The slow repricing
On 19 August 2026 we wrote that gold had risen 4.92% in a day to $4,580 an ounce, silver 4.47%, and the dollar index had fallen 0.86% to 98.80 — a three-month low, and down 2.3% over twenty trading days (19 August). The same afternoon the 30-year TIPS yield reached 3.09%, the highest since 2008 (19 August). That last number is the one to hold on to. It is the real yield, stripped of inflation compensation, and it says investors want more to lend the US government money for thirty years regardless of what happens to prices.
A currency falling while real long yields rise is not a growth story or a rate-differential story. It is a story about who wants to own the asset.
Our own beat notes through the week recorded the supporting evidence: 74% of reserve managers surveyed expect the dollar's share of reserves to fall, 89% expect central-bank gold holdings to rise, and reserve managers surveyed as net dollar sellers for the first time. None of that is panic. All of it is direction.
The pressure is not only on the sell side. Foreign-currency deposits at Japanese banks grew by ¥3.98tn ($25.1bn) year on year in the April-June 2026 quarter, the largest increase since Japan deregulated such accounts in 1998, with the total averaging ¥26.1tn ($165bn), up 18% (20 August). Japanese households buying dollars with the yen at a 39-year low past 163 is not a valuation trade. And to relieve the pressure on Japan — the largest foreign holder of Treasuries — bought yen on 31 July, the first US purchase in three decades according to Bloomberg. The US Treasury is now spending its reserves on a foreign currency to protect the bid for its own bonds.
The Treasury selloff and the gold bid
The June 2026 TIC data show the ranking: Japan at about $1.116tn, the United Kingdom at $939.9bn, mainland China at $633.4bn, then Belgium ($482.5bn), Canada ($459.6bn) and the Cayman Islands ($453.1bn).
The levels matter less than the distances travelled. China's holdings peaked at roughly $1.32tn in November 2013; at $633.4bn in June 2026 they are about $690bn lower, roughly halved. Japan's peaked at $1.325tn in November 2021 and are about $210bn below that, a fall of around 16%, with June alone down 2.3% from May's $1.143tn. The two largest official holders of American debt have been going the same way for years, one fast and one slow.
Where the money went is not a mystery. Central banks bought 1,080.0 tonnes of gold in 2022, 1,050.8 tonnes in 2023 and 1,089.4 tonnes in 2024, according to World Gold Council data, with 2025 reported at about 850 tonnes. In Q2 2026 the largest buyer was Poland at 51 tonnes, followed by China at 33 tonnes, Uzbekistan at 16, Kazakhstan at 15, and Jordan and the Czech Republic at 6 each. By Q1 2026 the WGC put gold at 29% of total global allocated reserves. It has not, on the evidence available to us, overtaken the euro for second place; that claim circulates and we cannot support it.
The headline number is the IMF's COFER series. The dollar's share of allocated global foreign exchange reserves was 57.13% in Q1 2026, up from 56.42% in Q4 2025 — a reminder that this line does not fall every quarter. The decade anchors are what make it a trend rather than noise: about 71% at the 1999-2000 peak, about 65% in 2014, about 58% across 2022-2024, and 57.13% now. Fourteen percentage points in a quarter of a century is not a collapse. It is a drift, and drifts are how this has always happened.
One thing we could not establish: what share of marketable Treasury debt foreign official institutions hold today versus 2010. The TIC country tables do not split official from private holdings, and the flow data in the June release does not answer the stock question. Anyone quoting a precise figure is estimating.
The $1 trillion gold valuation gap
The US government holds 261,498,926 fine troy ounces of gold and carries it on the books at $42.2222 an ounce, a statutory price set in 1973 and never changed. That produces a book value of about $11.04 billion.
At market prices the same metal is worth more than a trillion dollars. A July 2026 analysis using a gold price above $4,700 an ounce put it at roughly $1.2tn; an August 2026 piece using around $4,050 put it above $1tn and about 96 times the statutory valuation. The gap moves with the gold price, so treat any single figure as a snapshot. The order of magnitude is not in dispute: the asset is carried at about one percent of what it would fetch.
A gap that large invites schemes. The obvious one is to restate the gold at market, book the difference, and spend it. What the record actually shows is narrower than the chatter. A technical review published in July 2026 states that a senior official explicitly said there is no active plan for gold revaluation or a return to a gold peg, and that no legislative or Treasury announcement in 2025-2026 changed the statutory price. An August 2026 Chinese-language report says unnamed members of Congress have put forward bills to study raising funds through revaluation, but names no member and cites no bill number; we are treating that as rumour until someone produces the text. Separately, Kitco reported on 3 August 2026 an economic historian's point that a revaluation would benefit the federal government rather than private gold holders — a point to keep in hand whenever the idea is marketed to retail investors as a reason to buy bullion.
The Triffin dilemma and historical precedent
The structural problem has a name and a diagnosis dating to the late 1950s and 1960s. The Triffin dilemma says that a country whose currency the world uses as reserves has to supply the world with that currency, and the only way to do that in net terms is to run persistent external deficits — sending out more claims than it takes in. Those deficits are what make the system work. They are also what eventually makes people wonder whether the claims are good. The privilege and the problem are the same mechanism seen from two angles, which is why "exorbitant privilege", the French complaint of the 1960s about the dollar's cheap borrowing and settlement advantage, and "Triffin dilemma" are usually discussed together.
It did not have to be designed this way. At Bretton Woods in 1944, Keynes proposed an International Clearing Union issuing a supranational unit called bancor, with adjustment pressure on surplus and deficit countries alike, so that no single nation's balance of payments had to be the world's liquidity tap. Harry Dexter White's plan won: currencies pegged to the dollar, the dollar pegged to gold, and reserves accumulated in US liabilities. Triffin's point was that this combination was self-limiting, because the dollar claims required to lubricate world trade would eventually exceed the gold behind them. On 15 August 1971 Nixon closed the gold window and proved him right.
What happened next is the closest historical analogue to a slow repricing. Gold went from $35 an ounce under Bretton Woods to about $227 in January 1979, roughly $400 by October 1979, and an intraday $850 on 21 January 1980. The Deutsche Mark and yen each appreciated something like 80-90% against the dollar between 1971 and 1980. In the 1978 crisis the United States did something worth naming plainly: it borrowed in someone else's currency. The Carter administration's 1 November 1978 support package included DM-denominated Treasury issues — "Carter bonds" — of about 3.04bn DM in 1978, 2.5bn DM in February 1979 and a further 4bn DM announced on 26 October 1979, taking the total to nearly 10bn DM. A country that can print the world's reserve asset does not issue debt in unless foreigners have stopped taking its own paper on trust.
That is also why the petrodollar myth deserves puncturing. The popular version has a 1974 treaty binding Saudi Arabia to sell oil only in dollars in exchange for security guarantees. What the record supports is a US-Saudi Joint Commission on Economic Cooperation established on 8 June 1974, oil already being priced in dollars beforehand, and a confidential operating protocol agreed in Jeddah in December 1974 between a US Treasury delegation and the Saudi Arabian Monetary Agency for recycling surpluses into Treasuries, some bought outside the normal auction process so the amounts would not show in the public record. Recent reviews are explicit that the official record does not establish the simplified oil-for-security bargain. NPR walked through this in May 2026. The arrangement was real; the treaty was not.
The longer precedent is sterling, and it is the reason we describe this as a repricing rather than an event. Britain went off gold in September 1931. The 1945 Anglo-American Loan forced convertibility, which opened on 15 July 1947 and drained about $700m — a third of the entire loan — from the Bank of England in 36 days before being suspended on 20 August 1947. Suez followed: UK reserves fell from about $2.07bn in January 1956 to about $1.37bn that November, at one point burning $279m a month, with more than $100m gone in the first week of November alone defending the $2.80 peg. The final devaluation came on 18 November 1967, $2.80 to $2.40, a cut of 14.3%. From the first structural break to the last, thirty-six years. Nobody living through 1931 thought they were watching the handover.
Why it matters to this crash
On 19 August 2026 the Treasury said it was increasing "by at least double" its buybacks of bonds maturing in 10 to 20 and 20 to 30 years, taking repurchases from $2bn to at least $4bn and roughly $32bn a quarter from 9 September. The 30-year yield fell about 9 basis points to 5.2%, the 10-year 7bp to 4.64%, and the dollar had its worst day in three months (19 August).
That is the whole thesis in one trading session. A government with a large deficit and a has three places to put the adjustment: yields, the currency, or the real economy. Bessent has said repeatedly that 10-year yields are his benchmark of success. If the issuer steps in to hold yields down, the adjustment does not disappear; it moves to the price of the money. Holders who wanted more compensation for fiscal and inflation risk, and are instead offered less because the borrower is bidding for its own paper, respond by holding fewer dollar assets. Gold is the cleanest expression of that preference, because it has no yield to lose and no issuer to disappoint.
The fiscal arithmetic underneath is not improving. US gross federal debt passed $39.9tn on Monday 17 August 2026 and crossed $40tn that week, months ahead of the CBO's $39.4tn projection for the fiscal year, with annual interest projected above $1tn (19 August). Congress set the limit at $41.1tn; budget analysts now expect it to bind by early next year. And the buybacks that are propping up are being funded by issuing shorter paper into a front end where the New York Fed halted its reserve management purchases entirely from 14 August to 14 September 2026.
So the bond problem has been converted into a currency problem, and the currency market has noticed. That is not a fix. It is a change of venue.
What would make this dangerous
The repricing turns into something worse if the drift becomes a policy target. Stephen Miran's November 2024 Hudson Bay Capital essay, "A User's Guide to Restructuring the Global Trading System", is the document to watch here; summaries of it describe a plan for weakening the dollar that involves pressuring foreign governments to stop accumulating dollar reserves. Deliberately reducing foreign demand for your own liabilities while running a $40tn debt is a policy with a very narrow margin for error, and the error shows up in the exchange rate first.
The observable things that would this from slow to fast:
The US issues debt denominated in a foreign currency. This is the 1978 tell. Nearly 10bn DM of Carter bonds were the admission that dollar liabilities alone would not fund a dollar defence. There is no sign of this today, and it would be unmistakable.
Official selling accelerates in the TIC data rather than drifting. China down $690bn from its 2013 peak over thirteen years is a reallocation. The same magnitude over four quarters would be a run. Watch the month-on-month prints, not the levels.
COFER breaks decisively below its recent range. The dollar's share went up from 56.42% in Q4 2025 to 57.13% in Q1 2026. Two or three consecutive quarters of falls larger than the recent drift would say something the twenty-five-year trend line does not.
Gold rises while real yields rise, repeatedly. A 30-year TIPS yield at 3.09%, the highest since 2008, alongside gold at record levels is the combination that says compensation demands are rising for reasons no rate cut will fix.
is forced to choose between the currency and the debt ceiling. The Volcker resolution worked, and it cost: fed funds from 10.9% in August 1979 to 17.6% by April 1980 and above 19% in 1981, with CPI inflation falling from 14.8% in March 1980 to under 4% by 1983. Doing that now would collide with a debt limit that budget analysts expect to bind in early 2027 and interest costs already above $1tn a year. The 1979 Fed defended a currency against inflation. A 2027 Fed would be defending it against arithmetic.