Tashkent, Uzbekistan (UzDaily.uz) — The International Monetary Fund (IMF) has identified a potential correction in artificial intelligence-related assets as the main risk to financial markets in its latest Global Financial Stability Report. According to the fund, a sharp reversal in the AI boom could reduce global economic output by 0.4% and global investment by 1.3% by 2028. Developed Asian economies would be particularly vulnerable, the IMF said.
The fund ranks this risk ahead of rising bond yields and cyberattacks. In a blog post accompanying the report, IMF economists Athanasios Vamvakidis and Jason Wu warned that “there is no room for complacency.”
According to the IMF, AI is currently supporting the global economy and financial markets through investment in data centres, chips and infrastructure, as well as strong earnings from companies driving the boom. Markets have remained resilient thanks to robust results from AI-related businesses. However, if investment proves weaker or less sustainable than expected, or earnings fall short of forecasts, AI-related stocks and other assets could lose value. The IMF said rising debt burdens and concentrated risks could amplify such a correction.
The report shows that the AI boom is increasingly being financed through borrowing. The share of the world’s largest cloud computing companies, known as hyperscalers, in global investment-grade bond issuance rose from 1% in 2024 to 10%. At the same time, their free cash flow is declining as AI-related spending increases.
Investors are also making greater use of leverage. The share of individual stocks in leveraged exchange-traded funds (ETFs) rose from 16% to 28%. Hedge funds’ equity positions, combined with margin borrowing, increased from around 10% to 15% of GDP in advanced economies.
The IMF is particularly concerned about so-called circular financing in the AI industry. The fund examined 14 publicly disclosed deals concluded since 2023. Under these arrangements, hyperscalers invested around US$300 billion in the equity of AI developers and provided approximately US$650 billion in debt financing. In return, the recipients committed to spending US$1.4 trillion with their investors in the future, including on data centres, computing capacity and cloud services.
According to the report, around US$320 billion in debt is recorded on the balance sheets of the financed companies, with some of it partially guaranteed by investors or held off balance sheet.
The IMF said the problem lies in the mutual dependence created by these arrangements. The companies receiving financing are generally in a significantly weaker financial position than the hyperscalers. If one of them encounters difficulties, the investors’ expected revenue would decline, reducing the value of their stakes and weakening the reliability of loans linked to the deals.
The fund describes this as “wrong-way risk”: investments lose value precisely when the revenue associated with them dries up.
To reduce these risks, the IMF recommends that regulators strengthen oversight of borrowing used to purchase stocks and of leveraged ETFs, which can amplify both market gains and losses. Where risks are significant, the fund proposes introducing sector-specific capital requirements and limits on banks’ exposure to the AI industry.
According to the IMF’s stress tests, global banks would retain sufficient capital even under a scenario combining an AI market correction with a sharp increase in bond yields. Nevertheless, financial stability risks remain elevated.
The fund also warned that governments and central banks have less capacity to respond to a major shock than they have had in most periods over the past two decades, as fiscal and monetary policy room has narrowed.
IMF Managing Director Kristalina Georgieva also warned in a blog post that market valuations of companies are outpacing actual productivity growth.
Beyond market risks, the fund highlighted the threat of cyberattacks on critical infrastructure and the potential impact of AI on employment. According to the IMF, 40% of jobs worldwide and 60% of jobs in advanced economies are exposed to AI.
Georgieva called on regulators to strengthen supervision and cybersecurity while reforming tax policies to fund worker retraining without restricting innovation.
At the same time, the report notes that hyperscalers remain highly profitable in their core businesses, generate steady cash flow and have strong credit ratings, while their investors are generally financially sound. Because AI-related earnings have repeatedly exceeded forecasts, price-to-earnings ratios have even declined.
According to Tech Times, analysts at Goldman Sachs and Morgan Stanley consider concerns about an AI bubble premature, while JPMorgan’s indicator system has shown that the current AI boom does not yet meet the criteria for a bubble.