Gold Is Sliding. Are Central Banks Still Buying the Dip?

Gold has had a rough September. Yahoo Finance's daily series put the metal at $4,219.80 an ounce at 6:47 a.m. ET on September 30, 2026, which was 4.1% lower than a week earlier and 9.4% lower than a month earlier (Yahoo Finance). USAGOLD's daily report put Monday's (September 28) slide at 3.2%, to $4,149 (USAGOLD). On Tuesday, Kitco reported spot gold up 1.36% at $4,170.90 after a weak US consumer-confidence reading (Kitco, Sept. 29). That leaves the price roughly a quarter below this year's peak near $5,590–$5,600; sources differ slightly on the exact high (24/7 Wall St., StoneX).

That brings up a question readers keep asking: when gold drops like this, are the world's central banks still buying, and can they put a floor under the price?

The short answer

On the latest official data, yes, they are still buying, with China and Poland in the lead. Goldman Sachs also estimates that the real pace is well above what gets reported. But reported purchases are running behind last year, two large holders (Russia and Turkey) are net sellers, and no official data yet covers September, the month of the sell-off. Central-bank demand works like a slow, structural cushion under the market. It is not a trading-desk floor that catches every fall.

What the official numbers show

China: 22 months and speeding up

The People's Bank of China (PBoC) reported a 20-tonne purchase in August, which the World Gold Council called "the highest since October 2023." Official holdings reached 2,387 tonnes after rising for 22 consecutive months. Gold now makes up 9% of China's foreign-exchange reserves, up from 8% in July (World Gold Council, Sept. 14). Kitco reported the August addition as about 649,400 ounces (20.2 tonnes) and noted double-digit monthly purchases since May 2026 (Kitco, Sept. 8).

Private demand in China has been strong as well. Kitco reported on September 22 that China's gold imports topped 1,100 tonnes in the first eight months of 2026, which is already more than in all of 2025 (Kitco, Sept. 22). Chinese gold ETFs added 11 tonnes in August, bringing their holdings to 293 tonnes (WGC).

The rest of the official sector: buyers and sellers

The WGC's monthly central-bank statistics (published September 3 and covering July) show net purchases of 23 tonnes. China (20t), Poland (8t) and the Czech National Bank (2t) led the buying. Russia sold 6t, and Turkey, Jordan and Uzbekistan each sold 1t. Reported net purchases for the year to date were about 130 tonnes, against about 160 tonnes over the same period last year (WGC, Sept. 3).

• Poland has bought 90t this year and holds 640t, about 28% of its reserves. Its stated target is 700t.
• Russia has sold 50t so far this year, and Turkey has sold 85t.
• The Bank of Korea announced its first official gold allocation in 13 years, about 2t through gold-backed ETFs.

The gap between reported and estimated buying

Monthly figures only capture what central banks disclose. The WGC's quarterly Gold Demand Trends report also estimates unreported buying. On that basis it put official-sector purchases at a net 289 tonnes in Q2 2026, up 62% year on year (WGC Gold Demand Trends Q2 2026, July 30). That is far above the monthly reported tallies for the same period.

Goldman Sachs goes further. In research reported by Kitco on September 23, analysts Lina Thomas and Daan Struyven estimated that China bought about 35 tonnes in July, roughly 75% more than it disclosed. They reached that figure by tracking gold moving through London's over-the-counter market into domestic vaults and custodians. They put total central-bank buying at 44 tonnes for July and noted that gold held by the Bank of England for foreign central banks rose 63 tonnes that month. Goldman's fair-value forecast of $4,900 by the end of 2026 "assumes continued strong central bank demand," together with a recovery in ETF demand (Kitco, Sept. 23).

These are estimates built from trade and custody flows. They are not confirmed purchases.

Why that doesn't stop the price from falling

If official buyers are so persistent, why did gold still drop so hard? The answer is scale and timing. Central banks buy tonnes per month, while the paper market moves far more money every day. The WGC reported average daily gold trading volume of about $430 billion across all market segments in August (WGC, Sept. 9). A 20-tonne monthly purchase is worth roughly $2.7 billion at $4,200 an ounce (our calculation). Over a year that matters for supply, but it is small next to a single day's turnover.

Faster money has been driving the day-to-day moves. 24/7 Wall St. called Monday's drop "a purely macro, paper-market repricing." It cited October Fed rate-hike odds that jumped to 70.3% and a 10-year Treasury yield of 5.24% (24/7 Wall St., Sept. 30). After weaker US job-openings and consumer-confidence data on Tuesday, USAGOLD reported that those odds had dropped to roughly 51% from about 71% a day earlier (USAGOLD, Sept. 30). Western ETF investors added a record $18 billion in August and lifted global holdings to 4,189 tonnes. That kind of buyer can reverse quickly when rates rise, and the WGC itself attributed part of August's inflows to "momentum" (WGC, Sept. 9).

StoneX's Fawad Razaqzada wrote on September 28 that gold was still up 6.4% for the third quarter through September 25, despite a 4% drop in September. He described $4,000 as key support in his technical outlook (StoneX, Sept. 28). That is his view and we are not endorsing it as a forecast. This week's swings have since eaten into the quarterly gain, and the final Q3 number depends on today's close.

What we know vs. what we're inferring

Known (published data)

• The PBoC reported buying 20t in August, taking holdings to 2,387t and extending its streak to 22 months.
• Reported central-bank net buying was 23t in July and about 130t for the year to date, compared with about 160t a year earlier.
• Russia and Turkey are net sellers in 2026.
• Gold fell about 9.4% in the month to September 30 on Yahoo Finance's series.

Inferred or estimated

• That actual central-bank buying is well above reported figures. This rests on Goldman's estimates and the WGC's unreported-demand estimate.
• That central banks bought the September dip. There is no official September data yet.
• That official demand limits the downside. This is plausible over months, but it did not prevent this month's drop.

What to watch next

• Friday, Oct. 2: US jobs report for September at 8:30 a.m. ET / 9:30 a.m. UTC-3 (BLS). This is a key input for Fed rate-hike odds.
• Around Wednesday, Oct. 7: PBoC reserve data for September. China usually publishes month-end reserves around the 7th, but the date is not confirmed and could move around the National Day holiday. It will be the first hard evidence of whether China bought through the sell-off.
• Wednesday, Oct. 7: FOMC minutes from the September 15–16 meeting, at 2:00 p.m. ET / 3:00 p.m. UTC-3 (Federal Reserve).
• Early October: WGC releases of September ETF flows and August central-bank statistics. Exact dates are not yet confirmed.
• Wednesday, Oct. 14: US CPI for September at 8:30 a.m. ET / 9:30 a.m. UTC-3 (BLS).
• Oct. 27–28: FOMC meeting. The press conference is at 2:30 p.m. ET / 3:30 p.m. UTC-3 on October 28 (Federal Reserve).

Educational information only — not financial advice.