PredictMyGold
Special report

The worst month for gold is also its best year

We stress-tested $4,500 gold against a Fed hike, $150 oil and a Treasury exodus. In the scenario that hurts most, gold falls 12% in sixty days — then rises 43% in twelve months.

Gold is near $4,520 an ounce. It hit a record close to $5,600 in January, lost more than a fifth of its value in the second quarter — its worst quarter since 2013 — then rallied 14% in August. Now a new Federal Reserve chair is threatening to raise rates into a war-driven oil shock.

Most commentary treats each of these as a reason to buy gold. We ran the numbers, and the picture is more uncomfortable than that. In the near term, the war in Iran is working against gold, not for it. Over a longer horizon, the scenario that hurts most in the first sixty days is the one that pays best over twelve months.

Getting that sequence right matters more than getting the direction right.

The regime changed in May, not February

The Iran war began on 28 February. Gold's worst quarter in thirteen years happened during it. That single fact should retire the idea that geopolitical risk alone drives this market.

What changed is the Federal Reserve. Kevin Warsh took the chair in May and used his first Jackson Hole speech to say that summer inflation readings, while better than expected, did not tell him underlying trends had meaningfully improved. Markets heard a hawk. The 10-year Treasury yield touched 4.81% this week, its highest since October 2023. The 2-year sits at 4.35%, up from roughly 3.5% at the start of a year that opened with markets pricing cuts. After Friday's jobs report, September hike odds reached 63%.

The transmission mechanism is short, and for gold bulls it runs the wrong way.

How the war reaches the gold price

Hormuzescalation Brent above$110 CPI stayssticky Fed hikes,real yields up ETFs sellgold falls What the market did with the last four escalation headlines Ceasefiresignal Brent backtoward $85 Inflationexcuse gone Hike oddscollapse Goldrallies What actually lifted gold on 3 September Peace is the bullish trade. Escalation is not.
Gold rose on 3 September because President Trump said the latest strike would not take long, and traders read the escalation as short-lived. The attack itself did not move it up.

There is a moderating signal underneath the headlines. The US Energy Secretary said 17 million barrels transited the Strait of Hormuz on 1 September, the highest daily volume since the war began, and the Navy confirmed on 25 August that it had cleared mines from the main shipping lane. Flow is normalising even as strikes continue. That caps oil's upside unless Iran hits Gulf infrastructure directly.

What held the floor at $4,000

Gold has absorbed an 85 basis point repricing at the front of the curve and is still near $4,500. That resilience has one source.

Gold's year, in the three moves that are actually documented

3,5004,000 4,5005,0005,500 $5,600 $4,520 ~$4,000 January record End of Q2 After the August rally all-time high worst quarter since 2013 best month this century −20% in one quarter +14% in August
Only the three levels shown are anchored to reported figures. The path between them is not drawn, because month-by-month precision would be invented.

The World Gold Council recorded a quarterly record of 288.9 tonnes of official-sector buying in the second quarter — precisely the quarter in which the price was collapsing and exchange-traded fund investors were selling. Goldman Sachs expects roughly 60 tonnes a month for the rest of 2026. Central banks were the buyer underneath the sellers.

The structural number that matters more arrived at the start of this year.

Central banks now hold more gold than US Treasuries

Gold reserves $4.0 trillion US Treasury holdings $3.9 trillion First crossover on record
Holdings of the world's central banks at the start of 2026. The margin is small; the direction of travel is not.

Reserve managers have crossed over. That crossover is why a 20% drawdown found a floor near $4,000 rather than $3,000.

Repatriation is a signal, not a price driver

The Banque de France moved 129 tonnes out of the New York Fed between July 2025 and January 2026 and now stores all its gold at home. The Reserve Bank of India cut the share of reserves held abroad to 22% by March 2026, from 55% three years earlier. The Bundesbank faces mounting political pressure to bring back the 1,236 tonnes it holds in New York, more than a third of Germany's reserves, though no formal plan exists.

Two corrections to the popular version of this story. Federal Reserve data show foreign official gold in New York fell only 2% between the end of 2024 and April 2026 — this is a drift, not a stampede. And repatriation barely moves the price, because the metal changes vaults rather than owners. What it signals is that custody risk inside the dollar system is now being priced by allies, not only by adversaries. A formal German decision would be the headline that turns the drift into something faster.

Treasury selling is overstated now and decisive later

Foreign holdings of Treasuries hit a record $9.13 trillion in June 2025, even as the dollar had its worst half-year since 1973. The "foreigners are dumping bonds" story is really a composition story: China has cut from over $1.3 trillion to around $700 billion while diversifying into gold and euro assets, as UK and Belgian custodial accounts grew.

The development that matters more is domestic. In August the US Treasury expanded buybacks of long-dated securities, apparently to stop long-end yields rising, at the same moment the Fed is preparing to hike. A Treasury conducting quantitative easing by another name while the central bank tightens is the textbook definition of fiscal dominance. It is the most gold-positive structural event of 2026 — and it pays out after the hawkish phase breaks, not during it.

The stress test

We pushed all of these forces to extremes at once, across two horizons: the first sixty days, when rates and liquidity dominate, and twelve months, when fundamentals reassert.

The sensitivities are published ones. Goldman Sachs puts roughly $120 an ounce of price impact on every 50 basis points of Fed policy, and roughly 1.7% on every 100 tonnes of net committed-buyer demand. Academic estimates put around 6 basis points on the 10-year for every $100 billion of foreign official Treasury selling.

The model's one real insight is arithmetic. At $4,520, a tonne of gold costs about $145 million, so $100 billion of reserve money is roughly 690 tonnes — close to a full year of global central-bank buying. Treasury selling only reaches gold through that rotation, and the rotation is enormous relative to the size of the gold market.

Where gold ends up: sixty days versus twelve months

$4,000$4,500$5,000 $5,500$6,000$6,500 spot today, $4,520 60 days12 months 1979 redux  $6,450 Oil shock, Fed chases  $5,500 Reserve exodus  $5,380 Gulf sells its gold  $4,780 Base case  $4,540 Hawkish squeeze  $4,320 Peace + growth miracle  $3,930 The two boldest lines cross. The worst sixty days belongs to the best year.
Each line runs from the modelled 60-day outcome to the 12-month outcome for one scenario. Gold-coloured lines end higher than spot; red lines end lower.
Seven scenarios, with inputs
ScenarioFedBrentForeign UST selling, share into gold10-yearGold, 60 daysGold, 12 monthsDollar
Base case, oil holds near $97+25bp$97$50bn, 20% (~70t)4.9%$4,470 −1%$4,540 0%−1%
Hawkish squeeze: Hormuz normalises, Fed keeps hiking, ETFs sell 150t+50bp$85none4.9%$4,440 −2%$4,320 −4%+1%
Oil shock and the Fed chases it+75bp$130$150bn, 25% (~260t)5.5%$4,130 −9%$5,500 +22%−3%
Reserve exodus: Japan, China and the Gulf sell together+25bp$100$400bn, 30% (~830t)5.2%$4,600 +2%$5,380 +19%−10%
1979 redux: everything at once+100bp$150$500bn, 30% (~1,030t)6.1%$3,990 −12%$6,450 +43%−11%
Gulf states sell 300 tonnes of gold to fund the war+50bp$120$200bn, 20% (net ≈ 0t)5.4%$4,120 −9%$4,780 +6%−5%
Peace and a growth miracle: CB buying halves, ETFs sell 400t0$70inflows4.5%$4,560 +1%$3,930 −13%+2.5%

What the extremes reveal

The worst sixty days and the best twelve months are the same scenario.

In the 1979 redux, the first move is a forced deleveraging toward $4,000. That is the pattern of the second quarter of this year and of March 2020: funds, trend-followers and margin calls sell gold to raise cash regardless of what it is worth. The second move is a Fed that raises rates a full percentage point and still cannot get real rates positive against $150 oil. That is the regime in which gold reprices past $6,000.

Anyone running size faces a specific operational risk here: being stopped out in the first move and missing the second. The answer is not to hold and hope. It is to hold enough cash to be the buyer at $4,000.

Treasury selling is the only input that helps on both horizons. Every other shock hurts first. A $400 billion reserve exodus with 30% rotated into gold means roughly 830 tonnes of price-insensitive demand arriving in a market that clears on perhaps a thousand tonnes of official buying a year, and the dollar takes a 5 to 10% hit. The caveat is probability: foreign holdings set a record last year despite the narrative. This is a multi-year drift, not a shock.

Rate hikes alone are nearly harmless. Hikes with falling oil are the real bear case. The hawkish squeeze looks mild at −4%, but push it further. A Hormuz settlement takes Brent to $85, which removes the Fed's inflation excuse; the Fed hikes anyway to establish credibility; ETF money leaves; and there is no debasement premium left to offset any of it. Combine that with the growth miracle — central banks halving their purchases because there is nothing left to hedge — and gold is at $3,900 a year from now. State Street has already flagged an artificial-intelligence productivity surge that shrinks deficits through growth as the most structurally negative case for gold. The genuine downside tail comes from good news.

And the scenario nobody is pricing: the buyers become sellers. Central banks have already begun using gold to fund energy and defence spending this year. If Saudi Arabia, the UAE and Kuwait monetise 300 tonnes to pay for war budgets exactly when everyone else is rotating in, the official bid nets to zero and gold ends the year up 6% instead of 20%. Quarterly World Gold Council data showing any Gulf central bank flipping from buyer to seller is the single most important early warning available to a gold investor right now.

The dollar's asymmetry is the cleanest result in the table. The dollar strengthens in the mild scenarios and weakens in every extreme one. There is no extreme scenario in which it finishes the year higher. Its upside is bounded by how far Warsh can hike; its downside is not bounded at all.

What we would do at $4,520

The binary in front of us is next week's inflation print and the FOMC meeting on 15 and 16 September. Hike odds have swung between 31% and 66% in three weeks on single data releases. This is a coin flip on 25 basis points, decided by a chair who needs to establish credibility.

Into that event, the sensible posture is lighter tactical exposure — roughly a third less — with part of the delta replaced by put spreads or collars rather than outright puts. Goldman's research notes that the growing use of gold derivatives to hedge policy risk is making the metal more volatile, which means implied volatility is already expensive and outright options overpay for it.

If the Fed hikes and gold trades into $4,000 to $4,200, that is the re-entry. Goldman's desk frames a $4,000 floor as the trade, as central banks rewrite the demand picture. We agree the floor is real. We disagree that we are at the buying point yet.

The strategic core is a different question and should not be traded. Reserve rotation, custody risk and fiscal dominance are all compounding, and the last hike of this cycle will be the signal for the next leg. Anyone holding for that leg should do what the central banks themselves are doing: physical, allocated, in Zurich or Singapore rather than London or New York. A US-domiciled ETF gives you the price and none of the sovereignty hedge you are paying for.

For Indian investors there is an extra term in the equation. Every oil spike weakens the rupee, so gold in rupee terms carries a tailwind that dollar gold does not.

Our position

Next three months: a $4,200 to $4,800 range, with the Fed meeting and the oil price deciding which end. We would be lighter into the meeting and a buyer of the hike-driven fall.

Twelve to twenty-four months: a retest of $5,600 once the Fed pivots. Goldman's year-end 2026 target is $4,900. We would take the under this year and the over for 2027.

A note on the model

The sensitivities cited above are published. The interaction terms are ours: a deleveraging haircut when the 10-year passes 5.25% or Brent passes $130, a debasement premium above $110 oil, and a credibility-break adjustment when the Fed hikes hard into a stagflationary shock. They are calibrated to what 2026 has already demonstrated rather than fitted to history.

Read the ordering of the scenarios and the sign flips between horizons as the finding. The specific dollar figures are accurate to perhaps ten per cent either way. Anyone who quotes a gold forecast to the nearest dollar is selling something.