Gold fell 8.5% in September to end the month at $4,176 an ounce, yet physically backed ETFs took in roughly 67 tonnes over the same period while leveraged futures funds cut net length. The metal's investor base is moving in two directions at once.
TL;DR
- Price: gold ended September at $4,176/oz, down 8.5% on the month, against a LBMA AM benchmark high of $5,501.70 on 29 January.
- ETFs: September inflows of approximately $10bn, separately reported as about 67 tonnes, reversing Q2's 44.8-tonne outflow.
- Futures: Managed Money gold net length fell to 110,987 contracts on 6 October, 22,129 below its mid-September level.
- Official sector: 288.9t of central-bank buying in Q2, a second-quarter record, but H1 at 345t was the weakest first half since 2022.
- Yields: the US 10-year nominal yield rose 53 basis points to around 5.3% during September while the dollar index gained 2%.
Two Investor Bases Moving in Opposite Directions
Gold's September was not a uniform retreat. The World Gold Council reports the metal ending the month at $4,176 per troy ounce, down 8.5%, with higher US Treasury yields and a stronger dollar weighing on the market. Over the same month, physically backed gold ETFs attracted approximately $10 billion of inflows, separately reported as around 67 tonnes of demand. The Council publishes those two measures separately and both appear here as published, without reconciliation into a single figure.
Against that, the Council reports COMEX aggregate net-long positioning falling by roughly 100 tonnes during September, including an 84-tonne reduction in Managed Money net length. Different categories of investor moved in opposite directions within a single month. That coexistence is the central fact of the current market, and it is why a falling price cannot be read as a withdrawal of investor support — nor the ETF inflows as evidence of a floor.
| Measure | Latest | Change |
|---|
| Gold, end-September | $4,176/oz | −8.5% MoM |
| LBMA AM benchmark high, 29 January | $5,501.70/oz | — |
| LBMA benchmark low, late June | ~$3,994.50/oz | ~−27% from January |
| September ETF inflows | ~$10bn | +67t |
| Managed Money net length, 6 October | 110,987 contracts | −9,331 WoW |
| Central-bank net purchases, Q2 | 288.9t | +62% YoY |
| Central-bank net purchases, H1 | 345t | Lowest since 2022 |
| London vault holdings, end-August | 9,632t | +1.03% MoM |
| US 10-year nominal yield, end-September | ~5.3% | +53bp MoM |
Sources: World Gold Council, LBMA, CFTC. The January high and late-June low are LBMA benchmark fixings. The end-September price is as reported by the World Gold Council. Series are identified individually and are not interchangeable.
The Year in One Series
Measured on the LBMA benchmark, gold's year has been a parabolic advance followed by a long correction. The AM fix reached $5,501.70 on 29 January, the year's benchmark high. The move reversed within days and carried into early summer, with the benchmark fix bottoming near $3,994.50 in late June, a decline of roughly 27% from the January level.
Other series give different peaks for the same episode — COMEX futures recorded $5,318.40 and the intraday spot high was approximately $5,595 — because they measure different instruments at different moments. They are not contradictory, but they are not interchangeable, and mixing them produces drawdown figures that vary by several percentage points. The year-to-date price history in this analysis is therefore stated on the benchmark series, with intramonth and current market observations identified separately wherever they appear.
Futures Funds Keep Reducing
The reduction in speculative length continued into October. CFTC Disaggregated Commitments of Traders data show Managed Money gold net length at 110,987 contracts on 6 October, down 9,331 on the week, with funds cutting 4,610 gross longs while adding 4,721 gross shorts. The near-equal split matters: this was not profit-taking alone but a simultaneous increase in outright bearish exposure.
The sequence runs back through September. Net length stood at 133,116 contracts on 15 September and 120,318 on 29 September, a fall of 12,798 contracts, which at 100 troy ounces a contract is approximately 39.8 tonnes. That is consistent in direction and in part of its scale with the Council's 84-tonne monthly figure, but it is an illustration rather than a reconciliation: the balance cannot be assigned to earlier September dates without matching the Council's measurement window to the CFTC reporting dates. The 6 October reading is 22,129 contracts below mid-September, so the reduction has extended past the month end. Total gold futures open interest also fell, down 10,347 contracts to 396,109, so participation contracted alongside positioning.
The Official Sector Is Positive, Not Accelerating
Central banks bought a net 288.9 tonnes in Q2, a fivefold increase on Q1's revised 57 tonnes and 62% above the 177.9 tonnes of Q2 2025. The World Gold Council describes this as a record for a second quarter, which is a narrower claim than a record quarter. Poland was the largest buyer at 51 tonnes, taking reserves to 632 tonnes, followed by the People's Bank of China at 33 tonnes, its biggest quarterly addition since Q4 2023.
The half-year figure is the necessary counterweight. H1 net purchases totalled 345 tonnes, the lowest first half since 2022, with Turkey, Russia and Azerbaijan leading sales and the Bank of Russia the largest single seller at 22 tonnes. More recently the Council's monthly update reports 39 tonnes of net buying in August for 170 tonnes year to date. Those monthly figures rest on reported reserve changes and have different coverage from the quarterly demand estimates; they should not be added together. Official demand has remained positive throughout 2026, but describing it as uninterrupted acceleration would misstate the record.
China Shows Why Gold Is Not a Single Market
China's three demand channels moved independently in August. Domestic gold ETFs added 11 tonnes to reach 293 tonnes, and the People's Bank of China added approximately 20.2 tonnes, lifting reported reserves to 2,387 tonnes. Shanghai Gold Exchange withdrawals, the clearest available proxy for wholesale offtake, fell to 62 tonnes, down 22% month on month and 27% year on year.
Those readings are not in conflict. Reserve managers and financial investors can accumulate while fabricators and retail buyers step back, and high prices are a plausible contributor to the latter alongside seasonal, credit and channel factors. The Q2 global figures show the same split: bar and coin investment at 307.1 tonnes was down only 3% year on year, while jewellery consumption fell 17% to 278.2 tonnes. Treating gold as one demand market produces conclusions neither half of it supports.
Nominal Yields Rose; Real Yields Are the Question
The Council attributes September's decline to rising Treasury yields, a stronger dollar and falling futures positioning, with the US 10-year up 53 basis points to around 5.3% and the dollar index gaining 2%. The pressure has persisted: Reuters reported the 10-year near 5.36% on 9 October amid government debt and inflation concerns.
The distinction between nominal and real yields does real work here. Gold pays no contractual yield, so its carrying cost moves with real rates, not nominal ones. A nominal yield rising during an oil-driven inflation shock can leave real yields flat or lower if inflation expectations rise by as much or more. September's 53-basis-point move is therefore evidence of a changed nominal environment and only presumptive evidence of a changed real one, and the strength of the headwind on gold depends on which it was.
Mine supply offers no offsetting scarcity argument. Q2 production rose 2% to 965.6 tonnes while recycled supply fell 6% to 326.1 tonnes. Flow figures alone cannot establish availability — gold's above-ground stock, where it sits and who owns it bear more on deliverable supply than a single quarter's output — but neither series points to production constraint as a driver of the current price. Reuters' reporting from the LBMA conference on 5 October carried the counter-case: reserve diversification and geopolitical demand may be allowing gold to withstand rate conditions that would historically have produced greater weakness. That is a plausible reading, not evidence of a permanent decoupling.
Iran Cuts Both Ways
Geopolitical escalation is routinely treated as unambiguously supportive of gold. The LBMA's own Q2 analysis complicates that, connecting the gold-price reversal to the inflationary consequences of the Iran conflict. The mechanism runs through the distinction above: higher oil prices raise inflation expectations, which push nominal yields up, and whether real yields follow depends on which leg moves further. Escalation can therefore support gold through safe-haven demand and pressure it through the rate channel at the same time, with the net effect turning on the path of real rates rather than the severity of the headline.
That is directly relevant now. Reporting on 8 October had Brent above $103 on strikes speculation, while earlier commentary had pointed to progress in US–Iran discussions reducing the safe-haven premium. Both can be true at different moments in a fast-moving situation, and neither establishes a direction for gold on its own.
London Vault Holdings
LBMA data show 9,632 tonnes of gold held in London vaults at the end of August, an increase of 1.03% from July. The series includes commercial custodians and Bank of England holdings, and measures bullion physically stored in London rather than metal freely available for sale. The increase is consistent with an accumulation of bullion in London, but the aggregate figures do not identify the owners or distinguish ETF custody, official reserves and other holdings. Rising vault stocks therefore cannot be interpreted independently as evidence of weaker physical demand.
Outlook
Spot gold was reported by Reuters at around $4,194 on 9 October, up approximately 1.5% after an earlier two-month low. One session does not reverse an 8.5% month, and the September pattern of ETF accumulation alongside futures liquidation has not yet resolved.
Three questions shape the next quarter. Whether both flows persist, which the next CFTC releases and monthly ETF data will show directly — though neither series can demonstrate that one is absorbing the other, nor fix the causal relationship with price. How much support official-sector demand provides, which the Council's Q3 Gold Demand Trends will begin to answer. And whether physical consumption recovers as prices retreat, where Chinese wholesale withdrawals and jewellery volumes are the cleanest indicators. The rate path remains the most immediate variable, with the real-yield question unresolved on the evidence here.
Key Risks
Real-rate persistence. Gold's carrying cost tracks real rather than nominal yields. A nominal 10-year above 5.3% that proves to carry higher real rates would be the most direct threat to the investment bid; the same nominal level alongside rising inflation expectations would not.
Positioning concentration. Managed Money remains substantially net long at 32.1% of open interest against 4.1% short. A disorderly unwind of the remaining length would be larger than anything September produced.
Official-sector variability. H1 purchases were the lowest since 2022 and a substantial share of central-bank activity is never reported, so quarterly figures are estimates with real margins of error rather than precise measurements.
Geopolitical ambiguity. Escalation raises safe-haven demand and inflation expectations simultaneously, and the net effect on gold is not determinable in advance.
Data and source note: Monthly price, ETF and positioning context is from the World Gold Council's September market commentary and ETF reporting; quarterly supply and demand figures are from Gold Demand Trends Q2 2026. Price series are identified individually: the 29 January high of $5,501.70 and the late-June low near $3,994.50 are LBMA benchmark fixings, the end-September level of $4,176 is as reported by the World Gold Council, the COMEX high of $5,318.40 is a futures observation, and the 9 October price of approximately $4,194 is a spot observation reported by Reuters. These series measure different instruments and are not interchangeable. ETF dollar flows and tonnage changes are separately reported Council measures and are not converted between one another here. The 39.8-tonne contract conversion covers only the 15 to 29 September CFTC reporting dates and is illustrative of the Council's monthly figure rather than a full reconciliation of it. Central-bank monthly reported-reserve figures and quarterly demand estimates use different methodologies and coverage and are not additive. CFTC positions are Disaggregated Managed Money, futures only, as at 6 October 2026, released 9 October. Treasury yields cited are nominal. London vault holdings are the end-August 2026 LBMA observation, the most recent published at the time of writing.
Sources
This document is published by Bloodstone Research for informational and institutional research purposes only. It does not constitute investment advice, an investment recommendation, an offer or solicitation to buy or sell any financial instrument, commodity or security, or a forecast of future performance. Market conditions and data may change without notice. Readers should conduct their own analysis and, where appropriate, seek independent professional advice before making investment decisions. For institutional enquiries contact research@bloodstonecapital.co.uk.
