China’s holdings of U.S. Treasuries have fallen a long way. From a peak of $1.317tn in November 2013 to $633.4bn in June 2026. A decline of just over half. On paper, that looks like China walking away from the dollar.
The money has not left. Part of the headline decline is real, but it reflects how the money is counted rather than who owns it. TIC data attributes a Treasury bond to whichever country its custodian is based in, not to whoever actually owns it. When China shifts custody of a bond it still owns to an account based in Belgium, Luxembourg, the UK or France instead of one that reports directly as China, the TIC figure for China falls even though nothing has been sold.
A separate pool of dollar exposure sits outside this reporting category entirely. China’s state-owned commercial banks hold dollars in their own right, through their own lending and trade financing, ~$1.1tn. Its policy banks, state lenders focused on infrastructure and trade financing rather than ordinary commercial banking, hold a further ~$1tn, mostly in dollars, with no official breakdown published. China’s sovereign wealth fund, the China Investment Corporation, known as CIC, holds a further $450bn in foreign assets, also mostly dollar-denominated. None of this counts as part of China’s official reserves, because reserves are defined narrowly: assets held directly by China’s central bank, the People’s Bank of China, known as the PBoC, and its foreign exchange manager, the State Administration of Foreign Exchange, known as SAFE. Money sitting on a separate institution’s balance sheet does not qualify, however state-owned that institution is.
Add it all together, custody reattribution and separately-held pools both, and China’s real exposure to the dollar looks close to unchanged from before 2022. It has not shrunk, but scattered across more places than it used to sit in, which makes it harder to add up from outside and easy to misdiagnose for a Chinese retreat.
Simultaneously, the opposite is happening with gold. The PBoC has bought gold for 22 consecutive months, taking its declared reserves to ~2,386t (tonnes) by the end of August, ~8% of its own total reserves. But the published numbers do not match the amount of gold entering the country. In May alone, China imported 163t of gold, declared only 9.95t as an addition to reserves, and saw withdrawals from the Shanghai Gold Exchange, known as the SGE, the domestic wholesale market that normally accounts for where imported gold ends up, fall to 63.5t, the lowest reading since February 2020. Widen the lens to the full first half of the year (H1) and the same gap holds at scale: ~865t imported, ~598t moving through the SGE, and only 40t declared as a reserve addition. That leaves ~227t unaccounted for by either official channel, across H1.
Analysis of UK customs data, tracking physical gold shipments leaving London for China rather than relying on anything China itself discloses, put the total brought in through that route at ~1,050t since February 2022, as of last year. That estimate has not been refreshed since, and the real number today is certainly higher. Albeit, a genuine counter-argument exists: some of the PBoC’s reported increase in reserve value likely reflects the rising gold price itself, not new tonnes bought. It does not touch the import versus domestic demand gap above. That gap concerns physical gold entering the country, not how it is valued afterward.
These two facts do not confirm each other. Gold and dollars are not being compared here in a direct sense, and showing that one figure is overstated while the other is understated does not, on its own prove the why. The natural logic may be that China is swapping one for the other: reducing real dollar reliance and using gold as the backing behind that shift. However, the evidence does not support that reading. If gold were ‘replacing’ dollar exposure, the true scale of that exposure, declared reserves plus the hidden pools sitting in state banks, policy banks and CIC, should be falling as gold rises to take its place. It is not, it has stayed close to flat. That is what rules out a currency swap. What is happening is a jurisdiction swap. China is not shrinking its dollar exposure and buying gold in its place. It is holding dollar exposure roughly steady while separately building a gold position that sits inside China’s own vaults rather than a foreign custodian’s. The two moves manage the same underlying risk from opposite directions. Scatter the dollars across enough institutions and countries that no single freeze order can reach all of it. Bring the gold home so it never has to sit inside someone else’s jurisdiction.
The same underlying risk is live again this week, for Russia, whose frozen reserves in 2022 were the reason China drew this lesson in the first place. The EU is discussing moving roughly €210bn of Russia’s frozen assets out of Euroclear, the Belgium-based institution that holds and settles securities on behalf of investors worldwide, and into a new EU-run structure. Belgium does not want to keep carrying the legal exposure alone, and Russia has filed a $230bn lawsuit against Euroclear in a Moscow court. Nothing has been seized, but the assets remain frozen, not confiscated. The detail that matters is that the custodian wants out. Euroclear was built to be a neutral piece of plumbing, holding and settling securities without taking any side. Being the mechanism used to freeze a sovereign’s reserves has pulled it into a geopolitical dispute it was never designed to be part of. It is not only sovereign reserves that turned out to be exposed. The ‘apolitical’ infrastructure turned out to be exposed too.
Infrastructure tells us what politics will not. CIPS, short for the Cross-Border Interbank Payment System, lets banks settle yuan transactions without routing through the Western financial system, meaning the SWIFT messaging network alongside the dollar clearing systems, CHIPS and Fedwire, that actually move dollars between banks. CIPS processed ~$25.5tn in transactions in 2025 and is on pace for ~$30tn this year.
Compare that with mBridge, a separate project China has built alongside several other central banks. mBridge has a more ambitious aim than CIPS. Rather than building a parallel messaging system, it lets central banks settle payments directly with each other on a shared digital ledger, bypassing the correspondent banking system completely. Total volume moved through it since 2021 sits at only ~$55bn, a fraction of CIPS’s scale, and the Bank for International Settlements, known as the BIS, the institution often described as the central bank for central banks, withdrew from the project entirely in October 2024, citing concerns that it could be used to circumvent sanctions. Not everything China builds succeeds at the same pace. CIPS is scaling fast. mBridge, the more ambitious bet, has stayed small. Despite CIPS’s growth, the yuan’s share of global payments messaged through SWIFT has stayed flat at ~3% for years, close to fifth among world currencies, against an economy that accounts for ~18% of global GDP and ~13.5% of global trade. That gap between economic weight and currency weight is worth an eye.
The same direction shows up outside of payments infrastructure. In 2025, China converted dollar-denominated debt owed by Kenya, Angola and Ethiopia into yuan terms, part of a broader lending relationship in which China is now a major creditor to dozens of lower-income countries. China’s onshore bond market is now the world’s second largest, with more than $25tn outstanding. Foreign investors hold ~¥4.2tn of it, mostly through Bond Connect, a route that settles offshore in Hong Kong and avoids dealing with mainland capital controls directly. Bilateral trade settled directly in yuan has ~tripled as a share of China’s total trade since 2017, from a tenth to ~a third, though Beijing has attached no public target to how far it intends to take this. Russia settles more than 90% of a $245bn trading relationship with China in yuan and roubles. Brazil has run a real-yuan settlement arrangement since 2023, backed by a swap line worth ~¥190bn, covering ¥171bn of bilateral trade last year.
China does not hide the goal. It hides the pace. The Five-Year Plan calls for the yuan’s internationalisation to proceed “steadily” and “prudently.” Officials describe it as a process measured in decades, not a campaign with a deadline. A 2026 industry survey of more than 2,500 domestic and over 1,000 foreign firms found that more than 60% of the foreign companies polled plan to hold their yuan usage steady rather than increase it, which cuts against reading this as a sudden or accelerating shift. The approach is selective, favouring trade settlement over any move to open China’s capital account.
That caution holds real weight, because it reflects a constraint rather than a free choice. Fully opening China’s financial system would reopen the door to the capital flight the country fought hard to shut down a decade ago. Between mid-2014 and early 2017, China’s reserves fell by ~$1tn as the central bank sold dollars to defend the yuan, after a surprise devaluation in August 2015 triggered a rush of money leaving the country. Much of it left through Hong Kong, converted into dollars, some disguised as ordinary trade through manipulated invoices. A visible share went straight into property in Vancouver, Sydney, London and several U.S. cities, all of which saw a noticeable wave of Chinese buying during this window. Beijing has not forgotten that episode. Its slow, gradual approach to internationalising the yuan is that memory, still shaping policy today.
There is a historical precedent for this gap between what the official numbers show and what actually happened. The conventional account of the dollar’s rise says sterling still held somewhere between 80 and 90% of global currency reserves as late as 1945 to 1947, and that the dollar only took over afterward, formalised by the Bretton Woods agreement in 1944.
Look past the official reserve statistics and the picture changes. Research using bond market data, by the economic historians Barry Eichengreen and Marc Flandreau, found that the dollar had already overtaken sterling as the currency international borrowers preferred to issue debt in, by the mid-1920s, ~two decades before the reserve figures caught up. That is not the same as saying investors chose US government debt over British government debt. It means that governments and companies anywhere in the world, borrowing internationally with no particular connection to the United States, increasingly chose to denominate their bonds in dollars rather than sterling. That shift was effectively complete by the late 1920s. It simply did not show up anywhere official for another two decades.
When sterling’s decline finally did show up in the official reserve data, the catching-up process itself was not smooth. Sterling’s share of global reserves fell to ~58% by 1950 and to ~11% by 1973. The sharpest jolt came in 1956, during the Suez Crisis. Britain, France and Israel invaded Egypt without American backing, and Washington saw the whole coalition as a liability. Britain and France looked like colonial powers reasserting control, and Israeli presence made the episode look like a Western-aligned attack on an Arab nationalist leader, exactly the optics the US was trying to avoid across the wider region during the Cold War, for fear of pushing Arab nations toward Soviet support. The United States made its support conditional. It declined to back sterling and blocked Britain’s access to International Monetary Fund, known as IMF, assistance until British troops withdrew, even though Britain was one of America’s closest allies. Britain withdrew, and sterling suffered a serious run in the process. It took until 1976, and a full IMF bailout of the United Kingdom, for sterling’s fall from reserve-currency status to reach its final marker.
The lesson in that episode is not about Britain. It is that financial dominance gets used as leverage even against allies, whenever it serves the dominant power’s own strategic interest, and that the official numbers describing a currency’s status are reliably a lagging indicator, catching up years or decades after the underlying shift has already happened.
Which is where this has now landed us, a pattern. Central banks are not abandoning the dollar at the pace the headlines suggest, and China’s own numbers, read properly, show a country repositioning rather than retreating. But the pressure on the dollar’s reserve-currency status is real, and it is not coming from a rival currency winning the argument on its own merits. It is coming from a world moving away from a single, unquestioned centre of financial power toward one with several competing centres, at exactly the moment several of those centres have watched financial infrastructure become weaponised, and a live conflict over the Strait of Hormuz puts a fifth of the world’s oil supply, priced overwhelmingly in dollars, at risk. Nobody knows which currency, or which system, ends up ahead. That uncertainty is the condition under which gold thrives, and it is thriving again now.
The dollars leaving are the least interesting part of this story. Where trust is going is the part worth our time.