Gold rose 4.55% in the week to Friday 6 November and added more than 1% again on Monday morning, yet it remains approximately 27% below the record it set on 29 January 2026. The twelve-month return of 1.37% conceals that shape entirely: an extraordinary rally into late January followed by a large and sustained correction. The correction ran through three quarters in which central banks bought 675 tonnes, exchange-traded fund holdings reached a month-end record and Chinese official reserves rose for a 24th consecutive month. The useful question is therefore not whether gold is being bought, because it plainly is, but which categories are adding and which are leaving.
TL;DR
Price: The New York close rose from $3,883.35 on 30 October to $4,060.20 on 6 November, a gain of 4.55%. Against 7 November 2025 the gain is 1.37%. Monday morning quotations run between approximately $4,087 and $4,123 depending on observation time.
Drawdown: Spot gold touched an intraday record of $5,594.82 on Thursday 29 January 2026. The 6 November close is 27.4% below it. Platinum and silver are further below their own January peaks, at approximately 47% and 57%.
Official sector: Central banks bought an estimated 321.2 tonnes in the third quarter, up 42% year on year and about 41% above the five-year quarterly average of 227 tonnes. Estimated year-to-date demand is 675 tonnes.
Reported against estimated: Reported global net buying was 120 tonnes in the quarter and 228 tonnes for January to September, against the estimated 675 tonnes. China and Poland alone supplied approximately 86% of the quarter's reported net purchases.
China: Reported reserves rose 830,000 ounces, approximately 25.82 tonnes, in October, to roughly 2,435 tonnes, roughly 2.25 times the pace of the preceding nine months.
ETFs set a record: October brought a fourth consecutive month of inflows, $2.7bn and 19 tonnes, taking holdings to a month-end record of 4,275 tonnes. Europe added $1.8bn and Asia $1.3bn while North America lost $512m.
Futures exposure contracted: Reported Managed Money net length fell 1,948 contracts to 103,614 in the week to 3 November, with gross longs down 225 and reported gross shorts up 10.2%. The Council's broader COMEX measure fell the equivalent of 94 tonnes during October.
The category that halved: Over-the-counter and other demand fell to 143.5 tonnes in the third quarter from 305 tonnes in the second, the largest single swing in the demand table.
Market Overview
The week's gain needs setting against the level from which it came, and the instruments involved are not interchangeable.
Weekly and annual figures are one provider's New York closes, not an LBMA benchmark series. Monday quotations are time-specific and differ by observation time and quote convention rather than conflicting. The record is an intraday spot observation, not an exchange fixing. The quarterly figure is an average of the LBMA PM benchmark and is not a quarter-end price. See the data note.
The drawdown is not specific to gold, which is the first thing the figures establish.
From January 2026 peak
Peak
6 November close
Change
Gold
$5,594.82
$4,060.20
−27.4%
Platinum
$2,923.70
$1,538.15
−47.4%
Silver
$120.45
$51.89
−56.9%
Peaks are intraday spot observations for gold and silver, both recorded on 29 January 2026, and the January 2026 record as published in our platinum analysis of 2 November. Closing prices are New York closes. Comparing an intraday peak with a close overstates each decline slightly; the ranking between them is unaffected.
Gold has fallen least of the three by a wide margin. Whatever repriced the complex from late January did so across metals with very different demand structures, which is difficult to reconcile with an explanation resting on gold's own supply and demand.
Within the week the move was concentrated rather than distributed. Spot rose approximately 3% on Wednesday 4 November to $4,046.74, its largest single-day gain since 19 August, with December futures up 2.9% to $4,054.60. That session followed the conclusion of the United States midterm elections on Tuesday 3 November, and the same mid-week turn appears across the complex, with platinum up 7.97% on the week, silver 5.75% and palladium 4.91%. We are not asserting that the election result caused the move; the sequence establishes that the gain arrived in the New York session after a scheduled political event, with Shanghai following on the Thursday.
Reported Buying and Estimated Buying Are Different Numbers
The World Gold Council's Gold Demand Trends for the third quarter, published on 3 November, records the strongest quarter of central bank buying since 2024. The distinction that matters most in it is between what was disclosed and what was estimated.
Central banks, Q3 2026
Tonnes
Estimated total demand
321.2
Q3 2025 comparison
226.3 (+42%)
Five-year quarterly average
227
Reported net buying
120
Estimated demand, January to September
675 (+6%)
Reported net buying, January to September
228
People's Bank of China, reported
63
National Bank of Poland, reported
40
National Bank of Kazakhstan, reported
14
Central Bank of Uzbekistan, reported
14
Czech National Bank, reported
5
Bank of Russia, reported
−19
World Gold Council, Gold Demand Trends Q3 2026, and its central bank statistics for September published 4 November 2026. Reported country data were captured to 30 October and may be revised as delayed disclosures arrive. Russia's reduction was the only sizeable sale; Qatar, the Kyrgyz Republic and the United Arab Emirates made small reductions.
Two measures sit in that table and they are not interchangeable. Reported net buying for the first nine months was 228 tonnes. The Council's estimate of total official-sector demand over the same period was 675 tonnes, so the estimate exceeds disclosed net transactions by roughly a factor of three, and the Council attributes around 63% of the third quarter's demand to unreported activity. In aggregate, therefore, disclosed official purchases substantially understate estimated official demand. That is an aggregate relationship and not a per-institution one: it does not follow that any particular central bank is buying more than it reports, nor that the gap is constant between periods.
Within the disclosed figures the concentration is striking. China's 63 tonnes and Poland's 40 tonnes together account for approximately 86% of the quarter's 120 tonnes of reported net buying. That is a more revealing comparison than measuring either against the 321.2 tonne estimated total, because a reported country figure and a partly estimated global aggregate are different kinds of measurement. Against those additions, Russia reduced holdings by 19 tonnes in the only sizeable sale of the quarter, a notable direction for a sanctioned producer to take while its principal trading partner accumulates. September alone brought 61 tonnes of reported net buying, the highest monthly reported figure of 2026 to that point.
China's October Addition
The State Administration of Foreign Exchange reported reserves of 78.30 million troy ounces at the end of October against 77.47 million at the end of September. The increase of 830,000 ounces is approximately 25.82 tonnes and takes reported holdings to roughly 2,435 tonnes, in what is reported as the 24th consecutive month of increases. Because the figure is a quantity measured in ounces, a change in the dollar price cannot by itself produce it.
The comparison with preceding months requires care. Reported additions for January to October total approximately 129 tonnes, but October's own 25.82 tonnes sits inside that figure, so dividing it across ten months and measuring October against the result compares the month to an average it has itself raised. Removing October leaves approximately 103.2 tonnes across nine months, an average of about 11.5 tonnes a month, which makes October roughly 2.25 times the preceding pace. October has been described elsewhere as the largest monthly addition since the current sequence began in November 2024, which we are not in a position to confirm without the full monthly official series. The disclosure is in any case a month-end balance rather than a transaction ledger.
ETF Holdings Rose While Reported Futures Exposure Fell
Gold ETFs, October 2026
Figure
Net inflows
+$2.7bn
Gold added
+19t
Total holdings
4,275t (month-end record)
Assets under management
$542bn (−6% month on month)
Europe
+$1.8bn
Asia
+$1.3bn
North America
−$512m
Other regions
+$62m
January to October inflows
$41bn, +246t
World Gold Council, published 5 November 2026. Rounded regional figures are not intended to reconcile exactly with the global total.
October was the fourth consecutive month of global inflows and holdings reached a month-end record despite the price falling during the month. The relationship between the two headline figures makes that concrete: tonnage rose 19 tonnes while the dollar value of the holdings fell 6%, so these funds held more metal worth less money at the end of October than at the start. Europe has attracted inflows for 18 consecutive weeks, and across the first ten months the United Kingdom recorded the largest national inflows with China close behind. North America was the only region in outflow, which the Council suggests may reflect liquidity requirements during equity-market weakness, though motive cannot be established from flow data.
Reported futures exposure moved the other way. The Commodity Futures Trading Commission's report released on 6 November, covering positions held at the close of Tuesday 3 November, shows Managed Money net length down 1,948 contracts to 103,614, with gross longs falling 225 while reported gross shorts rose 1,723, an increase of 10.2%. The net decline is therefore accounted for by the increase in reported gross shorts rather than by a reduction in longs. The report is a snapshot of outstanding positions by category: it does not establish when any position was initiated, who initiated it, or whether a change reflects new exposure rather than a transfer between traders. At 100 troy ounces a contract the net long is equivalent to approximately 322 tonnes of notional exposure, which measures futures positioning rather than bullion owned or stored. Separately, the Council reports its broader COMEX net-long measure falling the equivalent of 94 tonnes during October to 560 tonnes; that measure covers a different scope and the two are not interchangeable.
What the two datasets jointly establish is that a contraction in reported COMEX futures exposure coincided with an increase in gold held by physically backed investment funds. They cover different instruments and different investor populations, and they cannot establish that the same capital moved from one to the other.
Demand Divided by Geography, and the Category That Halved
Treating consumer demand as uniformly weak would misread it. The global figure conceals a split.
Bar and coin, Q3 2026
Tonnes
Change y/y
Global
267.8
−19%
China
78.9
+7%
India
55.0
−40%
Europe
approx. 20
−31%
United States
approx. 8
−40%
World Gold Council, Gold Demand Trends Q3 2026. China also recorded 393 tonnes over the first nine months, its highest January to September total in the Council's series.
Chinese bar and coin demand rose 7% while the global figure fell 19%, and China alone accounted for approximately 29% of the world total. Indian demand fell 40%, though the Council notes that differences in the festival calendar affect that annual comparison, and United States physical investment fell to its weakest quarter since the fourth quarter of 2019. Jewellery consumption fell 17% to 312.9 tonnes, the weakest third quarter in the Council's series, with buyers moving toward lighter and lower-carat pieces; consumption value nonetheless rose 2% to $43bn, so buyers spent slightly more money on materially less metal. Technology at 84.8 tonnes was the only growing category, up 3% on manufacturing associated with artificial intelligence.
The largest movement in the demand table is in none of those categories. Over-the-counter and other demand fell to 143.5 tonnes in the third quarter from 305 tonnes in the second, a swing of more than 160 tonnes that exceeds the entire year-on-year decline in jewellery and bar and coin demand combined. That category is partly estimated and contains a statistical residual, so it cannot be read as a precisely observed liquidation of private holdings. It is nevertheless the single largest quarterly change in the table, and any account of why record official and fund buying did not hold the price has to contend with it.
Supply Grew and Scrap Stayed Subdued
Q3 2026 supply
Tonnes
Mine production
1,070.9
Recycling
314.5
Net producer hedging
−16.8
Total supply
1,368.6
World Gold Council, Gold Demand Trends Q3 2026. Net producer de-hedging reduced the quarterly supply measure by 16.8 tonnes.
Mine production was a third-quarter record, up 3%. Recycling fell 8% to its lowest since the fourth quarter of 2023, which the Council describes as remaining subdued despite high prices. That is the narrower and better-supported observation: scrap supply did not rise as the quarterly average price ran 23% above a year earlier. It does not establish that holders are unresponsive to price at every level, nor what they would do at a different one. Mining in any case responds over long investment cycles, and against each year's production the Council estimates total above-ground stocks of approximately 222,600 tonnes on its latest assessment, so the behaviour of existing holders governs far more metal than annual mine output. Total supply and total demand are equal in the Council's framework because it balances by construction, which is an accounting identity rather than evidence about physical market clearing.
What Has Changed Since January
Gold's peak was set on 29 January in conditions that no longer obtain. Reuters attributed the rally to safe-haven demand around United States and Iran tensions, weakening United States economic signals, a weaker dollar, expectations of Federal Reserve rate cuts and strong fund inflows, with SPDR Gold Trust holdings at their highest since May 2022.
Each of those has since moved. Negotiations between the United States and Iran are under way in Doha; Reuters attributed Monday's gain partly to lower oil prices after comments suggesting a reduced likelihood of further United States strikes. United States employment growth strengthened in October and unemployment fell to 4.1%. Rate-cut expectations have been replaced by expectations of an increase, with CME FedWatch implying a probability around 79% to 81% for December after the employment data against 71% before, a market-implied figure at a point in time rather than any commitment by the Federal Reserve. Goldman Sachs has cut its end-2026 gold forecast to $4,150 an ounce from $4,650 and its end-2027 forecast to $4,800 from $5,400, citing the implications of a more hawkish Federal Reserve stance; a revised bank forecast is an opinion and is reported here as one.
What the evidence supports is that these developments coincided with gold's repricing while official-sector and fund accumulation remained substantial. It does not apportion the decline between them, and it does not establish what the price would have been without the buying.
Outlook
Three propositions are testable. The first is whether official buying holds, and the Council's own framing is more qualified than the tonnage suggests: it expects central bank demand to at least match 2025 after a surprisingly strong third quarter, while explicitly acknowledging that official buying may itself be sensitive to price. That qualification matters, because an official bid that responds to price behaves differently from one that does not.
The second is whether the divergence within investment demand persists. The Council expects total investment demand to finish 2026 approximately 200 tonnes above 2025, which would be growth in the aggregate alongside the geographic split already visible, with European and Chinese buyers adding while North American funds and United States retail investors reduce.
The third is whether the risk premium returns. The Doha talks concern the same relationship whose deterioration accompanied the January peak, and oil was the transmission channel into gold on Monday. Against that sit October inflation releases on 10 and 13 November and a December policy decision the market currently prices as more likely than not to tighten.
Key Risks
The record is an intraday spot print. $5,594.82 was touched during trading on 29 January 2026 and is not an exchange fixing or a closing price. Comparing it with a New York close slightly overstates the drawdown, and other feeds place the January peak at marginally different levels.
Reported and estimated official demand are different measures. Reported net buying was 120 tonnes in the quarter and 228 tonnes for the nine months, against an estimated 675 tonnes. The gap is an aggregate estimate and cannot be attributed to any particular institution or assumed constant between periods. Reported country data were captured to 30 October and may be revised.
Reserve disclosures are not transaction records. China's reported increase establishes a change in a disclosed month-end balance. It does not establish timing within the month, prices, counterparties or mechanism.
Observation periods do not align. The central bank figures cover a quarter, China's reserve change is an October month-end balance, ETF data are monthly and positioning is a Tuesday snapshot. None observe the same period and they are not combined here.
Positioning and flows are not prediction. The reported net long of 103,614 contracts describes exposure at the Tuesday cut-off, two sessions before the week's decisive move. It does not reveal when positions were initiated or whether changes reflect transfers between traders. ETF flows record what was bought, not why.
OTC and other demand is partly estimated. The 143.5 tonne figure contains a statistical residual and is not a precisely observed category, so the quarterly swing in it should not be read as a measured liquidation.
The causes of the drawdown are not quantified here. Rate expectations, the dollar and geopolitical risk all changed alongside the price. Apportioning the decline between them would require matched yield and currency observations that this piece does not contain.
Intelligence Monitoring Points
The next reserve disclosure. Whether China's November addition sustains October's 25.82 tonnes or reverts toward the approximately 11.5 tonne monthly average of the preceding nine months.
Reported against estimated official buying. Whether the Council's reported net total rises materially from 228 tonnes for the nine months as delayed disclosures arrive, which would narrow the gap to its 675 tonne estimate without any change in underlying demand.
October inflation data. United States CPI on 10 November and PPI on 13 November, both feeding the rate expectations that have moved since January.
The CFTC report of 16 November. A delayed release owing to a federal holiday, covering positions as at 10 November, and the first to capture positioning after the 4 November move.
November ETF flows. Whether a fifth consecutive month of inflows arrives, whether European inflows extend past 18 weeks, and whether North America remains the only region in outflow.
FAQ
Why has gold fallen if central banks are buying record amounts?
The data establish that both happened rather than explaining the relationship between them. Official and fund accumulation continued while rate expectations, the dollar and geopolitical risk all moved against gold from late January. Platinum and silver fell considerably further from their own January peaks, which suggests a repricing across the complex rather than one driven by gold's own demand.
How far below its record is gold?
27.4%. Spot gold touched an intraday record of $5,594.82 on 29 January 2026 against a 6 November New York close of $4,060.20. Because the first is an intraday print and the second a close, the comparison slightly overstates the fall.
Did central banks really buy 321 tonnes in the third quarter?
That is the World Gold Council's estimate. Reported net buying was 120 tonnes, and the Council attributes around 63% of the quarter's demand to unreported activity. For the first nine months, reported net buying was 228 tonnes against an estimated 675 tonnes.
Are investors leaving gold?
Not uniformly. ETF holdings reached a month-end record of 4,275 tonnes in October on a fourth consecutive month of inflows, while reported COMEX futures exposure contracted and Western retail demand fell sharply. The composition of demand changed more than its level.
Is China's October addition the largest since purchases resumed?
It has been reported as such and we have not confirmed it, because establishing it requires the full monthly official series. What the data support is that October was roughly 2.25 times the average monthly pace of the preceding nine months.
What was the biggest change in the demand table?
Over-the-counter and other demand, which fell to 143.5 tonnes from 305 tonnes in the second quarter. That category is partly estimated and includes a statistical residual, so the swing is not a precisely observed figure, but it is larger than any other quarterly movement in the table.
Data and Source Note
Weekly and annual gold prices are one provider's New York closes for 30 October 2026, 6 November 2026 and 7 November 2025. That provider labels the annual comparison as a month-end series, which these observation dates are not, so it is described here by its dates. Monday 9 November quotations are time-stamped spot and futures observations from separate providers and are not a single price. Gold's record of $5,594.82 and silver's of $120.45 are intraday spot observations on 29 January 2026 as reported by Reuters; they are not exchange fixings, and comparing them with closing prices slightly overstates each decline. Platinum's January 2026 record of $2,923.70 is as published in our platinum analysis of 2 November 2026. The third-quarter figure of $4,262.2 is an average of the LBMA PM benchmark and is not a quarter-end price.
Demand, supply, central bank, bar and coin, jewellery, technology and over-the-counter figures are from the World Gold Council's Gold Demand Trends for the third quarter of 2026, published 3 November 2026, with monthly reported central bank statistics for September from its publication of 4 November 2026. Total demand including over-the-counter investment was 1,368.6 tonnes and excluding it 1,225.1 tonnes. The Council separately reports investment demand excluding OTC at 479 tonnes, which already contains ETF and bar and coin investment and is therefore not added to them here. Over-the-counter and other demand is partly estimated and includes a statistical residual. The equality of total supply and total demand in the Council's framework is an accounting identity rather than evidence about physical market clearing. Reported country data were captured to 30 October 2026 and may be revised as delayed disclosures arrive, and reported and estimated official demand are distinguished throughout.
ETF holdings, flows, assets under management and regional splits are from the Council's gold ETF report published 5 November 2026 and cover October 2026. Regional flow figures are rounded and are not intended to reconcile exactly with the global total. Above-ground stocks of approximately 222,600 tonnes are the Council's latest estimate.
Chinese reserve figures are reported holdings published by the State Administration of Foreign Exchange and carried by SMM on 9 November, converted at 31.1035 grams per troy ounce. The nine-month average of approximately 11.5 tonnes a month is derived by subtracting October's addition from the approximate January to October total before dividing, and rests on a year-to-date figure rather than a verified monthly series.
Positioning figures are from the CFTC Disaggregated Commitments of Traders, Futures Only, Managed Money category, for positions held at the close of Tuesday 3 November 2026 and released Friday 6 November. Prior-week figures are derived from the reported change columns. The notional tonne equivalent uses the 100 troy ounce contract size and describes exposure rather than metal owned. The Council's COMEX net-long measure of 560 tonnes covers a different scope and is labelled separately. The Commission's published calendar lists 16 November 2026 for the following report, marked as a delayed release owing to a federal holiday, the intervening holiday being Veterans Day on 11 November.
The January 2026 rally drivers, Monday's macroeconomic attribution, the October employment figures, the 4.1% unemployment rate, the CME FedWatch probabilities and the CPI and PPI dates are as reported by Reuters. Implied probabilities are market pricing at an observation time and are timestamp-sensitive; figures around 79% and 81% have been reported for the December decision. Goldman Sachs forecast revisions are as reported by Reuters and are stated opinions rather than observations. No nominal yield, real yield or dollar index series is used in this analysis, and no apportionment of the drawdown between macroeconomic variables is attempted. Instruments are quoted in their original currencies with no foreign-exchange conversion applied.
This document is for informational purposes only and does not constitute investment advice. Data and estimates may be revised. Independent financial advice should be sought before making any investment decision.
