Gold Fell 27%. Central Banks Kept Buying.

You do not have to guess which AI model wins to invest in the AI boom.

Every large language model needs a massive amount of specialized technology to keep running, scaling, and improving.

That means advanced memory, data storage, high-speed networking, semiconductor manufacturing, power systems, and cooling infrastructure.

MarketBeat’s new The Infrastructure’s Backbone: 10 Stocks Powering the AI Buildout report reveals 10 publicly traded companies positioned across the AI infrastructure stack.

These are the companies helping supply the technology AI providers need to keep growing.

The report normally sells for $29.97, but it is available free today.

See The 10 Stocks

Central banks are shifting away from U.S. Treasuries toward gold as their primary reserve asset. Sanctions risk after 2022 drove the move, and allied governments pulling gold out of U.S. custody have accelerated it. The 2026 price correction tested that thesis directly, and the sovereign bid held through it.

A structural trend that ran for three years in the background became a market event in 2026, first on the way up, then on the way down. Gold set a record above $5,600 an ounce in late January, then fell 27% into July. The correction came from the dollar and the Fed, and the reserve story kept moving the other way underneath it.

The first driver was France's completed repatriation. In January 2026 the Banque de France finished withdrawing all 129 tonnes of gold it held at the New York Federal Reserve, a process that began in July 2025. The stated reason was a 2024 internal audit: the New York bars, some dating to the late 1920s, no longer met the modern purity and weight standards used in gold trading. The operation netted roughly 13 billion euros (about $15 billion) in capital gains, because France sold the old bars at high prices in New York and bought new compliant bars in Europe as prices eased, so no bullion crossed the Atlantic. Every ounce of France's 2,437 tonnes now sits in domestic vaults south of Paris. Germany and Italy, which together keep 2,297 tonnes at the New York Fed, face building domestic pressure to do the same.

The second driver was the Federal Reserve. New Chair Kevin Warsh signaled a more hawkish path than markets expected, and in the second quarter that repriced everything gold-sensitive. The dollar rallied, real yields (interest rates after inflation) rose, and leveraged and ETF holders liquidated, driving gold's second leg down. The DXY dollar index, which measures the dollar against a basket of major currencies, sat near 100.6 in mid-July. Through all of it, official-sector buying continued at 244 tonnes in the first quarter, the fastest quarterly pace in over a year.

Exhibit 1: The gold-Treasury crossover is now official. For the first time since the mid-1990s, gold makes up a larger share of central bank reserves than U.S. Treasuries. The European Central Bank's June 2026 report put bullion at 27% of global reserve assets at the end of 2025, up from 20% a year earlier, with Treasuries down to 22% from 25%. The path here is striking: in late 2015, Treasuries were 33% of reserves and gold just 9%. Over the following decade gold's share tripled while the Treasury share fell by a third. ECB President Christine Lagarde tied it to one force: "Geopolitical tensions continue to drive strong central bank demand for gold."

Figure 1: Gold overtook Treasuries in central-bank reserves in 2025. Source: European Central Bank (June 2026); World Gold Council; IMF COFER.

Exhibit 2: The buying accelerated into the correction. The steady pace of official purchases is the engine behind the crossover, and it sped up as prices fell. Central banks bought a net 244 tonnes in the first quarter of 2026, up from 208 tonnes in the fourth quarter of 2025 and the fastest quarterly pace in over a year, led by Poland, Uzbekistan, and China (World Gold Council). Annual net purchases since 2022 tell the same story.

The 2010 to 2021 average was 473 tonnes per year, so even 2025's softer 863-tonne total ran 82% above that baseline. Turkey, Russia, and Azerbaijan sold a combined 115 tonnes in the quarter for idiosyncratic reasons (currency support, budget funding, and portfolio rebalancing), and net buying still hit its fastest pace in over a year.

Figure 2: Central banks bought the correction. Source: World Gold Council, Gold Demand Trends Q1 2026.

Exhibit 3: The allied repatriation chain. France's move is the headline, but the pattern runs across several allied nations. Repatriation means moving gold from foreign vaults back to domestic ones.

The direction of travel matters more than the tonnage alone. France and the Netherlands have completed full repatriation. Germany's Bundesbank has officially resisted while facing domestic pressure that feels different in kind from earlier debates. Poland is both repatriating and buying hard, adding 102 tonnes in 2025 to reach 550 tonnes, the largest single-country buyer for the second year running.

Exhibit 4: Country-level behavior. The most granular evidence comes from individual central banks. China's central bank has now bought gold for 20 straight months as of June 2026, lifting official holdings to 2,346 tonnes, and it cut its Treasury holdings from roughly $1.06 trillion in 2022 toward $760 billion over four years. Brazil sold $61 billion in Treasuries in 2025 while doubling its gold, lifting gold to about 7.2% of reserves from 3.6% a year earlier. India holds roughly $232 billion in Treasuries and has kept up a steady gold accumulation program. The diversification is gradual in each case and points in the same direction.

Exhibit 5: The dollar's eroding reserve share. A separate measure, the currency mix of allocated reserves (which counts cash and bonds but excludes gold), tells a slower version of the same story. The IMF's survey shows the dollar at 57.79% of allocated foreign exchange reserves in the first quarter of 2025, down from 64.69% in 2017 and 71.5% in 2001, and it has since slipped below 57% for the first time since the mid-1990s. Most of the recent drop came from exchange rate effects, the dollar weakening against other currencies, so central banks are largely holding their dollars steady while adding gold on top.

The path from sanctions risk to a flipped reserve mix runs through five reinforcing stages.

Stage 1: Sanctions create confiscation risk, and gold becomes the only unfreezable reserve. The 2022 Western sanctions on Russia froze about $300 billion in central bank reserves held in dollar- and euro-denominated assets, a demonstrated capability. Central banks with similar exposures faced a clear choice: accept the risk that their reserves could be frozen, or move into an asset no government can freeze. Gold is the only candidate with global liquidity, no issuing government, and no counterparty. Buying roughly doubled from the 473-tonne pre-2022 pace to over 1,000 tonnes in each of the next three years.

Stage 2: Sustained sovereign buying flips the reserve mix. Central banks buy gold for strategic protection, which makes their purchases largely insensitive to price. Sovereign demand absorbs roughly 20% of annual global mine production at a steady pace, and over three years that was enough to flip the mix. Gold's share of central bank reserves rose from about 15% in 2022 to 27% at the end of 2025, while the Treasury share fell from about 27% to 22%.

Stage 3: Diversification extends from the asset to the vault. The sanctions precedent showed that assets held in foreign custody can be seized or frozen. The logical next step is custodial diversification: moving gold from foreign vaults (mainly the New York Fed and the Bank of England) to domestic ones. France completed that step in January 2026, and each completed repatriation raises the pressure on the next country, because the France precedent removes the argument that repatriation is impractical.

Stage 4: Market shocks stress-test the system, and the bid holds. Two shocks hit gold in 2026: the Iran war in February and the Fed's hawkish turn under Warsh in the second quarter. Both pushed the dollar up and gold down, the second all the way to a 27% drawdown from the January peak. Both were dollar-strength and forced-selling episodes among leveraged and ETF holders. Central banks kept buying through them, adding 244 tonnes in the first quarter. The structural bid held.

Stage 5: Crisis infrastructure creates lasting non-dollar settlement capacity. During the Iran war, Iran ran a yuan-based toll corridor at the Strait of Hormuz through CIPS (China's cross-border payment network), a live settlement system operating outside the dollar. Western firms paying in yuan under crisis conditions normalized non-dollar settlement for global trade. That plumbing persists after the crisis ends, permanently raising the capacity for non-dollar transactions even after volumes fall back.

The closest parallel is France's gold conversion campaign under Charles de Gaulle from 1965 to 1971. In the late 1960s France aggressively converted its dollar reserves into physical gold, demanding delivery from the U.S. Treasury. De Gaulle's government called the dollar's reserve status an "exorbitant privilege" and argued that gold was the only honest reserve asset. France's actions, alongside similar moves by other nations, accelerated a run on U.S. gold that ultimately forced President Nixon to close the gold window on August 15, 1971, ending dollar-gold convertibility and the Bretton Woods system.

Three differences matter. The link today is behavioral. In 1971 a fixed gold-dollar peg could be broken in a single policy decision; today the shift comes through thousands of individual purchasing decisions, with no single "Nixon moment" to reverse it, which makes the process slower and more durable. The debt backdrop is far worse: U.S. public debt-to-GDP sat near 35% in the de Gaulle era and is near 101% today. And the stated rationale is technical this time: de Gaulle attacked the dollar's privilege directly, while France framed its 2026 move as an audit-driven upgrade of old bars, though the reserve outcome, all French gold home, is the same.

The commentary in this section describes how asset classes have historically behaved in similar macro environments, for educational purposes only. Nothing here is a recommendation to buy, sell, or hold any security.

Equities. During past periods of reserve-currency transition and rising term premiums (the extra yield investors demand for holding longer-term bonds), U.S. large-cap stocks have often seen their valuation multiples compress. The current setup rhymes with that pattern: the S&P 500 trades near 25 times earnings with a slightly negative equity risk premium (the extra return stocks offer over safe government bonds), and the Buffett Indicator (total market value against GDP) sits near 197%. Energy and materials sectors have historically outperformed in these environments, while rate-sensitive sectors such as technology have faced headwinds.

Rates and Fixed Income. The marginal buyer of Treasuries is shifting from price-insensitive central banks toward price-sensitive private investors who demand yield to compensate for risk. That handoff has historically produced permanently higher term premiums. The 10-year term premium near +0.7% is elevated by recent standards but still below the +1.00% level that would signal a full repricing. In similar environments, institutional desks have historically favored shorter-duration bonds (those less sensitive to interest rate moves), and inflation-protected Treasuries have tended to benefit from the mix of sticky inflation and supply stress.

Credit. Higher structural Treasury yields raise borrowing costs for heavily indebted companies and squeeze their ability to cover interest. High-yield spreads (the extra yield risky bonds pay over safe ones) near 290 basis points, about 2.9 percentage points, look complacent against this backdrop. In prior periods of rising term premiums, investment-grade credit has historically outperformed high-yield by a wide margin, as refinancing risk concentrates among lower-rated borrowers.

Foreign Exchange and Emerging Markets. The dollar's 2026 behavior shows an important nuance. Despite the structural decline in its reserve share, the dollar rallied to multi-month highs on the Iran war and again on the Warsh-driven hawkish turn, with the DXY near 100.6 in mid-July. Short-term crisis and rate-differential demand can override the structural trend for months at a time. Over the medium term, each bilateral settlement deal and each CIPS transaction trims marginal dollar demand.

Commodities. Gold is the central story. Structural demand from over 40 central banks creates a price floor that operates independently of speculative positioning. When leveraged traders and ETF holders sold the 2026 correction, sovereign buyers absorbed the supply. With official demand already taking roughly 20% of annual mine production off the market, that floor has historically created asymmetric upside over time, even after a drawdown as sharp as this year's.

Counter-Argument 1: Gold buying decelerates at elevated prices

Even after a 27% drop, gold above $4,000 an ounce is expensive by any historical standard, and central banks may face budget or political pressure to slow purchases. The 2025 total of 863 tonnes already marked a step down from three straight years above 1,000 tonnes. The thesis still holds because sovereign buyers have shown low price sensitivity: they bought through both the 2025 rally to record highs and the sharp 2026 correction, and the first quarter of 2026 accelerated to 244 tonnes. A large majority of surveyed central banks, 89% in the World Gold Council's 2026 survey, plan to raise gold holdings further.

Estimated probability counter-argument is correct: 15%

Counter-Argument 2: Persistent dollar strength compresses gold further

This risk has partly played out. The Iran war and then the Warsh-driven hawkish turn strengthened the dollar and drove gold down 27% from its January peak. If the Fed stays hawkish or rate differentials widen further, sustained dollar strength could keep pressure on the price even as central banks accumulate. The thesis separates price from structure: the drawdown is a dollar-and-rates event, while central bank buying predates it and is driven by sanctions risk. A durable break lower would require both sustained dollar strength and a resolution of the sanctions risk that triggered the buying, and only the first is currently in play. This is the live risk to the price, with limited power over the reserve shift itself.

Estimated probability counter-argument is correct: 30%

Counter-Argument 3: U.S. fiscal reform or Fed intervention restores Treasury demand

Meaningful fiscal consolidation, or a credible Fed that anchors long-term yields, could restore confidence in dollar-denominated reserves. A demand-driven recession that collapses inflation could also make Treasuries attractive again at lower yields. The thesis still holds because the base rate for major fiscal legislation in a divided Congress is historically near zero, and the deficit is projected to grow from $1.9 trillion toward $3.1 trillion by 2036. Even a hawkish Warsh Fed defends the dollar's yield, and the specific risk driving repatriation is custody safety, which higher yields do nothing to address.

Estimated probability counter-argument is correct: 15%

If official buying holds at or above the roughly 900-tonne annualized pace set in the first quarter and Germany announces even a partial repatriation, the thesis tilts toward acceleration. If the DXY holds above 103 while gold buying falls below 500 tonnes annualized for two straight quarters, it is time to reassess.

Methodology note: Reserve-share figures use the European Central Bank's June 2026 report, which measures gold, Treasuries, and dollar assets as shares of total reserve assets (gold included). That basis differs from IMF COFER, which measures only the currency mix of allocated reserves and excludes gold, which is why the dollar reads 42% on the ECB basis and near 57% on the COFER basis. Central bank purchase figures use World Gold Council net demand data, which can differ from gross totals reported by individual central banks. Price and dollar-index levels are as of mid-July 2026 and move with each new release.

This report is for informational and educational purposes only. It does not constitute investment advice, a recommendation, or a solicitation to buy or sell any security. All asset class commentary reflects historical patterns and educational analysis, not personal investment advice. Past performance does not guarantee future results. Readers should consult a qualified financial advisor before making investment decisions.

Benjamin Capital Research | July 22, 2026