CICC: “De-dollarization” is still the general trend

The Zhitong Finance App learned that CICC released a research report saying that “de-dollarization” is still the general trend: as US government debt continues to rise, eroding the “safe asset” consensus on US debt, the short-term stabilization of the share of US foreign reserves does not mean a change in the long-term trend. The decline in reserve currency status often starts with a few countries reducing their holdings, then gradually spreads to most countries reducing their holdings, and the share of foreign reserves maintains a downward trend amid fluctuations. Stocks are information-sensitive assets, and government bonds are information-insensitive assets. The characteristics of the two types of assets are fundamentally different, so it is difficult for US stocks to replace US bonds to assume the function of reserve assets. The industrial trend and debt impact of the AI revolution are highly variable, and there is an “reverse Balassa-Samuelson” effect, which may not necessarily strengthen the US dollar's position as an international currency.

CICC's main views are as follows:

There is no contradiction between the short-term recovery in the share of US dollar foreign reserves and the long-term downward trend

In 2025, the share of the US dollar in the world's allocated foreign exchange reserves fell from 58.4% to 56.4%, but it rebounded to 57.1% in the first quarter of 2026, which seems to reflect a major adjustment in the views of central banks around the world on the US dollar. However, after excluding the impact of changes in the exchange rate of the US dollar compared to other major currencies, the bank discovered that since the fourth quarter of 2024, the share of US dollar foreign exchange reserves by central banks around the world has only fluctuated slightly, and has not reversed the previous continuous downward trend, but there are indeed signs of a phased slowdown. Breaking up the global foreign reserve structure, it can be seen that only a few countries are reducing their holdings of US dollar reserves, and most countries have not significantly reduced their holdings. Based on this, some opinions suggest that “de-dollarization” is not established because the share of US dollar foreign reserves has stagnated, and only a few countries have reduced their holdings of US dollars (Goldberg and Hannaoui, 2024; Weiss, 2025).

The bank holds the opposite view, believing that the “de-dollarization” process continues, and that the phased slowdown does not contradict long-term trends. In the short term, only a few countries reduced their dollar reserves; on the contrary, it is in line with historical laws. In the process of the decline in the reserve status of the pound, the decline in the share of the British pound's foreign reserves also began with a few countries taking the lead in adjusting.

Before World War I, the share of the pound in known official foreign exchange reserves fell from 64% in 1899 to 48%. This was mainly driven by the increase in franc holdings by a few countries such as Russia, Italy, and Greece, and the corresponding share of the French franc in foreign reserves rose to 31% (Lindert, 1969). Since then, although the balance of foreign reserves of the pound was surpassed by the US dollar from 1924 to 1926, since France, the largest holder of foreign reserves, accumulated a large amount of pounds from 1926 to 1927, the share of the British pound's foreign reserves rebounded and surpassed the US dollar again. From 1929, France exchanged British pound reserves for gold, and the share of the pound's foreign reserves declined again. By 1931, when other gold-based countries also began selling the pound, the reduction in British pound holdings gradually spread from individual large holders to more countries (Eichengreen and Flandreau, 2009).

As can be seen, “de-poundization” continued for about 30 years from 1899 before spreading from individual countries to most countries. During this period, the share of the British pound in foreign reserves also rebounded. Therefore, just because the number of countries that have been de-dollarized is small, or that short-term changes in US dollar reserves have stagnated, are not enough to refute the long-term trend of de-dollarization.

In fact, behind “de-dollarization” and the declining share of US dollar reserves, the “anchor” reflecting the hegemony of the US dollar is shaking, and the threshold for reversing the trend is very high. The US dollar's status as an international reserve currency stems from the “double anchor” of the US dollar's hegemony: the “anchor of law” comes from the US national credit, supported by economic strength and public system; the “anchor of function” comes from the broad, deep, and efficient dollar asset market. This supports the two basic requirements of official foreign exchange reserves: it is necessary not only to pursue credit quality, that is, reserve assets can be preserved for a long time and even appreciated during crises, as “safe assets” (Brunnermeier et al. 2024); it is also necessary to pursue liquidity, which requires assets to be monetized at any time for external payments and market intervention. US bonds have both high credit quality and market liquidity, so they are the main assets that carry foreign reserves in the US dollar.

However, due to the continued expansion in the size of US debt in recent years, the performance of US dollar reserve assets is divided: the US dollar and short-term US bonds still have strong cash-like properties and can meet transaction, settlement, financing, and collateral needs, and the trend is relatively good; however, the trend of long-term US bonds is weak and declined when crises such as the enactment of “equal tariffs”, the Greenland incident, and the US-Iran conflict broke out. The “weaponization of the dollar” has further increased the risk of foreign official institutions holding long-term US bonds, driving them to other reserve assets such as gold (Arslanalp et al., 2023). Since 2020, the convenience income of medium- to long-term US bonds has declined markedly and turned negative, yet the convenience benefits of the US dollar itself have not contracted at the same time (Du et al., 2025). This shows that global investors still need dollar liquidity and are willing to pay a premium for it, but they no longer equally recognize the ability of long-term US bonds to store value.

This differentiation will change the allocation of US dollar reserve assets from a safe-haven allocation to a liquidity allocation: although countries choose to continue to hold large dollar reserve assets, they will also reduce the weight of long-term US debt in reserves. Furthermore, the recovery in the share of US dollar foreign reserves in early 2026 was also affected by changes in exchange rate valuations and asset allocation, reflecting more short-term fluctuations, which is not enough to prove that the reserve appeal of long-term US bonds has recovered. The expansion in the size of US debt and the specter of the weaponization of the US dollar remain, and the long-term downward trend in the share of US dollar foreign reserves will not change.

It is difficult for US stocks to replace the reserve asset status of US bonds

US stocks have performed well in the past few years. The S&P 500 index has continued to rise and reach new highs, and the risk premium has continued to drop to almost zero. According to some opinions, US stocks can partially replace US debt and continue to support the demand for US dollar reserve assets: the US has the world's largest stock market with high liquidity and good long-term returns; foreign official institutions can reduce the allocation of US bonds and increase the allocation of US stocks and corporate bonds; US corporate securities can continue to bear the demand for US dollar reserves. Judging from the data, foreign official investors continue to buy US stocks, which seems to support this view. According to the US Treasury's TIC data, from May 2025 to April 2026, foreign official institutions made net purchases of US stocks of 120.9 billion US dollars and 49.3 billion US dollars of corporate bonds, while net sales of long-term US bonds of 17.5 billion US dollars and institutional bonds of 33.3 billion US dollars.

The bank believes that as risky assets, US stocks are difficult to replace the role of US bonds as safe assets in global foreign exchange reserves. Holmström (2015) notes that stocks rely on continuous information production and price discovery, and are information-sensitive assets. Safe assets, on the other hand, need to maintain high credit quality and market liquidity, and maintain value when risk rises. This means that safe assets must be information-insensitive assets, which fluctuate less during economic and financial market crises, so that official agencies can use them at any time to interfere with foreign payments and exchange rates and assume the function of reserve assets. Judging from historical experience, stocks retraced significantly more than government bonds during the crisis. Baele et al. (2020) reviewed historical asset data from 23 countries and found that when the financial crisis occurred, government bonds obtained an average excess return of 2.7% compared to stocks. After the impact of trade policy in April 2025, US stocks, US bonds, and the US dollar fell simultaneously; US stocks were sold off as risk appetite declined, and did not show the ability to preserve value during the crisis (Shin et al., 2025), making it difficult to replace them as safe assets.

Looking back at history, the yen's experience shows that currency demand attracted by high stock returns makes it difficult to cross the technology and asset price cycle. In the 1980s, the competitiveness of Japan's electronics industry increased, international capital inflows drove stock market valuations to rise, and the share of yen in global foreign reserves continued to rise. The price-earnings ratio of the Japanese stock market in 1988 was about 32.6 times, which is beyond what can be explained by real interest rates and profit growth (French and Poterba, 1991). The Japanese stock market declined for a long time since 1990, and the share of yen foreign reserves also peaked at about 9% in 1991, then continued to decline against the backdrop of a long-term downturn in the Japanese economy and financial system, and fell to 3.9% in 2003 (Eichengreen and Mathieson, 2000; Park and Shin, 2012). As can be seen, after all, stock prices depend on technological prospects and corporate profits. Once the return advantage disappears, international capital will be reallocated. Although risk assets can gradually expand international demand for a currency, it is difficult to support its position as a reserve currency in the long term.

Furthermore, the increase in US stocks by some foreign official institutions is essentially still a pursuit of relative earnings. In 2025, the stock markets of the world's major economies generally outperformed US stocks, and in the first half of 2026, the Korean and Japanese stock markets gradually outperformed US stocks. This shows that even in the context of the AI revolution, high-return assets will rotate between different markets according to profit cycles and industry trends, and will not always be concentrated in the US stock market. For sovereign wealth funds, stabilization funds, and central bank revenue-enhancing accounts, capital can flow in due to higher returns from US stocks, and may also be reallocated when other markets improve relative returns. Such income-driven official capital inflows are unstable, making it difficult to generate the sustained dollar demand required for central banks' core external reserves.

“AI dollar” is difficult to reverse “de-dollarization”

If the “safe asset” consensus on US bonds is shaken, and US stocks cannot replace US bonds to assume the official reserve function, then are there any other forces that can reverse the “de-dollarization” trend? Recently, a new opinion has indeed emerged in the market. It is believed that the AI revolution will consolidate the US dollar's position as a reserve currency. The main logic is that the US and its allied companies control some key nodes of advanced chips, cloud services, and computing power infrastructure, so that related cross-border transactions can be mainly denominated in US dollars. According to some opinions, an “AI dollar” similar to a “petroleum dollar” can be formed based on this, increasing demand for the dollar and supporting the dollar reserve (Ong, 2026). The bank finds the above assertion equally difficult to establish for the following reasons:

First, computing power is significantly different from the supply characteristics of petroleum, making it difficult to form long-term irreplaceable properties similar to petroleum. Petroleum is a natural resource widely needed by the global economy. The distribution of production areas and transportation conditions are difficult to change in the short term, so it is easy to form a natural monopoly in petroleum production and marketing; while models, chips, and computing power services will continue to be replaced with technological progress and industrial competition, the US may not be able to maintain its lead and monopoly for a long time.

In 2025, US private AI investment reached US$285.9 billion, significantly higher than China's US$12.4 billion, but the capital investment gap did not translate into a model capability gap of the same magnitude (Stanford HAI, 2026): In July 2026, Kimi K3, an open source model released by the Chinese AI model company Dark Side of the Moon, rose to first place in the Arena code list, and approached or surpassed the US head closed source model (Arena, 2026; Moonshot AI, 2026). According to OpenRouter data, the US model once contributed about three-quarters of the platform tokens in 2025; as of June 14, 2026, the Chinese model's share of weekly tokens had risen to 55%, surpassing 43% of the US model (OpenRouter, 2026).

Second, the increase in US productivity brought about by AI may not support the dollar; on the contrary, it may depress the dollar through the “inverse Balassa-Samuelson” effect. Cerutti et al. (2025) believe that if productivity improvements brought about by AI mainly occur in non-trade sectors such as education and health care, prices in the non-trade sector fall compared to prices in the trade sector, leading to a decrease in the overall price level of the US compared to overseas, causing the actual effective exchange rate of the US dollar to decline. This is the opposite of the traditional Balassa-Samuelson effect. Once the US dollar enters a channel of continuous depreciation, the share of foreign reserves will also be under pressure: exchange rate valuations will directly reduce the share of US dollar assets in global foreign reserves, and weakening currency values will also weaken the dollar's attractiveness and push the central bank to reduce the dollar allocation (Chinn and Frankel, 2007; Eichengreen et al., 2014). Therefore, AI does not naturally strengthen the US dollar; if the result is a trend depreciation of the US dollar, it will accelerate the decline in the share of the US dollar's foreign reserves.

Finally, the impact of AI on America's long-term solvency is highly uncertain. On the one hand, if AI significantly increases total factor productivity and drives potential economic growth, corporate profits, and tax base to continue to rise, it will undoubtedly help improve America's debt sustainability; but on the other hand, America's debt burden itself will also continue to rise. The US Congressional Budget Office (CBO) predicts that the share of federal debt held by the public in GDP will rise from 101% in 2026 to 120% in 2036, and that the federal budget deficit will increase from 1.9 trillion US dollars to 3.1 trillion US dollars during the same period. Therefore, it is currently difficult to determine whether the growth and increase in fiscal revenue brought about by AI can catch up with or even reverse the trend of debt accumulation.

At the same time, the impact of AI on employment structures and income distribution may also require substantial structural adjustments in fiscal policies. The government may need to increase taxes on AI-related capital gains, excess profits, and high-income groups, while expanding support for the unemployed, affected industries, and workforce transformation. If new taxes can cover related expenses, AI transformation may help improve the fiscal situation; however, if taxes are insufficient and subsidies and social security spending grow rapidly, it may also widen the fiscal deficit over a period of time.

Seen from this perspective, the ultimate impact of AI on America's ability to repay its debts is still uncertain. It depends on multiple factors such as productivity growth, technology diffusion, income distribution, and fiscal policy adjustments. Currently, it is not enough to infer that the US fiscal sustainability and long-term US dollar storage capacity will inevitably improve, and it is also difficult to judge that the long-term downward trend in the share of US dollar foreign reserves will be reversed based on this. Therefore, the advantages of the AI industry can continue the influence of the US dollar in trading and investment, but it is difficult to ease the market's medium- to long-term concerns about US dollar assets as foreign exchange reserves.

In summary, the bank expects the long-term decline in the share of US dollar foreign reserves and the “de-dollarization” trend to continue.