Goldman Sachs Estimates: China's Central Bank Actual Gold Purchases in July Were About Twice the Official Figures
nashnova research
Goldman Sachs estimates the PBOC bought ~35 tonnes of gold in July — 1.75× the officially reported 19.9 tonnes; global central-bank buying has rebounded to ~91 tonnes per month, but overall gold demand is running below historical averages, making central banks the critical pillar under prices.
How big is the gap between China's real and reported gold buying?
Goldman's monthly model puts the PBOC's July gold purchases at ~35 tonnes. The official figure: just 19.9 tonnes — actual buying was roughly 1.75× the reported number.
This is not a one-month quirk. Through most of 2025, official data barely moved while Goldman's estimates showed far higher volumes; only in March 2026 did official figures jump — and they still lagged reality.
This means → China has been steadily accumulating gold through undisclosed channels (purchases not reflected in official reserve disclosures). The reported number is just the visible tip.
Could Goldman's own estimate still be too low?
Bank of England custodial gold holdings rose 63 tonnes in July — more than the drop at the New York Fed. That implies some central-bank buying happened outside Goldman's London OTC tracking model.
Separately, the Dutch central bank recently moved 90 tonnes from the New York Fed to the UK, adding noise to the data.
In plain terms = Goldman tracks London OTC flows, but some central banks buy through other routes. The true global total could exceed the 44-tonne headline estimate.
Global gold demand is weakening — so who is holding up the price?
Société Générale data shows H1 2026 global gold demand totalled 1,951 tonnes, below the ten-year average of roughly 2,157 tonnes. To match 2025's full-year demand, H2 would need an additional ~3,076 tonnes.
The softness is structural: jewellery demand remains suppressed by high prices; investment demand is elevated but largely coasting on momentum from the 2025 buying wave.
This means → With both the consumer and investment sides cooling, central-bank buying is virtually the only demand source still expanding — its marginal support for gold prices is becoming more critical.
Why did central-bank buying collapse in Q1 and then snap back?
Q1 2026 global central-bank purchases fell to just 57 tonnes — the lowest quarterly reading since 2020 — dragged down by selling from Turkey and a handful of other banks.
Q2 rebounded sharply to 288 tonnes. On a three-month seasonally adjusted basis, the current trend sits at roughly 91 tonnes per month, well above the pre-2022 average of 17 tonnes.
This reflects a pattern: central-bank gold buying is lumpy, not steady — short-term swings can be dramatic, but the medium-term trend remains far above pre-2022 norms.
Which central banks are doing most of the buying?
Global central-bank gold purchases are highly concentrated: the top five buyers typically account for over 80% of the total. In July alone, China contributed ~35 tonnes out of a global ~44 tonnes.
Poland stands out: ~90 tonnes in 2024 (28% of global CB buying), ~102 tonnes in 2025 (31%), and ~82 tonnes year-to-date in 2026 (43%). Its gold reserves have climbed to 11th globally. Since early 2018, Poland has added roughly 529 tonnes.
In plain terms = the global central-bank gold rush is not broad-based. A handful of countries are stockpiling aggressively — China and Poland are the two that stand out most.
What does this data gap mean for the market?
The persistent divergence between China's actual purchases and its official data means the market's estimate of total central-bank buying carries a systematic downward bias.
Whether this gap is gradually corrected in official disclosures will be a key variable for gauging the true strength of gold demand.
This means → If you rely on official figures to gauge how much gold central banks are buying, you are likely underestimating the real purchasing power — and that purchasing power is the single most important floor under current gold prices.
市场有风险，内容仅供研究参考，不构成投资建议。