India bought 52 tonnes of gold. Bullion fell anyway.
India bought 52 tonnes of gold over Dhanteras and Diwali and China added 25.8 tonnes in October, yet bullion hit a one-year low. Demand is price-sensitive, not price-setting, and Friday's PPI is the next test.
$3,881.70$3,888.70$3,853$3,995.96$3,853.56$3,883.37$3,852$3,800
Gold's third consecutive down session took spot to $3,881.70 an ounce by 13:36 ET, 2.7% lower and its weakest print since October 2025 (CNBC). The December contract went with it at $3,888.70, down 2.8%. In London the intraday low was $3,853, "almost $20 beneath last month's low" (Bullionvault).
The desk's own series agrees to the dollar. On Binance PAXG 15-minute bars, Thursday ran from a $3,995.96 high in the 01:00 UTC bar to $3,853.56 at 15:00, and last traded $3,883.37 — a $142 session against a 20-day mean of $63. Asia sold it, Europe sold it, New York sold the bounce.
In the same week, the physical market did the opposite.
52 tonnes, bought at a lower price
The India Bullion and Jewellers Association estimates consumers bought 52 tonnes of gold worth ₹72,000 crore over Dhanteras on 6 November and Diwali on 8 November, plus 300 tonnes of silver worth ₹5,400 crore. Volume was 30% higher than the same festivals last year, silver 10% (Times of India).
That is not a small number in a market that had been starving for tonnage. Indian sales by volume had fallen nearly 50% in the fiscal year through September, after the import duty went to 15% in May and the rupee price hit records. What restarted the buying was arithmetic: the domestic price fell from about ₹1.8 lakh to ₹1.42 lakh per 10 grams. The industry body now expects advance purchases for a wedding season running from mid-November to March.
The wholesale premium says the same thing from the other side. Dealers quoted up to $4 an ounce over official domestic prices into Dhanteras, against $7 the week before — shrinking because metal was arriving, not because demand was missing. Onshore China traded flat to a $9 premium (MINING.COM, Reuters).
The official sector added its own bid: China's central bank lifted reserves by 830,000 ounces in October, about 25.8 tonnes, a 24th consecutive month and its largest single month in three years (Bloomberg).
And the funds were not dumping metal, whatever the screen said. Global physically backed gold ETFs took in $2.7bn in October, a fourth straight month of inflows, lifting holdings 19 tonnes to a record 4,275 tonnes; North America was the only region to sell, at -$512mn, against +$1.8bn in Europe and +$1.3bn in Asia (World Gold Council). The largest fund, SPDR Gold Shares, held 1,041.08 tonnes on 11 November, 0.57 tonnes more than a week earlier, and is down 29.48 tonnes this year from the 1,070.56 tonnes it held at the end of 2025 (goldreporter).
So what was selling?
Not the metal. The paper.
The dollar index added 0.34% to 103.48, near a 19-month high (CNBC). The 10-year yield went back above 5.00%, the first time in three weeks, and futures put 94% odds on a quarter-point December hike and 6% on a half-point (Bullionvault); CNBC's capture of the same screen showed 85% for the quarter point. That gap is the clock time of the grab, not a disagreement about the world.
Brent rising more than 2% after the US president said the country did not need a deal with Iran made it worse, not better: higher crude now reads as higher inflation and therefore a longer hold on rates. "If crude oil prices spike higher on this news, that's negative for gold from an inflation perspective," Jim Wyckoff of American Gold Exchange told CNBC.
Now put the two clocks side by side. A 15% duty in India and a wedding calendar measured in months. A two-year yield and a dollar index moving intraday. The second sets the price; the first buys what the second marks down.
The council already warned about this
The World Gold Council's Q3 outlook says it plainly enough to quote: "Purchases nevertheless appear sensitive to price. This pattern could provide some price support and stability, although the relationship is not sufficiently stable to imply a firm floor for the gold price" (WGC Q3 outlook).
Its own quarterly table shows the first half of that. Central banks bought 57.7 tonnes in Q1, when gold was at its record. They bought 321.2 tonnes in Q3, after the price fell. Official demand is a buyer of dips: good for the floor, useless as a timing signal.
India shows the same elasticity. Total demand was 162 tonnes in Q3, up 24% on the quarter but 29% below a year earlier, and domestic prices traded at a discount to import parity averaging $43 an ounce — the highest quarterly level on record — with unofficial inflows and old-gold supply doing the discounting (WGC India Q3). A market where the consumer buys at a discount is one where the seller is still the keener party.
Demand that is a function of price cannot be the thing that stops the price falling. It follows it, on a lag measured in festivals and months. We made the same point about the Q3 records on 3 November, and the week since has not argued otherwise.
What actually turns this
Levels first, because they are the only part of this that is checkable by tomorrow.
- $3,852-$3,853. Thursday's low on both the desk's series and the London print. A close below opens $3,800.
- $3,922-$3,950. Thursday's broken morning shelf, the same zone today's signal sells into. Reclaiming $3,950 is the first evidence the sellers have run out.
- $3,995.96, then $4,000. The session high, and the line the market has now lost three times since late October. Until it is reclaimed the trend is down.
Then the calendar. October's producer prices land Friday 13:30 UTC — 08:30 ET, now that US daylight saving has ended — consensus +0.4% on the month from September's 1.3%, core +0.3% from 0.7% (BLS schedule, Yahoo Finance).
Do not treat the number as a directional signal in itself. The last two PPI releases moved gold in opposite directions: August's came in hotter on the year and gold fell 1.39%, while September's 1.3% was more than double consensus and gold rose about 1% because the 10-year fell in a tech-led selloff (Yahoo Finance). Bullion traded the yield both times, not the print.
What to take from it
- Do not use the flow headline as a timing tool. Fifty-two tonnes of festival demand is real, and it was spent while gold lost a year of range; the data is published after the fact.
- The marginal price is set by the front end and the dollar. Two-year yields are up about 25 basis points this week. Those two move gold intraday; tonnage does not.
- Physical demand is why a deeper break is hard, not why it stops. China has bought 24 months running and India has a wedding season ahead — a bid under the market at lower prices, which is exactly the support the council refuses to call a floor.
- Keep the distinction. A flow report is a fact about the last quarter; a signal is a fact about the next session. Mixing them is how a good macro read becomes a bad entry — see Tuesday's CPI reaction and the vault story.
Nothing in Thursday's tape says the metal has lost its buyers. It says they are not setting the price this week, and that a market at $3,883 with the dollar at a 19-month high has one honest test: whether the front end stops rising. The archive is at /news; the official-sector record sits on /topics/central-banks and /topics/etfs.
Sources: CNBC, Bullionvault, Times of India / IBJA, Reuters via MINING.COM, Bloomberg, World Gold Council — ETF flows, World Gold Council — Q3 outlook, World Gold Council — India Q3, BLS PPI schedule.
Not financial advice. Levels are levels, not predictions; size every position from your own stop.