In a recent address, the governor of the Bank for International Settlements (BIS) sounded an alarm about the rapid escalation of capital expenditures in artificial intelligence (AI) across the global financial system. The central message was clear: the current AI investment boom is being financed largely through opaque forms of debt, and this hidden leverage could sow the seeds of a systemic crisis if not addressed promptly. The BIS chief, Pablo Hernandez, underscored that the enthusiasm surrounding AI technologies—ranging from large‑scale language models to advanced robotics and predictive analytics—has spurred corporations, venture capitalists, and even sovereign wealth funds to pour unprecedented sums into research, development, and deployment.
While the potential productivity gains and competitive advantages are undeniable, Hernandez warned that the financing structures underpinning this surge are often shrouded in complexity, lacking the transparency required for prudent risk assessment. Historical analogies were drawn to illustrate the perils of hype‑driven spending.
Hernandez referenced the railway mania of the 19th century, when speculative investments in rail infrastructure led to massive over‑capacity, bankruptcies, and a severe market correction. He also cited the dot‑com bubble of the late 1990s, a period when internet‑related firms attracted massive capital despite having little to no revenue, ultimately culminating in a painful crash that erased trillions of dollars in market value.
In both cases, the underlying assets were over‑valued, and the financing mechanisms—often involving high‑risk debt—amplified the fallout when expectations failed to materialize. Applying these lessons to today’s AI landscape, Hernandez emphasized that many firms are issuing debt instruments that are not fully disclosed to regulators or investors. These instruments can include private placements, convertible notes, and other hybrid securities that obscure the true leverage ratios of the borrowing entities.
In some instances, AI startups have secured funding through “soft” loans from strategic partners, where the terms are kept confidential to avoid market scrutiny. This lack of visibility makes it difficult for central banks and supervisory bodies to gauge the aggregate exposure of the financial system to AI‑related debt. The systemic risk arises from several interrelated channels. First, if a significant number of AI‑focused companies experience a downturn—perhaps due to slower-than‑expected adoption, regulatory setbacks, or technological bottlenecks—their ability to service debt could deteriorate rapidly.
Since many of these firms are interconnected through supply chains, joint ventures, and shared infrastructure, distress could cascade across sectors, affecting not only technology firms but also traditional industries that rely on AI for optimization. Second, the concentration of AI debt in a relatively narrow set of financial institutions—often those with a high appetite for risk‑adjusted returns—means that a shock could impair the balance sheets of banks and non‑bank lenders.
This, in turn, could tighten credit conditions more broadly, amplifying the economic slowdown. Hernandez warned that the opacity of these exposures hampers the ability of regulators to implement timely macro‑prudential measures, such as counter‑cyclical capital buffers, that could mitigate the impact of a sudden credit crunch. Third, the rapid pace of AI innovation creates valuation uncertainty. Unlike more mature sectors where cash flows can be projected with some confidence, AI projects often involve long development horizons, uncertain market fit, and evolving regulatory landscapes.
This makes it challenging for lenders to accurately price the risk, leading to potential misallocation of capital. To address these concerns, Hernandez called for a multi‑pronged approach. Enhanced disclosure requirements for AI‑related debt would enable regulators to monitor leverage levels and identify emerging vulnerabilities. He also suggested that standard‑setting bodies develop specific guidelines for assessing the creditworthiness of AI investments, taking into account factors such as technology readiness, intellectual property protection, and regulatory compliance.
Furthermore, the BIS chief advocated for greater coordination among international supervisory authorities. Since AI financing often crosses borders—through multinational venture funds, cross‑listing of AI firms, and global supply chains—a fragmented regulatory response could leave gaps that risk‑takers could exploit. A unified framework would help ensure that risk‑weighting and capital adequacy standards are applied consistently, reducing the likelihood of regulatory arbitrage.
In addition to regulatory reforms, Hernandez highlighted the role of market discipline. Investors, analysts, and rating agencies should demand greater transparency and perform rigorous due diligence on AI projects, rather than relying on hype or narrative‑driven valuations. By pricing risk more accurately, the market can naturally curb excessive borrowing and promote sustainable growth. The warning comes at a time when AI is increasingly being positioned as a cornerstone of future economic productivity.
Governments worldwide are launching AI strategies, and corporations are pledging billions to integrate AI into their operations. While the potential benefits—enhanced efficiency, new product creation, and improved decision‑making—are substantial, the BIS chief’s message serves as a reminder that unchecked financial engineering can undermine those very gains. In summary, the BIS governor’s cautionary note underscores the importance of balancing ambition with prudence. The AI capital expenditure race, if financed through opaque debt structures, poses a tangible threat to financial stability.
By drawing lessons from past bubbles, improving transparency, strengthening supervisory coordination, and fostering disciplined market behavior, policymakers and industry leaders can help ensure that the AI revolution proceeds on a solid, resilient foundation rather than on a precarious house of cards.