The strangest market of 2026 is a war that sold gold. The metal peaked at $5,595 an ounce on January 29 — days before the first shots of the US-Iran war. It spiked on the initial safe-haven reflex, a move Morningstar called "entirely rational, a classic safe-haven reflex to a major geopolitical shock." Then it spent eight months unwinding. The second quarter was its worst since 2013 — from roughly $4,700 into April to about $4,015 at its end. It briefly broke $4,000 in August. And this week, with Hormuz effectively closed, Brent above $100 for a fourth consecutive day, US diesel above $6 a gallon for the first time in history, and central banks buying bullion at a record quarterly pace, gold is on track for its third straight weekly decline, trading near $4,350 after a Friday low of $4,292.
In any textbook, that combination is impossible. In this market, it is arithmetic.
The mechanism: an inflationary war is a hawkish war
The debasement trade was built for one configuration: fiscal dominance. Deficits, debt costs, a central bank that monetizes — the 2020s playbook in which gold and bitcoin rise because the state quietly dilutes its own liabilities. Ray Dalio restated the case in August, weeks after Treasury Secretary Bessent's expanded buyback operations: underweight bonds, hold 10-15% of a portfolio in gold plus "a bit" of bitcoin, because a US debt crisis could arrive within about three years. CNBC ran the framing plainly — "the debasement trade returns."
What 2026 delivered instead is the opposite configuration. The war did not arrive as a monetization event; it arrived as a supply shock. It destroyed physical redundancy — the pipelines and refineries documented in this series — and repriced diesel, the most inflationary commodity in the economy. An inflationary war forces the central bank to fight, not to accommodate. The Fed meets September 15-16 with futures pricing roughly a 90% chance of a hike — while, as Bannockburn's Marc Chandler notes, only 13 of 48 economists surveyed by Bloomberg expect one. Oil-driven inflation expectations did the repricing: PPI ran hot on energy, headline CPI held at 3.4%, and the 10-year TIPS yield reached 2.54% on September 10 — above its late-2023 peak, the highest zone in twelve years, up nearly a full percentage point year-over-year.
Real yields are the price of gold's problem. Erb and Harvey's long-run correlation between the 10-year real yield and gold is -0.82 — the strongest, most reliable negative relationship in commodities. When real yields rise, the opportunity cost of holding a non-yielding asset rises with them. The war, in other words, is bullish for gold's long-term thesis and bearish for its cyclical price, and the cyclical price is what prints on the tape.
The precedent is not hypothetical. After Russia invaded Ukraine in 2022, gold rallied on the initial shock, then fell roughly 20% from March 8 to November 3 as the energy-driven inflation push pushed yields and the dollar higher. ING's analysts made the comparison explicitly this spring: "We saw the same dynamic in 2022 after Russia invaded Ukraine... The same dynamic has played out here, only faster." Fisher Investments tallied the 2022-23 hiking cycle at a 1.7% total return for gold — a safe haven that delivered bond-like nothing through a war and a hike cycle. The pattern is old. The war is new. The Fed's reaction function is the variable.
The two bids: why the same war splits the market
Look under the price, though, and the paradox becomes a split rather than a collapse. Gold has two bids with opposite sensitivities to this war.
The cyclical bid is Western capital: ETFs, managed money, the momentum complex. It trades real yields, and it has been leaving. World Gold Council data show ETF demand collapsed to just 4 tonnes in Q2 — more than 80% below Q1's 20 tonnes and the lowest since Q2 2025, with March's US-listed outflows setting the tone. Managed money's net long on COMEX stood near 145,000 contracts in late August — rebuilt from 128,000 a fortnight earlier, but far below the exuberance of the $5,000s. This is the bid that the Fed's hike path is currently repricing.
The official bid is the counterpart: central banks, which trade sanctions risk and reserve diversification, not TIPS spreads. They bought 289 tonnes in Q2 — a record for any second quarter in the WGC's data series, a fivefold increase on Q1's revised 57 tonnes. Bar and coin demand held steady at 307 tonnes. The Netherlands moved part of its gold reserves out of the US and Canada citing "crisis preparedness." And note the honest caveat in the fine print: the WGC revised Q1's central-bank figure down from an initially reported 244 tonnes to 57 tonnes after reclassifying flows to the OTC market — the official bid is real, but we measure it with error bars wide enough to hide a mid-sized buyer.
January was the moment the two bids aligned: rate-cut expectations, fiscal fear, and central-bank accumulation all pointed the same way, and the metal printed $5,595. The war split them. The same sanctions overreach that drives Moscow's and Beijing's reserves into bullion also drives the oil price that forces the Fed to hike. Gold is being bought by the war's geopolitical victims and sold by its inflationary consequences, simultaneously. That is why the metal can set a record in January, lose a quarter of its value by June, and still have central banks buying a record 289 tonnes into the slide.
What the experts actually said
The Kitco weekly survey published Friday captures the professional split. Wall Street turned majority-bullish into the FOMC; Main Street clung to a slim bullish majority through the third weekly loss.
The hawks on gold are the rate hawks. Adam Button of investingLive: "Lower. The Fed will hike, and that puts a downward bias into gold — but if the statement and press conference isn't sufficiently hawkish, then gold could quickly rebound." That conditional is the whole trade in one sentence: gold's near-term path runs through the tone of a press conference.
The bulls are, strikingly, the ones arguing the Fed cannot finish what it started. Bob Haberkorn of StoneX: "I don't think they can raise rates that much considering the situation that the Fed is in, the debt levels that are out there... I think they might just have to deal with higher inflation and see how it goes... If they don't raise rates at the meeting on the 16th, we will see metals trade substantially higher into the end of the year." Darin Newsom of Barchart, a technician by trade, made the structural argument in two lines: "the rest of the world continues to sell the US. For obvious reasons. Meanwhile, central banks continue to buy gold. Also for obvious reasons." Adrian Day sees "a strong sign of underlying strength" in gold's refusal to break down under hawkish pressure — "slowly forming a higher base, at the $4,300 level, before moving higher again."
The sell side has spent the correction capitulating in both directions. Goldman Sachs cut its year-end target to $4,900 from $5,400 in June — the first cut since 2023. JPMorgan scrapped its ~$6,000 path in July for $4,300 (Q3) and $4,500 (Q4). Deutsche Bank sits at $4,300-4,800, Morgan Stanley near $5,200. UBS broke ranks in late July and raised. Bank of America's Michael Widmer put $5,000 "within reach" — conditional, tellingly, on a rebound in Western ETF inflows, which is precisely the bid that is currently absent. The dispersion of targets is itself information: nobody's model currently prices both halves of the paradox at once.
And there is a camp arguing the framework itself has broken. "The correlation that has governed this market for four decades just quietly broke," one widely-circulated analysis put it — 10-year real yields and gold climbing together through 2025-26. The more careful version, from an empirical study of the TIPS-gold relationship back to 1997, is that the correlation is "neither a stable regularity nor a fiction": real yields govern the cyclical component, central banks the structural floor, and the two channels can diverge for years. The 2026 market is what that divergence looks like in real time.
The honest counterweights
Three things keep this from being a simple bull story about the official bid.
First, floors do not make trends. Central banks also bought steadily through 2011-2015, and gold still fell roughly 45% peak-to-trough as real yields rose out of negative territory. The official bid is slow, price-insensitive and unleveraged — it absorbs supply, it does not reverse momentum. A floor of 289 tonnes a quarter limits the depth of a capitulation; it does not start the next leg up.
Second, the positioning is not washed out. Managed money rebuilt net longs through August. The 2022 analogue bottomed only after the spec complex capitulated and the Fed's terminal rate was visible. Neither condition is in place; the terminal rate is, if anything, still being discovered.
Third, the technical structure is fragile. Analysts tracking the metal describe a head-and-shoulders neckline under the $4,300 base, with a confirmed break targeting $3,950 — roughly 9% lower — and the $3,920-4,000 zone as the line where the multi-year bull thesis itself gets tested. Friday's rebound off $4,292 held the September 1 low close of $4,396 on the December contract, per Newsom's read — constructive, but one hawkish press conference from invalidation.
Scenarios
The paradox resolves at the FOMC on September 15-16, and the scenarios price the Fed's reaction function, not the war:
45% — hike-and-continue. The Fed delivers, signals more (October/December are already discussed), real yields extend. Gold tests the $3,950-4,000 zone; the official bid limits the cascade; the 2022 replay completes with a base forming late in the year.
35% — hike-and-stop, or hold. The economists' minority (13 of 48) is right, or one hike comes with a dovish dot path. The political economy — midterms, $1 trillion of interest expense, a Treasury Secretary whose bond market is misbehaving — caps the cycle. Squeeze through $4,540; the bank targets at $4,900-5,200 come back into play. Haberkorn's "substantially higher into the end of the year" is this path.
20% — escalation shock. Brent pushes $110-120, Hormuz fully closes for weeks. The haven bid and the real-yield channel fire simultaneously — the spike comes first, the fade follows, unless the Fed visibly flinches. This is the scenario where the paradox becomes a whipsaw rather than a trend.
The tells: the statement and Warsh's presser tone; the dot path; whether October/December pricing builds or collapses; weekly ETF flows (the missing Western bid — BofA's condition); the COT washout; and Brent versus the $105-108 escalation band into the Sunday reopen.
The deferral, not the defeat
The debasement trade is not failing. It is being deferred — and the thing deferring it is the very war that is supposed to justify it.
The fiscal arithmetic the trade runs on is getting worse during the correction, not better: interest expense past $1 trillion a year, a foreign bid for Treasuries that has dried from more than half the market to roughly 30%, buyback operations rebuffed by the tape, a counter-order in Delhi building plumbing around the dollar system, and central banks accumulating gold at a record pace precisely because the sanctions regime keeps demonstrating what reserve diversification is for. Every one of those is a debasement-trade pillar. None of them has gone away. What went away is the easing cycle that was supposed to monetize them — because a war that inflates forces the Fed's hand first.
Gold's paradox, stated cleanly: the war strengthens its structural case and breaks its cyclical transmission. The metal answers to TIPS, not to tanks, until the hiking cycle ends — and the debt arithmetic Haberkorn cites argues the ceiling on that cycle is closer than 90% futures pricing admits. When it ends, the deferred trade reactivates with the official bid already in place and the fiscal case stronger than it was at $5,595.
Until then, the tape says what it says. A war can be the best long-term argument for gold and still take $1,200 off its price on the way to proving it.
Buffett · September 12, 2026