The Number That Should Not Exist
Gold closed Monday at $4,613 an ounce. That is a 180% gain from the October 2022 intraday low of $1,656.43, posted during a week when the two-year Treasury was screaming higher and Jay Powell was promising pain. Over the same 42 months, headline CPI fell from 9.1% in June 2022 to 2.4% in February 2026 before the March print reaccelerated to 3.3%. The disinflation was the sharpest in 40 years. The Fed delivered 525 basis points of tightening into it. Real yields on the 10-year TIPS sat above 2% for most of 2023 and 2024 and currently print at 1.93%. Every variable on the standard inputs sheet pointed the wrong way for bullion, and bullion tripled anyway.
This is not how the asset class is supposed to behave. The framework taught at every CFA prep school in the country, the one Larry Summers and Robert Barsky codified in their 1988 paper, the one that drove a generation of macro books, said gold is a long-duration zero-coupon claim on the inverse of real interest rates. When real yields rise, the discount on gold's perpetual non-payment widens, and the price falls. That model was the consensus until roughly the second week of March 2022. It explained 1980. It explained the 2011 peak. It explained the 2013 crash when Bernanke whispered the word taper. It has explained almost nothing since.
Walk through the wreckage. The SPDR Gold Trust, ticker GLD, held 1,064 tons in March 2022. By the end of 2024 it held 871 tons. ETF demand, the marginal Western retail and institutional buyer that the old model assumed sat at the center of the market, was a net seller for most of the rally. iShares Gold Trust, IAU, told a similar story. The miners, GDX and GDXJ, lagged spot by a factor of two for stretches in 2023, the classic tell of a market where the believers in the asset are not the believers in the equity. Speculative net positioning on COMEX peaked early and faded. Roughly every Western fingerprint a 2010-vintage gold analyst would have looked for went the wrong way. And the price kept going up.
Something else was bidding.
What the Real-Yield Model Got Wrong
The Barsky-Summers framework was not stupid. It was a model of a market where the buyers it modeled were the price-setters. From roughly 1980 through 2008, that was approximately true. The London fix moved with TIPS yields, with the trade-weighted dollar, and with U.S. real growth expectations because the marginal contract was being struck by a London-based bank, hedged through a New York ETF authorized participant, and warehoused in a vault that answered to LBMA rules. The architecture of the market was Western. Therefore the price function of the market was Western.
That architecture started cracking quietly during the GFC, when emerging market reserve managers began noticing the speed at which dollar funding froze for anyone outside the Fed's swap-line circle. It cracked louder in 2013, when the PBOC stopped publishing reserve composition data with the regularity it once did. It cracked again during the 2018 Turkey crisis. By the time of the Russian invasion of Ukraine on February 24, 2022, the cracks were structural. The G7 then took a sledgehammer to the wall.
On March 2, 2022, the United States, the European Union, the United Kingdom, Canada, and Japan froze access to approximately $300 billion of Russian Federation foreign exchange reserves held in their banking systems. The Bank of Russia, an investment-grade sovereign by every reasonable definition twelve days earlier, was suddenly unable to touch more than half its rainy-day fund. The reserves were not seized. They were rendered illiquid by political decision. For every other reserve manager on earth, from the State Administration of Foreign Exchange in Beijing to the Reserve Bank of India in Mumbai to the Saudi Central Bank in Riyadh to the National Bank of Kazakhstan, the message was operational, not rhetorical. Dollar reserves can be turned off. Euro reserves can be turned off. The only sovereign-grade liquid asset on the planet that cannot be turned off by another sovereign is the bar of metal sitting in your own vault under your own flag.
The real-yield model has no input for that variable. There is no column on the CFA Level III spreadsheet labeled "counterparty risk on the dollar itself." When that variable went from zero to nonzero on a single Wednesday in March 2022, the demand curve for gold shifted up by an amount the old model could not see and could not price. The price went where the new demand curve told it to go. Real yields, ETF flows, and CPI prints became second-order noise around a first-order rewiring.
The World Gold Council numbers tell the story with brutal clarity once you know where to look.
The Reserve Bid: 1,082, 1,037, 1,092, 863
Central banks bought 1,082 tons of gold in 2022. They bought 1,037 tons in 2023. They bought 1,092 tons in 2024. They bought 863 tons in 2025. The four-year total is 4,074 tons. To put that in context, the prior decade, 2012 through 2021, averaged roughly 500 tons of net official-sector purchases per year, and that decade itself was already the largest sustained official buying since the collapse of Bretton Woods. The 2022-2025 run doubled the prior decade's pace and did so during a period when, on paper, every reason to own gold was being unwound by the Fed in real time.
The buyers are not anonymous. The PBOC has been a steady tape-printer, adding ounces almost every month since November 2022 with only the occasional cosmetic pause when the dollar got too disorderly. Turkey's central bank, the CBRT, has been the most volatile name on the list, oscillating between net buyer and net seller depending on the lira crisis du jour, but a structural buyer across full cycles. Poland's NBP under Adam Glapinski has been the loudest in print and the steadiest in execution; Warsaw added more than 130 tons in 2024 alone, with the stated objective of pushing gold above 20% of total reserves. India's RBI has accelerated under a multi-year mandate to repatriate metal from London to domestic vaults. Kazakhstan and Uzbekistan, both gold-producing economies, have been net buyers at the margin. Singapore's MAS surprised the tape in early 2023 with a 45-ton print that nobody on the desk had modeled. Even Russia's CBR, despite the sanctions and the war, has continued to accumulate domestically mined ounces through the National Wealth Fund.
What makes these flows different from the old model's assumptions is duration. ETF holders mark to market and sell on drawdowns. Central banks do not. The State Administration of Foreign Exchange has no quarterly redemption window. The RBI does not get fired for missing a benchmark. When the price of gold ran from $2,000 to $2,400 in the spring of 2024, every Western technical desk in London called a top, and the central bank bid kept coming. When it ran from $3,000 to $3,600 in late 2025, the same thing happened, and the same thing happened. The central bank flow function is not price-elastic in the way ETF flow is. It is mandate-driven, and the mandate, once set by a board in Beijing or Warsaw or New Delhi, runs on a fiscal-year cadence that does not care about Bloomberg's morning note.
This is the bid that explains the price. Not real yields. Not CPI. The bid.
Russia, the Treasury Freeze, and a New Sovereign Hedge
Replay the tape from Q1 2022. February 24, Russian forces cross the border. February 26 through 28, the SWIFT carve-outs are announced. March 2, the formal G7 reserve freeze is implemented and the relevant euro and dollar accounts are made inaccessible. By the end of that first week, the operating assumption inside every non-aligned reserve management department on earth had quietly changed. Daleep Singh, then the deputy national security adviser, gave a press conference in which he described the freeze as the financial equivalent of disabling another country's air force on the ground. The metaphor was internalized in Beijing and Riyadh within days.
The response did not show up in speeches. It showed up in monthly IMF COFER data, in PBOC reserve postings, and in shipments through Heathrow, Zurich, and Hong Kong. From Q2 2022 onward, the share of dollar-denominated assets in global FX reserves resumed a slow decline that had paused during the 2018-2021 dollar rally. The share of gold in total reserves, by contrast, began climbing at a rate not seen since the early 1970s. By the end of 2024, gold accounted for roughly 18% of total global reserves at market value, up from roughly 11% at the start of 2022, with the bulk of the increase concentrated in non-G7 holders. Poland, an EU and NATO member, was an outlier in the wrong direction for the consensus narrative; Glapinski's mandate predated the war and accelerated through it, suggesting the diversification logic transcends the simple East-versus-West framing.
The instrument matters. A Saudi reserve manager who wanted insurance against future sanctions risk could not buy bitcoin in size; the market is too thin and the venue risk too acute. They could not buy yuan in size either, since the CNY market itself is intermediated by a banking system that the same G7 could theoretically sanction. They could buy gold. They could store it domestically. They could custody it without a New York correspondent. They could settle bilateral trade against it. None of those features is a function of CPI or DFII10. All of them are functions of the freeze.
This is the cause that the textbook chapter does not have a page for. Sovereign hedging demand, denominated in tons rather than in basis points, with a planning horizon measured in decades rather than quarters, became the price-setting flow. The Western ETF complex became the price-taking flow. The relationship between the LBMA AM fix and the 10-year TIPS yield, which carried an r-squared above 0.7 from 2007 through early 2022, has carried an r-squared near zero since. Pull the regression on your terminal. The line has been broken for four years.
What April 2026 Hot CPI Will and Won't Prove
Which brings us to the print on the screen this morning. March CPI came in at 3.3% year over year, hotter than the 3.1% consensus and the second consecutive monthly reacceleration after the February low of 2.4%. The regime tag on the macro dashboard reads stagflation-stable. Fed funds sit at 3.50 to 3.75%. The 10-year Treasury yields 4.31%. The VIX is calm at 17.99, suggesting cross-asset desks have not yet decided whether March was noise or signal. Gold ticked higher into the print and held the gain.
Here is the cleanest test the market has given us in four years. If the old Barsky-Summers framework still has any explanatory power, a hot CPI alongside a 1.93% real yield should pull gold lower over the next two to four weeks as the Fed re-prices toward a longer hold and breakevens widen less than nominals. If the new framework, sovereign reserve flow as the marginal buyer, is doing all the work, the central bank bid will absorb whatever Western ETF outflow the hot print produces, and the price will drift sideways or higher into the April CPI release. The shape of that response is the cleanest signal we will get on which regime the tape is in.
Watch three things specifically. First, GLD tonnage. If the trust prints meaningful redemptions over the next ten sessions and spot holds, it confirms the bid is non-Western. Second, the spread between the LBMA AM fix and the Shanghai Gold Exchange premium. A widening Shanghai premium during a hot U.S. CPI print would be a tell that the PBOC and Chinese commercial demand are absorbing the Western flow. Third, the Polish and Indian central bank monthly reserve postings due in early June. If both print net additions despite price levels above $4,500, the mandate-driven thesis gets another confirmation.
What the hot print will not prove, no matter the reaction, is that the old model is back. A single tape can confirm that the new regime is dominant. It cannot resurrect a framework whose architectural assumption, the Western marginal buyer, has been falsified for 42 months. The most that bears who still trade real yields against bullion can hope for is a tactical pullback inside a structural bull. The most that bulls can hope for is a clean break above $4,700 that takes out the late-April high and forces the last of the systematic CTA shorts to cover.
Somewhere between those two outcomes sits the actual question for any portfolio that holds gold today. Not whether the price is right at $4,613. Whether the reason for the price is durable. The case for durability is a list of buyers (Beijing, New Delhi, Warsaw, Ankara, Astana, Riyadh), none of whom has indicated any intention to slow accumulation, several of whom have publicly committed to increase it, and all of whom are operating against a sanctions precedent that has not been unwound and shows no sign of being unwound. The case against durability is a single Fed pivot to genuine restrictive territory that drags real yields to 3% and tests whether the reserve bid can absorb a Western liquidation in size.
March's print is not that test. It is the dress rehearsal. The real test arrives when either the Fed has to chase a 4-handle CPI back into the system, or when one of the reserve buyers (and the betting on the desk has been on Turkey or India) runs into a domestic fiscal crisis that forces a sale. Until one of those happens, the regime on the screen is the regime in the data, and the regime in the data has been the same for four years. Tomorrow morning's April CPI revision and the GLD tape at the close will tell us whether month forty-three looks like the previous forty-two.
Explore these indicators together: Chart Gold (Spot), Gold ETF (GLD), and 3 more on the Indicators Dashboard