An analysis by the Gold Observatory showed that the announced net changes in central banks’ gold holdings amounted to around 130 tons from the beginning of 2026 through the end of July, compared with around 160 tons during the same period in 2025. The figure represents the increase after deducting sales, rather than the total quantities purchased by banks before accounting for what they sold.
These figures are based on the latest World Gold Council data on banks that disclosed their holdings. Therefore, they do not necessarily represent the full activity of the official sector, as some disclosures may be delayed or previous data may be revised.
Purchases and Sales Moving in Divergent Directions
The National Bank of Poland led the announced buyers, adding around 90 tons and raising its holdings to nearly 640 tons, or around 28% of its total reserves, bringing it closer to its target of 700 tons. The People’s Bank of China ranked second, adding around 60 tons and bringing its announced holdings to approximately 2,366 tons. China added around 20 tons in August, but this increase is not included in global net purchases through July.
Uzbekistan also increased its holdings by around 40 tons, despite selling one ton in July, bringing its holdings to approximately 431 tons, with gold accounting for around 87% of the value of its reserves. Kazakhstan added around 29 tons, the Czech Republic 12 tons, Malaysia 6 tons, and Bolivia around 2 tons.
By contrast, Russia recorded net sales of around 50 tons, and Turkey around 85 tons. These figures confirm that reserve management does not move according to a single global trend; buying may coincide with selling depending on liquidity needs and asset management.
Two Figures Measuring Different Activities
The World Gold Council distinguishes between announced changes in banks’ holdings and its broader estimate of net official-sector demand, which may include activity that was not disclosed in detail when the data was prepared. While announced net purchases amounted to around 102 tons in the first half of 2026, the council estimated net official-sector demand at around 345 tons for the same period, including approximately 57 tons in the first quarter and 289 tons in the second quarter.
The estimate of 345 tons does not mean that the council identified the countries that purchased all of this quantity; it is an aggregate figure subject to revision, and the first-quarter estimate was reduced after some flows were reclassified as over-the-counter trading. In practical terms, the 130-ton figure measures the announced changes available through July, while the 345-ton figure represents a broader estimate for the first half of the year.
Changing Location Does Not Mean Selling the Reserve
The cases cited in the analysis highlight the importance of distinguishing between net sales and the redistribution of gold. In the Netherlands, the central bank redistributed around 86 tons of its reserves between March and August 2026. It sold around 59 tons in New York and bought a similar quantity in London, and it also physically moved more than 27 tons from North America to Zeist, then moved a similar quantity from Zeist to London.
As a result, London’s share of the Netherlands’ reserves rose from 18.1% to 32.1%, while New York’s share declined from 31.3% to 18.5%, whereas total reserves remained unchanged at 612.4 tons. The objective was to improve tradability and readiness for use during crises, not to reduce the balance.
As for the Bank of France, between July 2025 and January 2026 it sold around 129 tons of old bars held in New York and bought a similar quantity of bars that met its standards and were stored in Paris. France’s reserves remained at around 2,437 tons, while a capital gain related to the transaction of approximately 12.8 billion euros was recorded, of which around 11 billion euros was recognized in the 2025 accounts.
What Matters in Practice?
The data shows that interpreting purchase and sales figures requires separating three elements: the net change in the quantity of gold, its storage location, and the type of bars used. Differences in the timing and location of execution may temporarily affect liquidity or prices, but they do not automatically mean a decline in reserves.
The results of the World Gold Council survey support this distinction: 45% of participating banks expected to increase their reserves within a year, compared with 1% that expected to reduce them. Nevertheless, these are participants’ expectations and are not a substitute for actual purchase and sales data. The survey also indicated that 57% of participating central banks that hold gold store some of it with the Bank of England, compared with 49% that hold some within their own countries, with the two options not being mutually exclusive.