The International Monetary Fund (IMF), in its latest Global Financial Stability Report, has identified a potential correction in artificial intelligence-related assets as a primary risk to financial markets. The Fund forecasts that a sharp downturn in the AI boom could lead to a 0.4% reduction in global economic output and a 1.3% decrease in global investment by 2028. The IMF specifically links particular vulnerability to developed Asian economies.
This IMF risk is prioritized above rising bond yields and cyberattacks. In an accompanying blog post, IMF economists Atanasios Vamvakidis and Jason Wu warned against complacency.
Current Market Conditions and Risks
Currently, AI supports the global economy and financial markets through investments in data centers, chips, and infrastructure, as well as strong profits from companies driving this boom. Markets have shown resilience due to solid business results related to AI. However, if investments prove weaker or less sustainable than expected, or if profits fail to meet forecasts, AI-related stocks and other assets could lose value. The IMF notes that growing debt burdens and concentration of risks could intensify such a correction.
The report shows that the AI boom is increasingly financed through borrowing. The share of major global cloud providers, known as hyperscalers, in issuing investment-grade bonds has risen from 1% in 2024 to 10%. Simultaneously, their free cash flow is shrinking amid increased AI-related expenditures.
Investors are also actively using leverage. The proportion of individual stocks in leveraged Exchange Traded Funds (ETFs) has increased from 16% to 28%. The share of hedge fund equity, combined with borrowed funds, has grown from approximately 10% to 15% of GDP in developed countries.
Circular Financing and Dependency
The IMF is particularly concerned about so-called circular financing within the AI industry. The Fund examined 14 publicly disclosed deals since 2023. Under these agreements, hyperscalers invested about $300 billion in AI developer equity and provided approximately $650 billion in debt financing. In return, recipients committed to spending $1.4 trillion together with their investors in the future, including funds for data centers, computing power, and cloud services.
According to the report, the financed companies have about $320 billion in debt on their balance sheets, part of which is partially guaranteed by investors or off-balance sheet. The IMF explains the problem of mutual dependence created by these agreements. Companies receiving funding are typically in a significantly weaker financial position than the hyperscalers. If one party encounters difficulties, expected investor returns will decline, reducing the value of their stakes and weakening the reliability of loans associated with these deals.
The Fund characterizes this as 'wrong-way risk': investments lose value precisely when the associated income ceases.
Recommendations and Forecasts
To mitigate these risks, the IMF recommends that regulators strengthen oversight of borrowing used to purchase stocks, as well as leveraged ETFs, as they can amplify both market rallies and declines. Where risks are significant, the Fund suggests introducing industry-specific capital requirements and limiting banks' exposure to the AI industry. According to IMF stress tests, global banks will maintain sufficient capital even in a scenario combining an AI market correction with a sharp rise in bond yields. Nevertheless, financial stability risks remain high.
The Fund also warned that governments and central banks have fewer options to respond to a major shock than in most periods over the last two decades, as the space for fiscal and monetary policy has narrowed. IMF Director Kristalina Georgieva also noted in a blog that corporate market valuations are outpacing actual productivity growth.
In addition to market risks, the Fund pointed to 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 developed countries are susceptible to AI influence. Georgieva called on regulators to enhance supervision and cybersecurity while reforming tax policy to finance worker retraining without restricting innovation.
The report emphasizes, however, that hyperscalers remain highly profitable in their core operations, generate stable cash flow, and possess high credit ratings, while their investors are generally financially sound. Since AI-related profits have repeatedly exceeded forecasts, the price-to-earnings ratio has even decreased.
According to Tech Times, analysts at Goldman Sachs and Morgan Stanley consider concerns about an AI bubble premature, and the JPMorgan indicator system showed that the current AI boom does not yet meet bubble criteria.