETF flows are a different animal. They are the allocations of wealth managers, pension funds and individuals deciding, deliberately, to own more gold.
Key Data
August global gold ETF inflows: $18 billion (121t), the second-largest month on record; holdings at a record 4,189t; assets under management up 16% to $615 billion. Europe +$7.9 billion (a record month; UK +$4.4 billion, France +$1.5 billion); North America close behind in its third-largest month; Asia +$2 billion. Year to date: $29 billion (160t). COMEX net longs +212t to 753t. Gold +13.3% in August to $4,563/oz. The People’s Bank of China bought 20.2t in August, its 22nd straight month and largest purchase since October 2023, lifting reported reserves to about 2,387t.
Not a bet on the Fed
Here is what makes August unusual. The classic reason for Western investors to buy gold is the expectation of falling real interest rates: the Fed cuts, the return on cash and bonds drops, and a metal that pays nothing looks better by comparison. That is not what happened. Through August the Federal Reserve moved toward tightening, not easing. Three of its officials had voted for a rate hike in July, and on August 28 Chair Kevin Warsh used his Jackson Hole speech to warn that inflation remained too high; bond markets responded by pricing a hike. By the rate trader’s usual logic, this was a month to sell gold.
The World Gold Council’s account of the drivers explains why investors did the opposite. It pointed to concerns about the joint US–Japanese intervention to prop up the yen at the end of July, rising Treasury yields and doubts about fiscal sustainability, worries over European sovereign debt, and momentum as the price broke through technical levels. Set the momentum aside and those are not a trader’s reasons. They are, almost word for word, the reasons central banks have given for buying gold for years: doubts about the currencies and government debt that make up the rest of their reserves. The fiscal side of that story — a 19-year high in the 30-year Treasury yield, a national debt through $40 trillion, and a Treasury buying back its own long bonds — is the subject of our long read, Gold and the Shadow of Fiscal Dominance.
Two clocks, briefly in step
In July we described the gold market as two clocks keeping different time: the ETF holder’s, which runs in afternoons and quarters and answers to the real yield, and the reserve manager’s, which runs in decades. For one month they struck the same hour. While Western funds were piling in, the People’s Bank of China bought 20.2 tonnes, its largest monthly purchase since October 2023 and its twenty-second consecutive month of buying, taking its reported reserves to about 2,387 tonnes.
The official side of the ledger is not uniformly bullish, and it would be a mistake to pretend otherwise. Reported central-bank buying has run slower in 2026 than in 2025 — about 130 tonnes in the first seven months, against roughly 160 tonnes over the same stretch last year — and a handful of banks, led by Turkey and Russia, have been net sellers as domestic pressures bite. The World Gold Council still expects around 850 tonnes of official buying this year once unreported purchases are included. The long clock is still ticking. It was simply not the only thing moving the price in August.
September’s test
Momentum cuts both ways, and September has begun to test the new money. From its late-August high above $4,700, gold fell for three straight weeks as a strong jobs report, a hot inflation reading and a fresh oil shock — Saudi Arabia shut the pipeline that carries its crude around the Strait of Hormuz, as we describe in The Pipeline Around Hormuz Closed — pushed the odds of a September rate hike toward 90 percent. On September 14 gold slipped below $4,300, roughly where it began the year, giving back about half of August’s gain.
So which clock was the August money on? The first evidence is encouraging for anyone hoping it was the long one. In the five trading days to September 8, as the price edged lower, SPDR’s two main US gold funds took in close to $2 billion between them. That is one week, from one family of funds, and no substitute for the World Gold Council’s full September figures due in early October. But investors who buy into a falling price are not behaving like momentum traders. They are behaving like people who have decided what gold is for.
The lesson of the past three months is not that the ceiling we described was imaginary, or that Western investors have turned into reserve managers overnight. It is that the reasons people hold gold can change faster than the flows suggest. In June the marginal Western buyer was a rate trader, and rate traders sell when real yields rise. In August a large share of the new money arrived for reasons a rate hike does not answer: the credibility of government debt, the stability of currencies, and the growing willingness of governments to intervene in markets they once left alone. If that is the money that stays, a Fed hike can test gold’s price without necessarily breaking its bid.
For individual owners the takeaway is the same as in July, with one amendment. Decide which clock you are on before the next red screen — and notice that the question has become more interesting, because this summer a great deal of Western money started behaving like the other clock. Whether it holds on through September will tell us whether it meant it. If you want the long-horizon context that money was buying into, it is mapped nation by nation in our gold reserves cluster, and the Fed’s new approach to telling markets — or not telling them — what comes next is the subject of The Fed Has Stopped Giving Directions.
