Goldman Estimate Lifts Central Bank Gold Buying to 79 Tonnes in August
Key Facts
Goldman Sachs estimated that central banks bought 79 tonnes of gold in August 2026, compared with a monthly average of 17 tonnes before 2022. The estimate puts purchases for the month at more than four times the previous level, but is based on a secondary-source report whose page could not be opened. Separately, the World Gold Council said reported net purchases totaled 39 tonnes in August. The gap between the figures leaves a wide range between estimated total purchases and what appeared in official disclosures. Central bank purchases announced since the start of 2026 through August 31 also totaled 170 tonnes.
China's data reflect continued additions to official reserves, with the People's Bank of China buying gold for a 22nd consecutive month through the end of August. Official holdings stood at about 2,387 tonnes at month-end, according to the World Gold Council. Gold represented 9% of China's total foreign exchange reserves on that date. The estimate attributed to Goldman indicates that China was the largest identifiable buyer in August's calculations, but the figure in the secondary-source report alone does not establish the presence of unreported purchases. The published record of official holdings remains separate from Goldman's estimate of total purchases for the month.
Goldman's forecast for 2026 puts central bank purchases at an average of 50 tonnes a month, compared with 17 tonnes a month before 2022. That means the August estimate of 79 tonnes exceeded the bank's projected monthly average for the year, although the comparison is between an estimate for one month and an annual forecast. On a yearly basis, central banks and other institutions bought 863.3 tonnes in 2025, down 21% from 1,092.4 tonnes in 2024. Still, the 2025 total was well above the annual average of 473 tonnes recorded between 2010 and 2021. The figures show purchases fell from their 2024 peak while remaining above levels from before 2022.
Managing gold reserves involves more than buying decisions: the Dutch central bank announced that it moved about 86 tonnes of gold from New York and Ottawa to London between March and August. The bank said the move was intended to improve the tradability of its reserves and prepare for crises. The announcement describes a specific step to redistribute part of the Netherlands' holdings and does not establish a broader change in inventories at the Federal Reserve Bank of New York or the Bank of England. The move adds an operational dimension to decisions about holding the metal, as storage location affects a bank's ability to access and transact in its reserves. Moving the gold does not, by itself, change the Dutch central bank's total holdings.
August's purchases came in a month when gold rose sharply, gaining 13% in dollars and 8.4% in yuan, its strongest monthly performance since January, according to the World Gold Council. The council attributed the rebound mainly to improving momentum and positioning in gold funds and futures. Prices then fell in September, down 8.5% in dollars and 5.5% in yuan, before stabilizing in early October with the international price remaining above $4,100 an ounce. In EL7 closing data, the GCUSD gold contract ended trading on October 23 at $4,007.9, after closing at $4,160.1 on October 19. The weekly decline shows that strong central bank buying does not prevent short-term price moves.
Official demand expectations point to continued interest in gold in the coming months, according to the World Gold Council's 2026 survey. Some 89% of respondents expected global reserves to rise over the following year, while 45% said they expected to increase their own holdings over the same period. The findings support the council's expectation that central bank demand will remain above its long-term average. Goldman's estimate of monthly buying in 2026 offers another benchmark for assessing those expectations, at an average of 50 tonnes a month. Data for the coming months and subsequent official disclosures are the next dated milestones for comparing expectations with reported purchases.