Gold's floor is sovereign buying, not ETF noise

Gold bulls keep being told to watch Western fund flows, yields and the dollar. We think that misses the real balance-sheet story. Reserve managers are still the structural bid in gold, and they are absorbing supply on a scale that makes day-to-day ETF chatter look small. Official holdings across 15 tracked central banks are up 216.4 tonnes this year, and 7 of those 15 added again in their latest reported month, according to IMF International Reserves data.

Reuters captured the standard market script this week: “Strong dollar, high yields drag on gold outlook, but central-bank demand lends support,” and its survey said analysts had been lowering near-term gold price forecasts because of high Treasury yields and a strong dollar. We think that gets the pecking order wrong. Official buying is not a side note that “lends support”. It is the floor. If you stare only at GLD and macro prints, you will keep missing who is taking metal out of the market.

Official demand is not slow. It is large.

The World Gold Council said central banks and other official institutions added a net 321t in the third quarter, up 42% year on year and the highest quarterly total since Q4 2024. That is not dull background noise. It is a quarter in which official buyers took more gold than jewellers did, because jewellery demand was only 313t and fell 17% year on year.

Go wider and the picture gets stronger. The WGC said year-to-date central-bank demand reached 675t in Q3, up 6% year on year. It also raised its expectation for full-year central-bank demand to at least match 2025 after stronger-than-expected Q3 buying. In other words, the official sector is not fading as price rises. It is still there, and the industry body that tracks it has had to mark its estimate up, not down.

Then look at the country data. IMF International Reserves figures show net official gold holdings among 15 tracked central banks are up 216.4t this year, with Poland alone accounting for 122.7t. That matters for two reasons. First, this is not just a China story. Second, the bid is concentrated in sovereign balance sheets that do not trade like hedge funds. They do not stop out because yields backed up for a day.

China is not nibbling

China matters because it is persistent, and persistence is what creates a floor. The People’s Bank of China’s reserve-assets table shows gold reserves at 78.3 million ounces at end-October, with 25.8t added in the month. SBS, citing the PBOC, said that extended China’s buying streak to 24 straight months.

Twenty-four months is the point. This is not tactical dip-buying. It is a reserve-allocation programme. Before October, IMF International Reserves data showed China at 2,410t in September, up 103.3t this year after 23 straight monthly increases. October then added another 25.8t, according to the PBOC. When a sovereign buyer keeps adding through record prices, it is telling you something about motive: diversification first, entry level second.

Set that against the Western flow obsession. SPDR Gold Shares, the largest gold ETF, held 1,040.5t on Nov. 5, 2026, down 20.6t over the month, while its latest weekly flow was a modest -0.6t and its four-week flow was -14.9t, according to SPDR Gold Shares. China’s October purchase alone, at 25.8t, was larger than GLD’s monthly decline. That does not mean ETF flows do not matter at the margin. It means they are not the whole market, and lately they are not even the clearest one.

The price action should make sceptics uneasy

If the orthodox macro story were enough on its own, gold should have struggled harder. US data were hardly soft. JOLTS job openings came in at 7.52 million versus a 7.15 million forecast on Nov. 3, 2026. Initial claims were 189,000 against 197,000 expected on Nov. 5, 2026. Non-farm payrolls then printed 165,000 versus a 65,000 forecast, unemployment was 4.1% against 4.2% expected, and average hourly earnings rose 0.4% month on month against a 0.3% forecast on Nov. 6, 2026.

That is the sort of run that usually bolsters the dollar-and-yields case against gold. Yet COMEX gold was $3,945.40 before the first major central-bank-demand headline on Nov. 4, 2026, hit $4,154.40 by early Nov. 5, 2026, and was still at $4,074.90 late Thursday, Nov. 6, 2026, according to exchange futures data. It settled up $132.90, or 3.37%, on Nov. 4, 2026, and then held roughly that ground into the close. We should be careful about simple one-cause stories, but this is not what a market looks like when official demand is irrelevant.

There is a second-order effect here. When central banks keep absorbing metal, fewer ounces need to be cleared through price-sensitive private demand. That helps explain why weak spots in jewellery and patchy Western ETF flows have not produced the kind of sustained downside that the macro bears keep promising. Even Shanghai warehouse gold stocks rising to 118.4t, up 1.5t on the week and the highest since November 2025, suggest metal is being carried and financed inside the system, not thrown back at the market in disgust.

What would actually test this floor

The next test is not whether GLD has another soft week. It is whether stronger US data and sticky inflation can finally force gold through levels that central-bank buying has helped defend. The immediate event is US CPI on Nov. 10, 2026, with headline CPI forecast at 0.3% month on month and core at 0.2%.

If hot inflation and last week’s strong payrolls still cannot drive a deeper break than the brief dip to $3,928.90 seen on Nov. 4, 2026, then the sovereign floor thesis hardens. And if the next reserve updates show that China’s 24-month run was not a one-off flourish, while more than 7 of 15 tracked central banks keep adding, the market will have to stop pretending official buying is merely polite background support.