Gold Reserves by Country
Worldwide Data and Rankings

That list is the entire substance of almost every gold reserves article on the internet. Sort the tonnes from high to low, paste a table, maybe write a sentence about hedging risk. The ranking table is interesting for entirely different reasons than the numbers themselves, and almost nobody writing about it gets into any of them.

The top four names have not changed since the Bretton Woods system collapsed. Check 1975. United States, Germany, Italy, France. Check 1985, 1995, 2005, 2015, 2025. Same four, same order.

They did not earn those positions through fifty years of accumulation. Before the 2008 financial crisis, advanced economy central banks were net sellers. The UK dumped 415 tonnes between 1999 and 2002. Switzerland sold 1,550 tonnes between 2000 and 2008. After the crisis, the selling stopped, but buying did not start. Holdings just sat there.

What those top-four reserves represent is the Bretton Woods era, 1944 to 1971, when the dollar was convertible to gold at $35 an ounce and European trade surpluses with America could be exchanged for physical metal from the US Treasury. Germany, Italy, and France ran surpluses and took gold. The convertibility window shut in 1971. No comparable mechanism for large-scale gold transfer has existed since.

All the net purchasing activity of the past fifteen years has come from emerging markets. The top of this table is a geological deposit from the mid-twentieth century. Everything happening now is happening further down.

China officially reports about 2,306 tonnes as of Q4 2025. The PBoC reported purchases for twelve consecutive months in 2025, adding roughly 24 tonnes on an official basis for the year.

The World Gold Council's demand estimates consistently exceed what the IMF collects through voluntary country reporting. In Q2 2024, the gap was enormous: 67% of total official sector demand, as estimated by the WGC, was unreported. Goldman Sachs puts global central bank buying at roughly 80 tonnes a month, most of it undisclosed, with trade data pointing heavily toward China.

Where does the evidence for hidden Chinese buying actually come from? Independent researchers tracked UK customs data and found something that commercial logic cannot easily account for. London's bullion market trades 400-ounce bars. The Shanghai Gold Exchange private sector uses 1-kilogram bars exclusively. Large-format bars shipped directly from Britain to China have no private-sector buyer waiting in Shanghai. They do not match the exchange's specifications.

November 2024 made the picture sharper. UK gold exports to China stayed elevated while Shanghai gold was trading at a discount to London. Nobody in the metals business ships gold from a premium market to a discount market to lose money. The bars were not going to a commercial buyer.

Estimated covert PBoC purchases from Q3 2022 through end-2024: approximately 1,800 tonnes. Five times the official number for the same period. Two industry insiders with deep connections gave total holding estimates at different times. One in late 2015 said roughly 3,300 tonnes. Another in June 2024 said approximately 4,000 tonnes.

If the higher figure is in the right neighborhood, China is not sixth on this table. The ranking shifts substantially. Nobody can verify it because the PBoC does not submit to external audits and IMF reporting is voluntary. China's track record includes the 2015 episode where reserves jumped from 1,054 to 1,658 tonnes in a single announcement after six years of silence.

Saudi Arabia also buys gold without reporting to the IMF. The WGC's 67% unreported figure covers the entire official sector, not just China.

The ranking table runs on voluntary self-reporting. Two-thirds of the flow bypasses the reporting system.

Bank of England vault delivery queues hit four to six weeks in January 2025. One of the world's largest refineries told clients that London warehouses were empty. The Bank of England acknowledged delays of four to eight weeks.

What happened: Trump tariff expectations from November 2024 onward created a COMEX futures premium over London spot. Arbitrageurs started airlifting gold from London to New York. Between December 2024 and March 2025, about $80 billion of gold crossed the Atlantic. COMEX stocks surged from roughly 530 tonnes to 926 tonnes.

The Bank of England vault sits under Threadneedle Street, nine underground chambers, over 400,000 bars, roughly 100 clients including 72-plus overseas central banks. Most of that gold has an owner. It belongs to central banks, or ETF funds, or foreign treasuries. When arbitrage traders needed physical metal fast and the amount available for immediate release was not enough, the queue appeared.

Gold lending is part of why this situation gets complicated. Central banks lend their vault gold to LBMA bullion banks. The bullion banks sell or redeploy the bars. The central bank's vault record still shows ownership but what it holds is now a claim in fine ounces, not a specific bar sitting on a specific shelf. Whether the physical metal is present on any given day depends on whether the lent bars have been cycled back.

The Bank of England says nothing about how much gold has been lent out. LBMA bullion banks say nothing. Central bank lenders say nothing.

Bank of England vault history gives some sense of scale: about 6,250 tonnes at the start of 2013, dropping to roughly 4,693 tonnes by March 2016. Over 1,500 tonnes flowed out in three years. UK net gold exports during the same period were about 2,500 tonnes. The 1,000-tonne gap came from London's commercial vaults, primarily the HSBC and JPMorgan facilities that house ETF gold. Whether ETF shareholders were aware their underlying metal had moved continents is an open question.

A gold reserve number on the ranking table is a legal and accounting entry. Whether corresponding metal is on the shelf it is supposed to be on, at the moment you need it, is governed by an entirely separate set of facts.

New York Fed vault: 507,000 bars, about 6,331 tonnes, all custodial. The world's biggest pile of monetary gold, and the Fed owns none of it. Germany has roughly 1,200 tonnes stored there. Italy about 1,060 tonnes. The vault peaked above 12,000 tonnes in 1973.

Fort Knox: about 4,580 tonnes, 56% of US federal gold. Congressman Thomas Massie introduced HR 3795 in 2025, calling for full physical count and assay of all three US depositories, plus forensic accounting of every gold transaction for the past fifty years. First time in over sixty years anyone has pushed this through legislation. Trump has commented publicly on whether the gold is there.

Repatriation reporting surfaces a recurring anomaly: countries getting their gold back have found bars with recent casting dates, not original deposit dates. Several possible explanations exist. Bars recast due to wear. Bars re-refined to meet current delivery standards. Or original bars deployed elsewhere at some point and replaced later with fresh ones. No public information confirms which.

Germany repatriated 674 tonnes from New York and Paris between 2013 and 2017. The Netherlands secretly moved 122.5 tonnes out of New York. Hungary increased reserves tenfold in 2018 and brought everything to Budapest. Austria targets 50% domestic storage. Nigeria announced overseas repatriation in 2024. Germany in January 2026 saw renewed political demands to bring back the remaining roughly 1,236 tonnes from New York.

The 2022 freezing of Russia's central bank foreign exchange assets changed the calculus for every reserve manager watching. Gold stored in another country's vault is now understood to carry a risk category that did not operationally exist before: sovereign inaccessibility. France, Russia, and China keep essentially all their gold at home. If the ranking table were redrawn by domestically controlled tonnes rather than reported tonnes, the ordering would shift.

Poland is worth lingering on because its behavior is anomalous for a European central bank. About 90 tonnes added in 2024, 102 tonnes in 2025, total now 550 tonnes. Governor Glapiński started with a 20% target for gold's share of reserves, hit it, raised to 30%, then publicly stated he wants 700 tonnes total. Every time he explains it to media, the phrase is "national security." Poland is on NATO's eastern flank with the Russia-Ukraine conflict next door. The security framing is not abstract for Warsaw.

Kazakhstan added 57 tonnes in 2025, its annual record since 1993. Both Kazakhstan and Uzbekistan buy directly from domestic miners without going through international markets, which avoids price impact and the information leakage that comes with open-market purchases. The Kazakh governor said buying continues until global tensions ease.

Brazil bought 43 tonnes between September and November 2025 after four years of inactivity. Gold at 7% of reserves. Czech Republic has bought for 30 consecutive months, targeting 100 tonnes by end-2028. Serbia wants to double to 100 tonnes by 2030. South Korea and Madagascar signaled interest at the 2025 LBMA Kyoto conference.

The WGC's 2025 central bank survey: 95% expect global gold reserves to keep growing. 43% plan to add to their own holdings, up from 29% the previous year. Zero respondents plan to reduce. Eight years of this survey, and it has never been this one-directional.

The buying wave in aggregate: 1,136 tonnes in 2022 (the all-time record going back to the 1950s), roughly 1,050 in 2023, 1,044.6 in 2024. Fifteen consecutive years of net additions. 2025 came in at 863 tonnes, below the 1,000-tonne threshold of the prior three years.

The 2025 slowdown needs context. Gold went up over 55% that year and broke $4,000 an ounce in October. Central banks were buying into a parabolic price move and chose to moderate the pace rather than stop. That is not a signal of waning demand. That is price sensitivity operating within an unchanged strategic direction.

The Philippines sold a net 29.4 tonnes in 2024, the year's biggest net seller. The Philippine central bank buys from domestic miners as routine and selectively retains. With gold up 50% that year, taking profits is unremarkable.

Russia sits at fifth with 2,332.7 tonnes reported. Underneath: the National Wealth Fund started the war with 405.7 tonnes and by November 2025 had 173.1 tonnes left. Ranking unchanged, reserve eroding.

The UK is instructive over a longer time horizon. Chancellor Gordon Brown in 1999 announced the sale of 415 tonnes. By end-2002, reserves were down to 355.25 tonnes. Gold prices then began a twenty-year climb. The market named his selling window "Brown's Bottom" because it coincided almost exactly with the generational low. Estimated cost to the UK: around £2 billion. UK holdings now sit at about 310 tonnes, ranking roughly seventeenth. That is what one call by one finance minister does to a country's position on this table, and it does not reverse.

Finland trimmed about 10% in 2024, dropping to 44 tonnes, its lowest since 1984, swapping gold for foreign exchange liquidity. Small economies face a genuine tradeoff between gold's preservation properties and the operational utility of liquid FX. The direction of travel is not uniform across all countries.

US gold at 76% of foreign exchange reserves is the global maximum. Three-quarters of America's reserve pool is a metal that cannot directly intervene in currency markets, cannot settle import bills on demand, and requires over-the-counter block processes to liquidate. Good when gold is rising. Constraining when you need cash.

China at about 5% on official figures. Japan around 4%. South Korea under 2%. The largest foreign exchange reserve holders in the world carry the thinnest gold allocations. JPMorgan calculated that bringing every sub-10% central bank to 10% would require about 2,600 additional tonnes of buying, around $335 billion at $4,000 per ounce.

Gold's share of total global reserves crossed 21% in 2024, up 4 percentage points in one year, the largest annual jump in over forty years. Brookings puts the end-2025 figure at roughly a quarter. Most of that expansion came from the gold price rising, not from tonnage increases. Price-driven passive expansion and purchase-driven active expansion mean different things. Both were happening at the same time.

Bank of America data: central banks have been net sellers of US Treasuries since March 2025, with selling picking up after the tariff episode. Dollar share of global reserves down from about 72% around 2000 to roughly 58% now. Gold is the fastest-growing reserve asset class. Same chart, opposite slopes.

JPMorgan forecasts quarterly gold demand from central banks and investors averaging about 585 tonnes in 2026, with prices potentially hitting $5,000 per ounce by Q4. If that materializes, gold's share of reserves inflates further just from the price, without any country buying another gram. Tonnes stay the same on the ranking table. What those tonnes mean changes.

The first is a data crack. Reporting to the IMF is voluntary. No enforcement, no mandatory audit. The WGC's own estimates say two-thirds of official sector gold demand goes unreported. China may hold far more than its stated figure. The ranking positions could be wrong, and not by a small margin.

The second is a physical crack. The table records who reported how many tonnes. It does not record where those tonnes are stored, whether corresponding metal is present in the vault, or whether bars have been lent into the secondary market. Germany has 37% of its gold in New York. Italy has 43% in New York. Reorder the table by domestically held tonnes and it looks different. Gold lending deepens this crack because vault records can show bars that are physically elsewhere.

The third is a time crack. Fort Knox, seventy-two years without an independent audit. Bar casting dates not matching deposit dates, unexplained. London vaults in January 2025 quantifying the distance between paper gold and physical gold as a four-to-eight-week wait.

The top of this ranking table has been static for half a century. The middle is being rewritten by emerging market central banks that mostly still have gold below 10% of reserves and room to buy for years. The largest uncertainty sits in the unreported portion, especially China, where the gap between the official figure and the trade-data-implied figure is wide enough to rearrange the top five if it ever gets confirmed. The repatriation movement is redrawing not the tonnes column but the control column, which the ranking table does not have but probably should.

The table is useful. It is also incomplete in ways that matter. Knowing where it is incomplete is more informative than the numbers it contains.