In a recent address, the head of the Bank for International Settlements (BIS) sounded the alarm over the rapid escalation of capital expenditures in artificial intelligence (AI) technologies. According to the governor, the current wave of AI investment is being driven not by solid, profit‑generating projects but by a fervent belief that the sector will inevitably deliver outsized returns. This optimism, he argued, is encouraging firms and governments to tap into financing sources that are often opaque, complex, and poorly understood. The result, he warned, is a burgeoning debt burden that could pose serious systemic risks to the global financial system.
The BIS chief, Pablo Hernandez, highlighted that the present AI spending spree bears striking resemblance to past speculative episodes, such as the railway boom of the 19th century and the dot‑com bubble of the late 1990s. In each of those historical cases, investors poured massive sums into emerging technologies without a clear view of how those investments would translate into sustainable cash flows.
When the anticipated revenues failed to materialise, markets experienced sharp corrections, leading to widespread financial distress and, in some instances, prolonged recessions. In the AI context, the governor noted that many corporations are allocating billions of dollars to develop large‑scale models, acquire cutting‑edge hardware, and recruit top talent. While these initiatives can indeed generate valuable innovations, a substantial portion of the spending appears to be motivated by a fear of falling behind competitors rather than by concrete business cases.
This “catch‑up” mentality fuels a competitive arms race, where firms feel compelled to outspend rivals to secure a perceived strategic advantage. One of the most concerning aspects of this race, Hernandez explained, is the way it is being financed.
A growing number of companies are issuing high‑yield bonds, tapping private‑equity funds, or relying on shadow‑banking channels that lack the transparency and regulatory oversight of traditional banking institutions. The terms of these debt instruments are often complex, with covenants that are difficult for investors to evaluate.
Moreover, the rapid pace of AI development means that the underlying assets—such as proprietary models or data sets—can become obsolete quickly, increasing the risk that borrowers will be unable to service their obligations. The BIS governor stressed that this opacity hampers the ability of regulators and market participants to assess the true exposure of the financial system to AI‑related debt. Without clear visibility into the scale and quality of the liabilities, stress‑testing exercises become less reliable, and the probability of a sudden loss of confidence rises. In a worst‑case scenario, a cascade of defaults could spread across sectors, amplifying the shock and potentially destabilising banks, pension funds, and other institutional investors.
To illustrate the potential magnitude of the risk, Hernandez referenced data from recent surveys indicating that AI‑related capital spending has grown at double‑digit annual rates in many advanced economies. If this trajectory continues, the cumulative debt linked to AI projects could reach several trillion dollars within a decade.
Even a modest correction—say, a 10% decline in the valuation of AI assets—could translate into hundreds of billions of dollars in write‑downs, pressuring balance sheets and eroding confidence. The governor also warned that the systemic implications extend beyond the financial sector. Many AI initiatives are embedded in critical infrastructure, including healthcare, transportation, and energy. A disruption in funding could stall or abandon projects that are essential for public services, creating broader societal costs.
Additionally, the concentration of AI capabilities in a handful of large firms raises concerns about market power and the potential for anti‑competitive behaviour, which could further distort investment incentives. In response to these challenges, the BIS chief called for a coordinated policy approach. First, he urged regulators to enhance disclosure requirements for AI‑related debt, ensuring that investors have access to standardized information about the purpose, maturity, and collateral backing each instrument.
Second, he suggested that central banks incorporate AI exposure into their macro‑prudential monitoring frameworks, allowing for early detection of emerging vulnerabilities. Third, Hernandez advocated for international cooperation to develop common guidelines on responsible AI financing, similar to the standards that have been adopted for climate‑related risks. Hernandez concluded by emphasizing that while AI holds tremendous promise for productivity gains, medical breakthroughs, and societal advancement, the path to those benefits must be paved with prudent financial practices.
He cautioned that the allure of rapid, hype‑driven spending should not eclipse the fundamentals of sound investment: clear revenue streams, manageable leverage, and transparent risk assessment. By addressing the opacity of AI‑linked debt now, policymakers can help avert a future crisis that mirrors the painful corrections of past technology bubbles, safeguarding both the stability of the financial system and the long‑term health of the AI ecosystem.