The advance of artificial intelligence investments is creating new risks for global financial stability, Pablo Hernández de Cos, general manager of the Bank for International Settlements (BIS), said on Thursday (10). The warning involves high valuations, market concentration and a growing reliance on debt to finance the sector's expansion.

The five largest global technology companies are expected to spend more than US$ 1 trillion on AI-related investments between 2025 and 2026, according to an estimate by the BIS. Industry projections indicate that global investment could jump from about US$ 500 billion currently to US$ 3 trillion to US$ 4 trillion by 2030.

The scale is already sufficient to influence growth, trade and financial markets. Data centers, specialized semiconductors, cloud infrastructure and equipment account for a significant share of the investments, while AI-related companies have gained weight in stock markets and in the capital expenditures of major economies.

Debt and concentration amplify the risk

According to Hernández de Cos, spending by the largest companies is beginning to exceed their cash flows, increasing the use of debt and private credit. Part of the financing also occurs through structures considered opaque and interconnected among chip manufacturers, computing providers and AI companies.

One of the examples cited by the BIS is so-called circular financing, in which chip manufacturers and large infrastructure providers buy stakes in AI companies that, in turn, commit to purchasing their equipment or computing capacity. These relationships can make it difficult to assess the financial exposures among the companies.

The risk increases if the technology's commercial returns fall below expectations. In that scenario, a reduction in investments could cause a drop in company valuations and hit creditors and suppliers exposed to the expansion of AI infrastructure. The BIS also notes that the concentration of valuations in a few companies could amplify the transmission of a potential correction to other markets and to consumption.

The institution does not claim that a crisis is inevitable. Hernández de Cos, however, compared the current dynamics to other cycles of major technological investment, such as the 19th-century British railway expansion and the internet company boom of the late 1990s, periods in which the volume of capital invested ultimately exceeded the returns subsequently obtained.

At the same time, the BIS acknowledges the economic potential of artificial intelligence. Studies cited by the institution point to productivity gains of 10% to 65% in specific tasks, especially in programming, consulting and the production of professional texts. The question is how much of this gain can turn into a sustained increase in productivity across the entire economy.

More from Radar