Dedollarization in 2026: What the Data Actually Shows

This site shows the government debt of 138 countries in US dollars. It cannot show you Japan's debt in yen, or Brazil's in reais. That is a presentation choice, not a fact about the debt — and it is worth stating plainly before writing anything about whether the world is moving away from the dollar, because the measurement problem and the subject are the same problem.

So: is dedollarization happening? The honest answer, from the primary data rather than the commentary, is yes, slowly, on one measure — and the opposite on most others. Meanwhile the measure everyone quotes changed underneath the entire debate last year, and almost nobody noticed.

The number everyone cites, and what it actually says

The IMF's COFER survey reports the currency composition of official foreign exchange reserves. It is the statistic behind nearly every dedollarization headline, and it does show a genuine decline. Pulled from the IMF's own data API, the dollar's share of world reserves ran:

That is a fall of 6.80 percentage points over a decade — roughly 0.7 points a year. On the annual series the drift is longer still: 69.74% in 2000, 64.15% in 2015. This is real. Anyone claiming nothing is happening is not reading the data.

But two details rarely survive the trip into a headline. First, the series low was 2025 Q4, at 56.42%, and the most recent quarter went the other way, rising to 57.13%. Second, exactly three quarters have ever printed below 57% — 2025 Q2, Q3 and Q4 — so "the dollar has fallen below 57%" describes a narrow, recent and currently-reversing episode, not a trend line.

The measurement changed underneath everyone

This is the part almost no coverage has picked up, and it matters more than any single quarter.

Effective 2025 Q3, and revised back to 2000 Q1, the IMF eliminated the "unallocated reserves" category from COFER. Historically a large share of world reserves belonged to countries that did not report their currency breakdown, and COFER reported those separately. Now IMF staff impute them. In the most recent quarters the imputed portion was 9.99% (2025 Q3), 10.49% (2025 Q4) and 10.65% (2026 Q1) of total reserves. The advanced-economy versus emerging-market breakdown was deleted entirely, back to 2000.

Three consequences follow. Every dedollarization chart published before December 2025 sits on the old allocated basis and is not directly comparable with one published after. Roughly a tenth of the current figure is now modelled rather than reported. And the revision stops at 2000 — the 1995–99 data still carries allocated coverage of only 56–60%, while quarterly COFER itself only begins in 1999 Q1. That makes any claim of the form "the lowest quarterly share since 1995" simply undefined. We went looking for that quarterly data; it does not exist.

On most other channels, the dollar got stronger

Reserves are one channel. They are not the only one, and the others point the other way.

Currency trading. The BIS Triennial Survey (April 2025) found the dollar on one side of 89.2% of all foreign exchange turnover, up from 88.4% in 2022, on daily volumes of $9.6 trillion — itself a 28% increase.

Offshore borrowing. The BIS Global Liquidity Indicators put dollar credit to non-bank borrowers outside the United States at $14.3 trillion at end-2025, growing 8.5% year on year — which the BIS describes as the fastest annual growth since the third quarter of 2014. The rest of the world is taking on more dollar obligations, not fewer.

Payments. The dollar's share of SWIFT payment traffic rose from 31.8% in 2010 to 47.0% in 2024. One caveat we will supply against our own argument: most of that rise happened early. Measured from the mid-2010s (41.6–44.0%) the increase is only about three points, so "the dollar gained ten points in payments" is largely an artefact of choosing 2010 as the starting line.

Gold overtook US Treasuries — with an asterisk

The most striking genuine development is in the composition of reserve assets rather than currencies. Per ECB analysis published in June 2026, gold reached 27% of global official reserves, overtaking US Treasuries at 22%, with the euro at 15%.

That sounds like a historic reallocation. The ECB's own counterfactual suggests otherwise: valued at end-2023 gold prices, gold would sit at 16% — level with the euro — and Treasuries at 26%. In other words the crossover is substantially a price move, not central banks dumping Treasuries for bullion.

A second inconvenient fact: two of the countries most often named as leading the charge were net gold sellers. Through 31 May 2026, World Gold Council data has Türkiye down 81 tonnes on the year and Russia down 34 tonnes.

The one real crack

Having spent several sections deflating the story, here is the evidence that genuinely cuts the other way — and it deserves its weight.

In April 2025 the dollar and US Treasuries briefly stopped behaving like safe havens. The dollar fell about 12% against the euro over the year, roughly seven points of that after 1 April. The ten-year Treasury yield rose around 50 basis points in a single week, which the BIS noted was the third largest weekly increase since 1986. The ECB observed that the dollar "temporarily failed to act as a natural hedge." The BIS found that the usual correlation between Treasury yields and the VIX collapsed to zero, while the equivalent relationship for German bunds strengthened.

For a few weeks, the asset that is supposed to rally in a crisis did not. Most of those flows reversed by June and no broad reallocation followed. But it was a real stress test, and the dollar did not pass it cleanly.

Claims that do not survive checking

Several widely repeated assertions did not hold up when we went to primary sources.

"BRICS settle 90% of trade in local currencies." The 90% figure comes from the Kremlin's own published transcript, in which President Putin is describing Russia's settlements with its partners — not the bloc's internal trade. It has been laundered into a statistic about BRICS as a whole.

"China's CIPS has overtaken SWIFT." This compares different things: SWIFT is a messaging network, CIPS a settlement system. The frequently quoted rejoinder — that most CIPS transactions still ride SWIFT messaging — traces back to a 2022 ANZ Research estimate relayed by CSIS, which is now four years old, so we would not lean on that specific percentage either.

mBridge and BRICS Pay. The cross-border central bank digital currency platform mBridge is real but small, and the BIS exited the project in October 2024. Circulating volume figures for it originate from a single news agency rather than the platform itself. BRICS Pay is not operational, and published timelines for it contradict one another.

Two pieces of genuine 2025–26 context are worth adding, both cutting against the simple narrative. On 20 February 2026 the US Supreme Court held 6–3 in Learning Resources v. Trump that the International Emergency Economic Powers Act does not authorise tariffs — meaning much of the 2025 dedollarization commentary was premised on a tariff regime that has since been struck down. And the GENIUS Act (Public Law 119-27, signed 18 July 2025) requires payment stablecoins to hold reserves of at least one to one, with Treasury bills capped at 93 days, creating a new and purely dollar-denominated channel of demand for short-dated US government paper.

What this means for a site that counts in dollars

Which brings us back to the lens. Every counter on this site is anchored to a dollar baseline and a dollar growth rate, both fixed until our next quarterly review. A practical consequence is that our counters do not move when exchange rates move — currency shifts are absorbed at the review instead, which is why a revision can occasionally reflect the dollar rather than new borrowing. We set this out in full on our methodology page.

If dedollarization were happening at the speed the louder commentary suggests, that would make our unit of account a serious problem: comparing 138 governments through a currency the world was abandoning would distort every number on the site. At roughly 0.7 percentage points a year, with trading, offshore credit and payments all moving the other way, it is not that problem yet.

But "not yet" is not "never," and the honest position is that our measurement choice has a shelf life. If the reserve trend runs for another two decades, or if one of the payment alternatives stops being a press release and starts being infrastructure, the right way to display this data would change — and we would rather say so now than defend a dollar-denominated view of the world long after it stopped describing one.