Why Central Banks Bought Record Gold as Prices Fell — and What Indian Savers Should Do
Markets · Gold · Global reserves and Indian savers
Why Central Banks Bought Record Gold as Prices Fell — and What Indian Savers Should Do
Official-sector buying hit a record 289 tonnes in the second quarter of 2026, in the same quarter gold posted its steepest quarterly fall in a decade. This is a check of what the data confirms, what it does not, and how the split changes an Indian household’s gold decision.
Gold in India is trading near ₹1,56,000 per 10 grams for 24 karat, close to the most Indian households have ever paid for it. Global gold, over the same stretch, sits roughly a fifth below where it stood in January. Both things are true at once, and the gap between them is where most of the confusion about this year’s market lives. A second contradiction sits underneath it: central banks bought more gold in the second quarter of 2026 than in any second quarter on record, during the three months the metal fell hardest in a decade.
Quick Summary
Central banks bought a record 289 tonnes in the second quarter of 2026, a 62% jump year on year, in the same three months gold suffered its worst quarterly slide in ten years. The popular reading of that, ordinary investors fleeing while governments quietly accumulate, does not survive the flow data. Global gold ETF flows in the first half of 2026 were positive at US$8 billion. The selling came from one region alone, and Asian funds recorded their strongest half on record.
The 2026 gold market split in two, and most explanations cover only one half
Two things happened at once this year, and they point in opposite directions. Price collapsed. Sovereign demand accelerated. Any explanation that only accounts for one of them is incomplete.
On the price side, gold reached an all-time peak of nearly US$5,595 an ounce in January 2026, then unwound. By late June it briefly traded under US$4,000, a drawdown of close to 30% in five months. The drivers were unglamorous and well documented: a stronger dollar, higher real yields, and the market reading the new Federal Reserve chair as hawkish while the Middle East conflict pushed inflation expectations up rather than down. Gold pays no coupon, so when Treasury bills start paying a real return, holding it costs something and money moves.
On the demand side, none of that appeared to matter to reserve managers. The World Gold Council reported net official-sector purchases of 244 tonnes in the first quarter and 289 tonnes in the second, the strongest second quarter in its data series. Central banks were buying into the fall, not out of it.
Five claims doing the rounds this year, checked against the record
A cluster of assertions has travelled widely through finance commentary this year, usually bundled as proof that something large is imminent. Taken one at a time, two hold cleanly, two describe real events with a misattributed motive, and one matches no published number we could locate.
| The claim | Verdict | What the record actually shows |
|---|---|---|
| Poland, Kazakhstan and China are aggressively buying gold | Confirmed | Poland added 51 tonnes in Q2 2026 alone, reaching 632 tonnes by end-June, against a stated 700-tonne target. China’s central bank added 10 tonnes in May, extending a streak beyond 18 consecutive months. |
| Retail investors pulled US$18 billion out of gold ETFs | Unverified | No World Gold Council release matches that figure for 2026. June’s global outflow was US$8.9 billion. North American funds lost US$7.7 billion across the whole first half. Globally, H1 flows were positive at US$8 billion. |
| China banned paper gold trading at four major banks | Real event, contested motive | ICBC, Postal Savings Bank, Ping An Bank and China Guangfa Bank all exited retail precious-metals trading tied to the Shanghai Gold Exchange from 24 July 2026. The stated trigger was leverage and volatility risk, not a push into bullion. |
| US dollar devaluation strategies are driving capital into gold | Not supported | No announced policy exists. Through the first half of 2026 the dollar strengthened, and that strength was among the main forces pushing gold down. |
| A structural shift toward gold-settled international trade | Pilot stage only | The BRICS “Unit”, launched as a research pilot on 31 October 2025, is backed 40% by gold and 60% by member currencies. Russian officials have pointed to 2030 for operational status. |
It was not retail against nations. It was North America against Asia.
This is the correction that changes the whole story. The framing of ordinary investors panicking while wise governments accumulate is emotionally satisfying and geographically wrong.
Across the first half of 2026, North American gold ETFs shed US$7.7 billion, the region’s weakest first half since 2013. In the same six months, Asian-listed funds took in US$12 billion, their strongest first half on record, and European funds added US$3.2 billion in healthy inflows. Retail and institutional investors in Mumbai, Shanghai and Tokyo were doing the opposite of retail and institutional investors in New York.
Zoom out and the ETF story softens further. Global holdings peaked at 4,176 tonnes on 27 February 2026, dipped through the spring, and were back at 4,068 tonnes by the end of July after US$3 billion of inflows reversed two months of redemptions. Year to date through July, global gold ETFs had taken in US$11 billion. That is not an exodus. It is a rotation, and it moved east.
Why gold’s 27 percent reserve share is smaller than it sounds
In June 2026 the European Central Bank confirmed something that had been speculated about for two years. Measured at market prices, gold made up 27% of global central bank reserve assets at the end of 2025, up from 20% a year earlier, while the share held in US Treasuries fell from 25% to 22%. For the first time in the modern dollar era, bullion outranked American government debt in official portfolios.
That is a genuine milestone, and it is also where careful readers should slow down. A share is a ratio, and a ratio moves when either the numerator or the denominator changes. Gold’s price roughly doubled over two years. Existing holdings were revalued upward without a single extra bar entering a vault.
The valuation effect, and why the ECB stripped it out
The ECB ran the same calculation holding gold at its end-2023 price. On that basis gold accounts for about 16% of reserves, the euro also 16%, and US Treasuries 26%. In other words, roughly half the headline shift is price appreciation rather than reallocation. The reallocation is real, but it is slower and smaller than the 27% figure implies on its own.
Two further details deserve attention. Dollar-denominated assets in aggregate still command about 42% of global reserves, nearly three times the euro’s 15%. And central banks now hold more than 36,000 tonnes between them, approaching the roughly 38,000 tonnes of the Bretton Woods era. Direction of travel matters more than any single year’s ratio.
How the world got here, in six dated steps
The accumulation cycle has a clear starting gun, and it was not a market event.
What reserve managers themselves say they are buying against
Rather than infer motives, it is worth reading the survey where central banks state them. The World Gold Council’s 2026 Central Bank Gold Reserves Survey drew responses from 76 institutions, the largest participation in its nine-year history.
Notice what is absent. No respondent category cites imminent monetary collapse, a return to a gold standard, or a coordinated dollar replacement. The stated reasons are prosaic risk management: an asset with no issuer, no counterparty and no sanctions exposure.
The buyer list is also widening in a way that supports the boring explanation rather than the dramatic one. Guatemala, Indonesia, Malaysia, Cambodia, Uganda and Kenya all appeared as buyers in the first quarter of 2026, several for the first time in institutional memory. Small reserve managers adding a first tranche of gold look like diversification, not insurgency. And the flow runs both ways: Turkey, which had added roughly 220 tonnes since 2022, sold or lent about 130 tonnes in early 2026 to manage currency pressure. Even in an accumulation cycle, gold gets spent when a country needs liquidity.
China’s paper gold shutdown: a real event with a misread motive
The China claim is the one most worth understanding properly, because it is accurate on its facts and wrong in its interpretation.
From 24 July 2026, ICBC stopped offering individual precious-metals trading linked to the Shanghai Gold Exchange. Postal Savings Bank of China, Ping An Bank and China Guangfa Bank made parallel exits. Customers were told to close positions, sell, or take physical delivery. Margin requirements on some products had already been raised to between 120% and 140%, which erases leverage entirely, and reportedly as high as 190% on certain ICBC contracts.
The popular reading is that Beijing is herding citizens into bullion. The sequence argues otherwise. New retail accounts for these products have been frozen since late 2020, immediately after the “Crude Oil Treasure” episode in which a Bank of China product tied to oil futures inflicted losses on roughly 60,000 retail customers when crude briefly traded negative. What happened in mid-2026 was the final chapter of a five-year wind-down, accelerated by a metal that had swung from US$5,600 to under US$4,000 in months. Regulators were closing a leveraged retail product, not opening a bullion window.
The pattern worth carrying away
Every one of these claims is built on something real. The distortion is in the join. A leverage crackdown becomes a sovereign gold strategy, a regional ETF rotation becomes a global retail panic, and a research pilot becomes an imminent settlement system. When a chain of true facts produces an extraordinary conclusion, check the joins rather than the facts.
What we know
- Central banks bought a net 244 tonnes in Q1 2026 and a record 289 tonnes in Q2 2026, a 62% year-on-year rise, per World Gold Council Gold Demand Trends.
- Full-year 2025 official buying was 863 tonnes. The Council’s 2026 forecast is approximately 850 tonnes, against a pre-2022 average near 500 tonnes.
- Gold peaked near US$5,595 an ounce in January 2026 and traded below US$4,000 by late June, recovering to roughly US$4,370 to US$4,450 in early September.
- The ECB confirmed gold at 27% of global reserve assets at end-2025 against 22% for US Treasuries, while noting that at constant end-2023 prices the figure is closer to 16%.
- Global gold ETF flows were positive in H1 2026 at US$8 billion. North America was the only region in net outflow at US$7.7 billion, its weakest half since 2013.
- The RBI held 880.52 tonnes as of 31 March 2026, virtually unchanged in weight, with gold at 16.7% of India’s foreign exchange reserves against under 6% in March 2021.
What is still unclear
- The origin of the widely quoted US$18 billion ETF outflow figure. It matches no World Gold Council monthly, quarterly or half-yearly release for 2026 that we could identify.
- China’s true accumulation pace. Reported additions run through IMF disclosure channels, and analysts have long argued that off-book purchases may not appear in published tonnage.
- Whether the BRICS Unit becomes operational. Russian officials have indicated 2030 as a target, with pilots possible before the end of 2026, and no member state has committed to gold settlement as policy.
- Where the price goes next. Roughly 298 tonnes of ETF gold was estimated to be sitting at a loss at mid-2026 price levels, which can create selling pressure on any rally.
- Whether the RBI resumes active buying. The physical stock has barely moved in a year, and the central bank has publicly denied 2026 reports that it sold gold to defend the rupee.
Where an Indian saver actually sits in this
Return to the contradiction this piece opened with. Indian prices sit near records in rupees only because the rupee weakened while the metal fell in dollars. A domestic buyer never gets the clean international chart, and someone buying at ₹1,56,000 per 10 grams because central banks are stockpiling is taking currency risk they have not priced.
The useful question is not whether to own gold. It is how much, and the answer is a range rather than a number.
Thin
Core
Conviction
Concentrated
Exposed
Worked example: what the 2026 round trip did to a real allocation
Take an investor with ₹20 lakh invested and 12% in a gold ETF, so ₹2.4 lakh, bought when the metal was near its January highs. A 28% fall in the underlying takes that holding to about ₹1,72,800, a loss of ₹67,200. Against the full portfolio that is a 3.4% dent, uncomfortable but survivable. Now run the same fall on a 30% allocation of ₹6 lakh. The holding drops to ₹4,32,000 and the portfolio takes a hit of 8.4% from one asset. The metal behaved identically in both cases. Only the position size decided whether it was a hedge or a problem.
Six ways to hold gold in India, compared on what actually costs you money
Format matters more than most buyers assume, and the tax clock is where the biggest gap sits. Listed gold ETFs reach long-term treatment in twelve months. Physical gold, digital gold and gold fund-of-funds need twenty-four for the same 12.5% rate.
| Format | Cost to get in | Ongoing cost | Long-term clock | Practical liquidity |
|---|---|---|---|---|
| Jewellery | 3% GST on value plus 5% GST on making charges | Locker or insurance, plus making charges lost on resale | More than 24 months | Weakest. Jeweller buyback deducts making and often purity |
| Coins and bars | 3% GST | Storage and insurance | More than 24 months | Moderate. Dealer spread applies on both sides |
| Gold ETF (listed) | No GST. Brokerage only | Expense ratio, typically well under 1% a year | More than 12 months | Strongest. Sells on exchange during market hours |
| Gold fund or FoF | No GST | Two layers of expense in a fund-of-funds structure | More than 24 months | Good. Redemption at NAV, no demat account needed |
| Digital gold | 3% GST | Platform spread and storage fee | More than 24 months | Good on paper, but outside SEBI and RBI product regulation |
| SGB (secondary market) | No GST. Exchange price, often at a discount or premium | None | More than 12 months if sold on exchange | Thin. Fresh issuance stopped, so only old tranches trade |
The single most common mistake
Applying one holding-period rule across every gold holding in the same tax return. A listed ETF and a secondary-market SGB cross into long-term treatment at twelve months. Physical, digital and fund-of-fund gold need twenty-four. Selling a gold fund at month eighteen because “gold is long-term after a year” converts a 12.5% liability into a slab-rate one.
The habits that separate a gold allocation from a gold bet
Questions readers are asking about the 2026 gold market
Why did central banks buy record gold in 2026 while the price was falling?
Reserve managers are not trading for return. In the World Gold Council’s 2026 survey, 92% cited interest-rate volatility and 90% cited gold’s performance during crises, while 95% of emerging-market respondents pointed to geopolitical instability. Gold has no issuer and cannot be frozen by another government. A lower price makes a planned allocation cheaper to complete, so falling prices encourage rather than deter this kind of buying.
Did investors really pull US$18 billion out of gold ETFs in 2026?
No published World Gold Council release for 2026 matches that figure. The actual numbers are that June saw US$8.9 billion of global outflows, North American funds lost US$7.7 billion across the first half, and global flows for the half were positive at US$8 billion because Asian funds took in US$12 billion. By the end of July, year-to-date global inflows stood at US$11 billion.
Has gold really overtaken US Treasuries as the world’s top reserve asset?
At market prices, yes. The ECB reported gold at 27% of global reserve assets at end-2025 against 22% for Treasuries. Held at constant end-2023 gold prices, the picture is different: gold falls to about 16%, level with the euro, and Treasuries rise to 26%. Roughly half the shift is price revaluation. Dollar assets in aggregate still make up about 42% of reserves.
Why is gold expensive in India when global prices have fallen?
Because the domestic price combines the international price with the rupee-dollar rate and import duties. Gold fell from roughly US$5,595 an ounce in January to under US$4,000 in June, but rupee depreciation over the same period absorbed much of that fall. The result is 24 karat gold trading near ₹15,600 per gram in early September 2026 even as the dollar chart shows a substantial decline.
Is China forcing its citizens to buy physical gold?
The evidence points to leverage control rather than promotion. ICBC, Postal Savings Bank, Ping An Bank and China Guangfa Bank all withdrew retail Shanghai Gold Exchange trading services from 24 July 2026, but new accounts in these products had been frozen since late 2020 following the Crude Oil Treasure losses. Margin requirements were raised to 120% to 140%, which removes leverage entirely. Physical purchase and unleveraged ETFs remain freely available.
How much gold does the Reserve Bank of India hold?
The RBI held 880.52 tonnes as of 31 March 2026, barely changed from 879.58 tonnes a year earlier. What changed sharply is the share: gold rose to 16.7% of India’s foreign exchange reserves from under 6% in March 2021, driven by the higher gold price and a weaker rupee rather than new purchases. Around 77% of that gold is now stored domestically, up from roughly 59%.
Is a gold ETF or physical gold better for tax in India?
Listed gold ETFs are more tax-efficient for a patient investor. They qualify for long-term treatment after twelve months, while physical gold, digital gold and gold fund-of-funds require twenty-four months for the same 12.5% rate. ETFs also avoid the 3% GST charged on physical and digital purchases and carry no making charges. Physical gold retains one advantage: it can be pledged for a gold loan without being sold.
How much of my portfolio should be in gold?
There is no regulatory answer, but common advisory practice puts the hedge band at 5% to 10% of investable assets, excluding jewellery you wear rather than hold as an investment. Above roughly 15% your returns start depending heavily on one metal and one currency pair. The 2026 drawdown of nearly 30% is the stress test to run against any allocation before deciding.
The short version
None of the popular claims is a lie, but stacked together they compress. Central banks genuinely are accumulating at roughly double their pre-2022 pace, and gold genuinely has passed US Treasuries in reserve portfolios at market prices. What the compressed version leaves out is that half that shift is revaluation, that the ETF selling came from one region while Asia bought a record amount, that China’s paper gold closures were a leverage crackdown five years in the making, and that gold fell nearly 30% while all of this was happening. Central banks can afford a 30% drawdown across a thirty-year horizon. Most households buying at ₹1,56,000 per 10 grams cannot. Size the position first, and let the tonnage headlines be interesting rather than instructive.
