The Bank for International Settlements (BIS) warned of risks to financial stability associated with the surge in artificial intelligence investments. The organization believes that a correction in the AI market could impact the global economy.
BIS General Manager Pablo Hernández de Cos stated on September 10 at the Global Fintech Fest in India that AI is already significantly influencing global macroeconomic conditions. According to him, the technological boom is simultaneously altering demand, supply, and financial markets, complicating the work of central banks.
AI Investments Could Reach $4 Trillion
According to BIS estimates, the world’s five largest tech companies will allocate over $1 trillion to AI capital investments in 2025–2026. Industry participants expect further acceleration, with global AI investments potentially growing from about $500 billion now to $3–4 trillion by 2030.
Hernández de Cos noted that an increasing portion of AI investments is financed not through company profits but via debt financing, including private loans.
This increases the potential vulnerability of the financial system if the returns on AI investments fall short of expectations.
BIS Sees Risk of a Bubble
According to the BIS head, valuations of AI-related companies are already high and concentrated among a relatively small number of players. The largest companies are simultaneously increasing capital expenditures and relying more on borrowed funds.
A particular issue is the so-called circular financing. Chip manufacturers and cloud infrastructure operators may acquire stakes in AI companies, which in turn commit to purchasing their computing power and equipment. This creates complex interdependencies that are difficult to assess and monitor.
Hernández de Cos warned that if AI investment returns are lower than expected, reduced investments could turn the current investment boom into a downturn. He also identified the investment race among AI companies as an additional risk factor, which could lead to excessive spending.
BIS compared the current situation to past technological investment booms—such as the railway mania in 19th-century Britain, the electrification of the 1920s, and the dot-com bubble of the late 1990s. In all these cases, the technologies were real, but the capital raised ultimately exceeded what future revenues could justify.
AI Simultaneously Accelerates Growth
However, BIS does not view AI development solely as a threat. The organization highlights the significant potential of the technology to enhance productivity.
According to research cited by Hernández de Cos, the use of generative AI increases productivity on specific tasks by 10% to 65%, with time savings of about 20% to 50%. The median estimate of the long-term effect on total factor productivity indicates an increase of approximately 0.5 percentage points per year.
At the same time, AI may reduce demand for workers engaged in routine cognitive tasks. According to the BIS head, early signs of employee displacement are already appearing in customer support, programming, and administrative work. Nearly 80% of companies, according to his data, are discussing automating production processes and replacing some labor.
Central Banks Face Greater Forecasting Challenges
Another effect of AI is directly related to monetary policy. The technology simultaneously affects productivity, demand, supply, and financial markets, making it more challenging for central banks to assess potential economic output, output gaps, and the natural interest rate.
BIS considers several development scenarios. In one, AI ensures sustainable productivity growth; in another, it leads to self-reinforcing growth; and in a third, a “demand trap” arises: automation redistributes income from labor to AI investments, reducing consumer demand and ultimately slowing economic growth.
Finally, BIS specifically warned about cyber risks. AI can assist both attackers and defenders, but the advantage may lie with the attackers: they only need to find one vulnerability, while defenders must secure the entire system.
However, BIS does not claim that the current boom will necessarily end in a crisis. According to Hernández de Cos, the scale and speed of AI investments require caution, and the ultimate impact of the technology will depend not only on its capabilities but also on policy, competition, benefit distribution, and the quality of institutions.
In August, Bridgewater Associates founder Ray Dalio stated that enthusiasm around AI has inflated a bubble comparable to those of 1929 and 2000, and identified specific mechanisms that could lead to its collapse.
