Reserve managers are rotating out of the dollar’s asset while keeping its unit. The data say so, and the distinction matters for what comes next.
Two facts from the reserve data sit awkwardly together until the distinction is drawn. In 2025 gold passed US Treasuries as a share of official reserves at market value, 27 percent against 22 percent at year-end according to the ECB, a first in the modern data even allowing for the price effect. In the same period the dollar’s share of allocated currency reserves held near 57 percent, where it has sat for years; the IMF’s COFER series puts it at 57.13 percent in the first quarter of 2026, against 71.01 percent at the end of 1999. The de-dollarisation debate reads these as one story. They are two.
The dollar as an asset is the Treasury: the obligation of a sovereign that crossed $40 trillion of debt on 19 August 2026 and pays more than $1 trillion a year in interest, with the ten-year yield closing above 5 percent on 15 September 2026 for the first time since 2007. That is the thing reserve managers are quietly reducing. The dollar as a unit is the language of the world’s obligations: the currency in which trade is invoiced, debts written and collateral margined. None of those uses requires holding the Treasury. They require only that everyone else keep using the dollar, which is how a language survives: because its speakers keep using it.
Tranchco calls this the divorce. It matters because the two layers can be renewed separately. The unit needs an income beneath it: the United States’ privilege was always funded by earning more on its risky foreign assets than it paid on the safe liabilities the world held, and as the real return on the safe liability erodes, the search for another income line becomes urgent, which is where the system stood in 1974. The first recycling found that income in emerging-market borrowers, through unsecured dollar loans with the mismatch on the borrower, and it failed in 1982. The second recycling, if it comes, places dollar-denominated claims on local sovereign collateral, with the currency tail retained at home by election. The unit is kept; the collateral beneath it changes.
The hypothesis has a public falsifier. A denominator does eventually follow its asset down; sterling’s did, over roughly half a century. The test is dollar invoicing: the day it follows the reserve rotation downward, the lag is closing and the thesis is wrong. That series has not moved. The data behind this page are tabulated, with sources and dates, on the data page; the argument is The Dollar Denominator, September 2026, which is public.