In a recent address to policymakers and financial market participants, the chief of the Bank for International Settlements (BIS) sounded an alarm about the accelerating pace of capital expenditure in artificial intelligence (AI) technologies. The warning centered on a growing reliance on opaque forms of borrowing to fund AI projects, a practice that could embed hidden vulnerabilities into the global financial system and potentially spark a broad‑based correction if not addressed promptly. The BIS chief, Pablo Hernandez, highlighted that the current wave of AI investment resembles past speculative episodes, such as the 19th‑century railway boom and the late‑1990s dot‑com surge.
In each of those periods, optimism about future profits spurred massive spending, often financed through debt that was not fully transparent to investors or regulators. While those historical bubbles eventually burst, the damage was amplified because the underlying financing structures were poorly understood and inadequately monitored. Today, AI is being touted as a transformative force across industries—from autonomous vehicles and advanced medical diagnostics to algorithmic trading and personalized marketing.
Corporations, venture capitalists, and even sovereign wealth funds are pouring billions into AI research, development, and deployment. However, much of this capital is being raised through complex financial instruments, including private placements, special purpose vehicles, and shadow banking channels that lack the same level of disclosure required for public bond issuances. Hernandez explained that the opacity of these financing arrangements makes it difficult for regulators to gauge the true level of exposure that banks, insurers, and other financial intermediaries have to AI‑related projects.
When debt is hidden in off‑balance‑sheet entities or bundled within structured products, the risk of a sudden loss of confidence can spread rapidly. A shock—such as a major AI system failure, a regulatory crackdown, or a sudden shift in market sentiment—could force borrowers to default, triggering a cascade of write‑downs across the financial sector.
The BIS chief also warned that the competitive pressure to outspend rivals in AI capabilities could create a self‑reinforcing cycle of ever‑higher capital allocation, irrespective of the underlying economic fundamentals. Companies may feel compelled to take on additional debt simply to keep pace with peers, even when the projected returns on those AI investments are uncertain or speculative.
This dynamic mirrors the “arms race” mentality observed during the railway expansion era, where firms raced to lay tracks and acquire locomotives, often financing construction through highly leveraged arrangements that later strained their balance sheets. To illustrate the potential systemic implications, Hernandez cited a hypothetical scenario: imagine a large multinational bank that has allocated a significant portion of its loan portfolio to AI‑focused startups, many of which are funded through undisclosed mezzanine financing. If a regulatory body were to impose stricter capital requirements on AI‑related exposures, the bank could be forced to revalue its assets, leading to a sudden contraction in lending capacity. The ripple effect could extend to other institutions that rely on the same funding sources, amplifying liquidity pressures throughout the financial system.
In response to these concerns, the BIS chief called for greater transparency and standardized reporting of AI‑related debt. He advocated for the development of a global framework that would require firms to disclose the terms, maturity profiles, and risk characteristics of any financing tied to AI projects. Such a framework could be modeled on existing Basel III disclosures but expanded to capture the unique features of AI investments, including intellectual property valuation, algorithmic risk, and data‑dependency considerations. Hernandez also urged central banks and supervisory authorities to incorporate AI‑related credit exposures into their macro‑prudential monitoring tools.
By tracking the aggregate level of opaque AI debt across jurisdictions, regulators could identify emerging concentrations of risk and intervene before they become entrenched. Additionally, stress‑testing exercises should be updated to reflect scenarios where AI projects underperform or encounter regulatory setbacks, thereby assessing the resilience of financial institutions under adverse conditions. Beyond regulatory measures, the BIS chief emphasized the role of market discipline.
Investors, rating agencies, and analysts should demand clearer information about the financing structures behind AI initiatives. When the cost of opaque debt rises due to heightened perceived risk, it can act as a natural brake on excessive borrowing, encouraging firms to align their AI spending more closely with realistic profit expectations.
In concluding his remarks, Hernandez reminded the audience that innovation is essential for long‑term economic growth, but it must be pursued within a sound financial architecture. The lessons from the railway and dot‑com bubbles illustrate that unchecked hype, when coupled with hidden leverage, can lead to severe corrections that reverberate across the economy. By fostering transparency, improving risk assessment, and ensuring that AI capital allocation is grounded in solid fundamentals, policymakers can help safeguard the financial system while still allowing the transformative potential of artificial intelligence to be realized.
The BIS chief’s warning serves as a timely reminder that the excitement surrounding AI should not eclipse prudent financial management. As the sector continues to evolve, balancing ambition with caution will be key to avoiding a repeat of past financial upheavals and ensuring that AI delivers sustainable benefits without compromising systemic stability.