In late January 2026, gold did something loud. It touched an intraday record of roughly US$5,595 per troy ounce, capping a two-year run that turned the oldest asset in finance into one of the best-performing things you could hold [source: World Gold Council, 2026]. Then it reversed. By late June the price had slid back toward US$4,000, leaving gold down about 7% on the year [source: World Gold Council, 2026]. The whipsaw is the story in miniature: a metal that earns no interest, pays no dividend and costs money to store had rallied hard enough to be crowned, by one central-bank measure, the world's largest reserve asset — and then fell fast enough to remind everyone why "safe" is a slippery word.
This article looks at what actually drove the rally, who was buying, what the macro backdrop was, and why the "safe haven" label is both apt and misleading. It keeps a hard line between verified data — official statistics from the World Gold Council, central banks and the IMF — and the forecasts and price targets that fill market commentary. It is not investment advice, and it is not a recommendation to buy or sell anything.
Table of Contents
- Record highs, then a reversal: reading nominal versus real
- Who was buying: central banks and the ETF surge
- The macro backdrop: real rates, the dollar, and geopolitics
- The de-dollarization debate: how much is really data?
- Why "safe haven" — and where the label breaks down
- What to watch
Record highs, then a reversal: reading nominal versus real
The scale of the move is hard to overstate. Across 2025 the LBMA (PM) gold benchmark set 53 new all-time highs, and the annual average price reached US$3,431 per ounce, up 44% year over year and the highest annual average on record [source: World Gold Council, 2026]. Momentum carried into the new year: gold rose about 60% in 2025 after a roughly 30% gain in 2024 [source: European Central Bank, 2026], before printing that intraday peak near US$5,595 on January 28, 2026 [source: World Gold Council, 2026].
Then came the reversal. By late June 2026 gold traded around US$4,100, roughly 7% lower year to date, after dipping toward US$4,000 [source: World Gold Council, 2026]. At the peak, realized volatility ran above 50% — a figure more associated with speculative tech stocks than with the asset people reach for when they want calm [source: World Gold Council, 2026]. The lesson is not that the rally was fake; it is that gold can climb and crash within a single year.
The nominal-versus-real caveat
A word of caution about the word "record." Those all-time highs are nominal — they compare today's dollar price with yesterday's, without adjusting for inflation. Gold's previous famous peak, around US$850 in January 1980, looks tiny next to US$5,000, but a dollar in 1980 bought far more than a dollar today, so a straight comparison overstates how far gold has really traveled. The honest way to compare across decades is in inflation-adjusted terms. On that basis the 2025–2026 surge was large enough to set genuine new real highs too, not merely nominal ones — but the general rule stands: when a headline says "record," check whether it means record price or record purchasing power. They are not the same thing.
Who was buying: central banks and the ETF surge
Two engines powered the rally, and they ran on different fuel: official-sector buying by central banks, and investment demand through exchange-traded funds and coins.
Central banks
The official sector has been accumulating gold at a pace not seen in half a century. Central banks bought more than 1,000 tonnes in each of 2022, 2023 and 2024, with net purchases of 1,092.4 tonnes in 2024 and a peak near 1,136 tonnes in 2022 [source: World Gold Council, 2026]. In 2025 the pace moderated to 863.3 tonnes, down 21% and the lowest since 2021 — yet that was still the fourth-largest annual total on record and far above the 2010–2021 average of 473 tonnes a year [source: World Gold Council, 2026]. The World Gold Council tied the slowdown to price itself: "elevated valuations" of gold reserves "prompted a more cautious approach," even as long-term strategic interest held firm [source: World Gold Council, 2026].
Crucially, the buying is emerging-market-led. The National Bank of Poland added 102 tonnes in 2025 — the largest single buyer for a second straight year — lifting its holdings to 550 tonnes, followed by Kazakhstan (57 tonnes), Brazil (43 tonnes) and Turkey [source: World Gold Council, 2026]. This is the deepest source of the rally, and the steadiest: since 2022, central banks have absorbed roughly 1,000 tonnes a year on average [source: World Gold Council, 2026].
ETFs, bars and jewellery
Investors provided the second engine, and in 2025 they came back with force. Global gold-backed ETF holdings climbed to an all-time high of about 4,025 tonnes, with annual inflows of 801 tonnes — worth roughly US$89 billion and the second-strongest year on record [source: World Gold Council, 2026]. Bar and coin demand rose 16% to 1,374.1 tonnes [source: World Gold Council, 2026]. Add it up and total gold demand topped 5,000 tonnes for the first time ever, worth an unprecedented US$555 billion, up 45% on the year [source: World Gold Council, 2026].
One part of the market moved the other way, and it is revealing. Jewellery demand fell 18% by weight to 1,542.3 tonnes as high prices deterred buyers — yet the value of that smaller volume still rose 18% to a record US$172 billion [source: World Gold Council, 2026]. The same high price attracts investors and repels jewellery shoppers at once: proof that "gold demand" is not one thing but several, pulling in different directions.
The macro backdrop: real rates, the dollar, and geopolitics
In the textbook, gold competes with interest-bearing assets, so it tends to shine when real (inflation-adjusted) yields fall and the dollar weakens — because a metal that pays no interest looks better when cash and bonds pay little either. In 2025 the textbook broke. Gold rallied alongside a relatively firm dollar, and analysts attributed the decoupling not to currency moves but to central-bank buying and worries about sovereign debt [source: Reuters, 2026].
The World Gold Council's own decomposition of the first half of 2026 tells a similar story: momentum contributed roughly 24% of the performance, risk and uncertainty about 17%, and foreign-exchange effects about 14%, while interest rates added only around 3% [source: World Gold Council, 2026]. Geopolitics did real work — a US–Iran conflict that escalated early in 2026 supported prices — but with a twist. That same conflict pushed oil and inflation expectations up, which led markets to price out Federal Reserve rate cuts and raised the opportunity cost of holding non-yielding gold, contributing to the pullback [source: Reuters, 2026]. The Fed's own December 2025 projections had penciled in only about one 25-basis-point cut for all of 2026 [source: U.S. Federal Reserve, 2025]. The Council flagged real rates and the dollar as "cyclically high," a potential headwind if they stay there [source: World Gold Council, 2026].
The de-dollarization debate: how much is really data?
No gold story is complete without the claim that the world is fleeing the dollar and rushing into bullion. Here, more than anywhere, it pays to separate a striking data point from the interpretation stapled to it.
The striking data point is real. In June 2026 the European Central Bank reported that, measured at market value, gold had overtaken both US Treasuries and the euro to become the largest official reserve asset at the end of 2025 — about 27% of global reserves, ahead of Treasuries at 22% and the euro at 15% [source: European Central Bank, 2026]. But the ECB was careful to add the interpretation that headlines often drop: the shift "mainly reflects valuation effects, rather than a direct replacement of Treasury holdings." In plain terms, a 60% price rally mechanically inflated gold's share; central banks did not dump their Treasuries to buy it [source: European Central Bank, 2026].
The flow data reinforces the caution. The dollar's share of allocated reserves reported to the IMF eased to roughly 57% by the third quarter of 2025, but much of the 2025 decline came from exchange-rate effects rather than active portfolio shifts [source: International Monetary Fund, 2025]. Real diversification is happening — central banks are net buyers of gold, led by emerging markets, and China has added more than 350 tonnes since Russia's invasion of Ukraine [source: European Central Bank, 2026] — but "diversification" is not "abandonment," and the process is slow.
Intentions run ahead of flows, too. In the World Gold Council's 2026 survey of reserve managers, which drew a record 76 responses, 89% expected global central-bank gold holdings to rise over the next year, a record 45% expected their own institution to add, and 84% believed gold would hold a larger share of reserves in five years, up from 76% a year earlier [source: World Gold Council, 2026]. That is sentiment and intention — a useful signal, but not the same as realized purchases, and it should be read as such.
Why "safe haven" — and where the label breaks down
So why is gold called a safe haven at all? Because it carries no counterparty: it is nobody's liability and cannot default the way a bond or a bank deposit can. Historically it has preserved value through inflation, currency debasement and crisis; it tends to hold up when stocks fall hardest; and it is liquid almost everywhere on earth. Those properties are why central banks and investors reached for it in a year of geopolitical stress.
But the label has limits, and 2026 put them on display. First, gold is volatile: a metal that can fall roughly a quarter from its peak inside six months is not a stable store of value quarter to quarter, whatever it does over decades [source: World Gold Council, 2026]. Second, it pays nothing — no coupon, no dividend — so holding it carries an opportunity cost that rises whenever safe interest rates rise. Third, physical gold costs money to store and insure. "Safe" describes gold's behavior over long horizons and extreme events, not its month-to-month price.
This is also where forecasts must be quarantined from facts. After the January peak, major banks raised their targets — J.P. Morgan lifted its 2026 year-end target to US$6,300 an ounce, and LBMA delegates had earlier forecast roughly US$4,980 for 2026 [source: Reuters, 2026]. Those are predictions, not data. Even the World Gold Council, whose interest lies in gold, explicitly calls its own valuation scenarios "hypothetical illustrations" rather than price forecasts [source: World Gold Council, 2026]. A price target is a claim about the future; a purchase reported by a central bank is a fact about the past. Keep them in separate columns.
What to watch
Gold's rally is real, multi-year and grounded in verifiable demand — record central-bank accumulation, the strongest investment year in a generation, and a genuine geopolitical bid. But the same evidence counsels humility: the records are nominal, the metal is volatile and yields nothing, and the de-dollarization narrative is more gradual than the headlines imply.
A few things are worth watching from here. Do central banks return above 1,000 tonnes in 2026, or does high pricing keep them cautious? Do Western ETF investors keep adding, or take profits? Where do real rates and the Fed's path settle, given that rates are the classic lever on a non-yielding asset? Does the geopolitical bid persist, and does its inflation side-effect keep cutting both ways? And in the reserve data from the IMF and ECB, how much of gold's rising share reflects real diversification versus the arithmetic of a higher price? None of that is a reason to buy or sell. It is a reminder that a record price answers fewer questions than it raises.