In January 2026, gold reached its highest price ever, about $5,589 an ounce. Most people would stop buying at such a price, but central banks did not. Between April and June 2026, they bought 289 tonnes of gold, the most ever for that period, even though the price had fallen by about 16%. Why? Because central banks are not buying gold to make a profit. They are buying it for safety, like insurance, in case their other savings are frozen, lose value or are not paid back. This article explains the problem they are trying to solve, how gold helps, what the evidence shows, where it falls short, and what India should learn from it.

Reserves that can be frozen

A central bank’s reserves are, for the most part, claims on someone else: a deposit at another country’s central bank, or a bond issued by another government. That works well as long as the issuer stays solvent, the currency holds its value, and the relationship stays friendly. The last few years have tested all three at once. The turning point was 2022, when the Ukraine war began and Russia’s central bank assets held abroad were frozen. The ECB has noted that the surge in official gold demand started in 2022, and its analysis, as summarised by U.S. Global Investors, found that in five of the ten largest annual jumps in gold’s share of reserves since 1999, the country had been sanctioned that year or the one before. The dollar’s share of global reserves slipped to 46% by the end of 2024.

So the problem is not how to earn a return on reserves. It is how to hold part of a country’s savings in a form that no foreign government can freeze, no issuer can default on, and no central bank can print more of. Framed like that, the question in the headline, why buy gold at record prices, is the wrong one. The price is a secondary variable.

Accumulate gold

The strategy has two parts. The first is to accumulate gold as a strategic reserve asset. The World Gold Council’s 2026 survey of central banks lists gold’s performance in crises, its value as a diversifier and its role as an inflation hedge as the main reasons to hold it, with geopolitical risk and reserve diversification as the reasons to hold more. The second part is newer: hold the gold at home, or at least in more than one place. In the same survey, the Bank of England is still the most popular vault at 57% of respondents, but domestic storage is close behind at 49%, and the Swiss National Bank’s share fell to 6% from 12%. A bar in a foreign vault is still, in the end, an asset that depends on a foreign custodian’s goodwill.

Where and when: 2022 to 2026, led by a few institutions

The trend is global, but the buying is concentrated. Central banks have averaged about 1,000 tonnes a year over the past four years, against roughly 500 tonnes a year in the previous decade. In the first half of 2026, Poland was the largest reported buyer at 82 tonnes, followed by Uzbekistan (41), China (40) and Kazakhstan (27). Poland’s purchases are part of a plan to reach 700 tonnes, which analysts link to security concerns on NATO’s eastern flank. China has reported additions for 21 straight months, taking its declared holdings to about 2,346 tonnes.

The second story is about location. India’s central bank brought 168 tonnes home in FY26, the third year in a row of large transfers, raising the domestic share of its gold to 77% from 38% in March 2023.

For whom: surplus-rich, exposed, and not short of cash

The strategy suits a specific kind of central bank. It is one that has more reserves than it needs for day-to-day defence of its currency, and that faces a geopolitical risk it cannot diversify away with more dollars. Poland (a frontline state) and China (a country wary of dollar exposure) fit that description, even though they worry about opposite things. The survey suggests the appetite is widespread: a record 89% of respondents expect global central bank gold reserves to rise over the next year, and a record 45% expect their own to grow. Only 1% expect a decline.

The strategy does not suit the opposite kind of central bank, and the data shows it. In the first half of 2026, Turkey was the largest net seller at 83 tonnes and Russia sold 44. Analysts read those sales as a response to fiscal and currency pressure, namely funding war spending under sanctions and supporting the lira, though neither central bank has confirmed that reading; see the Visual Capitalist summary. Gold is a reserve for those who can afford to keep it, and the first thing sold by those who cannot.

Under what conditions: law, liquidity and time horizon

Three conditions make this work. The first is the legal room. Many central banks are allowed to hold gold as a reserve asset, and some are required to. India is one of them: Section 33 of the RBI Act requires the Issue Department, which backs the currency in circulation, to hold gold and foreign securities worth at least Rs 200 crore, of which gold must be at least Rs 115 crore, and it requires that not less than seventeen-twentieths of that gold be held in India. Of the RBI’s 880.52 tonnes, 312.32 tonnes back notes in circulation, so the legal floor for domestic storage on that part is about 265 tonnes. That is a rough calculation of mine, but it makes the point that the RBI’s actual domestic holding of 680 tonnes goes well beyond what the statute demands.

The second condition is a stable liquidity position. Buyers are those who do not need to sell in a crisis. The third is a long time horizon, because gold pays no interest and its price can fall sharply. Gold peaked at $5,589 on 28 January 2026 and was quoted around $4,100 at the end of September, about a quarter lower. An institution that cannot live through a drawdown of that size should not be holding gold as a strategic asset.

Evidence: what the numbers show

The evidence is strongest on scale and consistency. Central banks bought 863 tonnes in 2025, a slow year by the new standard. In the second quarter of 2026 they bought 289 tonnes, up 62% on a year earlier and a record for a second quarter, in a quarter when the price fell about 16%. At the level of the whole reserve system, gold made up about 20% of global official reserves at market prices at the end of 2024, overtaking the euro’s 16%. Those purchases were roughly a fifth of annual mine output, which suggests the buyers are now part of the price they are paying.

The evidence on price sensitivity is mixed. The weak first quarter came at the peak, and the strong second quarter came after the fall, which looks like buying the dip. The fair reading is that central banks are price-tolerant rather than price-blind: the price affects when they buy, but not whether.

Limitations: what still holds, and what has changed

The first limitation is the oldest: gold pays no interest. That was the argument the UK Treasury used when it sold 395 tonnes between 1999 and 2002 at an average of about $275 an ounce, near a twenty-year low. The argument is still valid in the narrow sense, since gold earns nothing while it sits in a vault, but the episode shows what it costs to act on it at the wrong time. I found no source showing that the yield problem has been solved. There is a partial workaround: the RBI holds about 2.8 tonnes as gold deposits, but that is a rounding error next to 880 tonnes.

The second limitation is volatility, which is still valid and was on display this year. The third is data quality. The first quarter’s early estimate of 244 tonnes was later revised down to 57, and the first-half total of 345 tonnes was the lowest since 2022. The World Gold Council also flags that unreported buying remains elevated, so official figures understate the picture in one direction and can overstate it in another.

The fourth is custody risk, and this is where a new answer has emerged. Repatriation and multi-location vaulting did not feature much in earlier reserve management, and they are a direct response to the freezing episode. The fifth is crowding: when the biggest buyers are also a large share of demand, the price and the policy become entangled, and a fiscal shock at a large holder can push the price the other way, as Russia and Turkey showed.

If India has to adapt it, not adopt it

India cannot simply copy Poland or China, because India’s position is different in ways that matter. It imports 700 to 1,000 tonnes of gold a year for private use, which strains the current account. Its households own an estimated 25,000 to 30,000 tonnes, far more than the central bank’s stock. And the RBI has to defend the rupee, which means reserves must stay liquid. With those differences in mind, here are constructive suggestions. They are my own reading of the evidence, not settled policy.

First, separate the price effect from the policy decision. The RBI’s gold share of reserves rose to 16.7% at the end of March 2026 from under 6% in March 2021, but the physical stock moved by less than a tonne in FY26. A share that rises because of price, and not because of choice, can drift beyond what the RBI intends. Publishing a target range in both tonnes and value, with a stated rule for what happens when revaluation pushes the share out of the range, would make the policy legible and stop the headline number from doing the RBI’s talking.

Second, buy on a rule, not on a mood. The RBI added nearly 58 tonnes in FY25 and almost nothing in FY26, and the public data cannot say whether that was about price, the rupee or something else. A pre-announced pace, in the way Poland has a 700-tonne target, adjusted for the state of the foreign-exchange buffer, would reduce the risk of mistiming and let markets read the policy.

Third, keep repatriating for security but not for its own sake. Since the law requires only about 265 tonnes of the note-backing gold to be at home, the rest of the repatriation is a policy choice. Gold stored abroad is also easier to lend or swap when the RBI needs foreign currency, so the aim should be a deliberate split between security and liquidity, not a race to 100% domestic.

Fourth, and most importantly for India, fix the household-gold problem. The Gold Monetisation Scheme mobilised only about 39 tonnes in over a decade, and the government discontinued the medium- and long-term deposit options on 26 March 2025. A blog-level summary puts interest on the remaining short-term deposits at 0.5% to 0.6%, which is hard to see as a reason for a household to hand over gold. Reports say the government is considering a jeweller-led revamp that could mobilise more than 1,000 tonnes. The lesson from the central banks is that gold is valued for being safe and outside anyone’s control. A scheme that asks households to give up that safety for a tiny return will not work, whereas one that offers convenience, purity testing and a meaningful return might. Note that mobilised household gold mostly reduces imports and does not count as RBI reserves unless the RBI itself takes it.

Finally, India should be wary of copying the volume race. Poland and China are building holdings in a world where the price has already reset. For India, the more valuable adaptation is not more tonnes but a clearer rulebook: a stated allocation range, a rule-based pace, a considered split between home and abroad, and a household-gold policy that works. The central banks buying at these prices are not betting on gold. They are betting on the world gold is meant to survive, and a good policy for India would be built on that logic, not on the price.