In a recent address to the global financial community, the governor of the Bank for International Settlements (BIS) sounded a stark warning about the rapidly accelerating flow of capital into artificial‑intelligence (AI) technologies. While the enthusiasm surrounding AI is understandable—its potential to reshape industries, boost productivity, and generate new revenue streams is undeniable—the BIS chief cautioned that the current surge in AI‑related capital expenditures (capex) is being financed in ways that are both opaque and potentially destabilising for the broader financial system.
At the heart of the concern is the nature of the debt that is being used to fund AI projects. Unlike traditional corporate borrowing, where lenders can assess the cash‑flow prospects of a venture and price risk accordingly, much of the financing for AI initiatives is being sourced through complex, layered instruments that obscure the true level of leverage. These include special‑purpose vehicles, convertible notes, and other hybrid securities that blend equity‑like features with debt‑like obligations. The lack of transparency makes it difficult for regulators, investors, and even the borrowing firms themselves to gauge the real exposure they are taking on.
The BIS governor drew parallels with two historic episodes of speculative excess: the railway boom of the 19th century and the dot‑com bubble of the late 1990s. In both cases, investors poured massive sums into emerging technologies that promised transformative change. However, the underlying business models were often unproven, and the financing structures were riddled with hidden liabilities.
When the anticipated returns failed to materialise, the markets experienced sharp corrections that rippled through the economy, leading to bankruptcies, job losses, and a loss of confidence in the financial system. In the railway era, investors were lured by the prospect of connecting distant markets and unlocking new trade routes. Yet many railway companies were over‑capitalised, relying on optimistic traffic forecasts and government subsidies that never fully materialised.
The resulting debt overhang forced numerous firms into insolvency, prompting a wave of bank failures and a contraction in credit that took years to recover. Similarly, the dot‑com frenzy saw venture capital and public‑market funds rush into internet startups with little regard for profitability.
Companies were valued on projected user growth rather than actual revenue, and many financed their expansion through convertible debt that later turned into equity at inflated prices. When the bubble burst, the sudden devaluation of these assets left banks and investors holding large, impaired loan portfolios, contributing to a broader recession. The BIS chief warned that the AI capex race bears unsettling resemblances to those past bubbles. First, the hype surrounding AI—driven by high‑profile successes in natural language processing, computer vision, and autonomous systems—has created a narrative that AI will be a universal growth engine.
This narrative, while partially true, often eclipses the reality that many AI projects are still in experimental stages, with uncertain paths to commercial viability. Second, the financing landscape for AI is increasingly characterised by “shadow” borrowing. Startups and even established firms are tapping private equity, venture debt, and non‑bank lenders that operate with less regulatory oversight than traditional banks.
These sources frequently employ covenant‑light agreements, meaning that borrowers have considerable leeway before lenders can intervene, even if the borrower’s financial health deteriorates. The result is a buildup of hidden leverage that may not be reflected on balance sheets or disclosed in public filings. Third, the systemic risk emerges not only from the potential default of individual AI firms but also from the interconnectedness of the financial institutions that fund them. Large banks and sovereign wealth funds are beginning to allocate significant portions of their portfolios to AI‑related assets, often through indirect channels such as pooled investment funds or structured products.
If a substantial number of AI projects were to underperform simultaneously, the losses could cascade through these institutions, amplifying market stress. To mitigate these risks, the BIS governor called for greater transparency in AI financing.
He urged regulators to require clearer reporting of AI‑related debt, including the terms, maturities, and underlying assumptions about cash‑flow generation. Moreover, he advocated for stress‑testing frameworks that specifically incorporate AI investment scenarios, allowing banks and other lenders to assess the resilience of their portfolios under adverse AI market conditions.
In addition to regulatory measures, the speech highlighted the role of corporate governance. Companies embarking on large AI initiatives should adopt rigorous internal evaluation processes, ensuring that projected returns are grounded in realistic assumptions rather than speculative optimism. Boards should scrutinise the cost structures of AI projects, the timeline for monetisation, and the potential for technological obsolescence.
Finally, the BIS chief reminded policymakers that innovation and capital formation are essential for long‑term economic growth. The objective is not to stifle AI development but to channel investment in a manner that safeguards financial stability. By fostering a transparent, well‑regulated financing environment, economies can reap the benefits of AI while avoiding the pitfalls of a debt‑driven, hype‑fueled arm race.
In summary, the BIS leader’s warning serves as a timely reminder that the excitement surrounding AI must be balanced with prudent risk management. Historical lessons from the railway mania and the dot‑com bubble illustrate the dangers of financing speculative technology with opaque debt.
As AI continues to evolve, stakeholders—from regulators and banks to corporate executives and investors—must work collaboratively to ensure that the surge in AI spending does not sow the seeds of a future systemic crisis.