For months, if not years, Artificial Intelligence (AI) has dominated forecasts and assessments for the future, while simultaneously fueling unprecedented investment speculation in the present. Enterprises linked to AI are largely responsible for the rapid ascent of the S&P 500 stock market index. At the same time, investments in data centers constitute one of the primary factors behind rising interest rates in the US, as borrowing for their construction competes in the credit market with the federal government, which faces continuously widening fiscal deficits. When such a large portion of investment relies on a rapidly evolving new technology, the critical question is what will happen if the cash flows and profits of hyperscalers, meaning the tech giants building data centers, as well as the AI firms leasing these infrastructures, prove lower than expected. These two categories of businesses, after all, overlap in several cases. What will the consequences be for financial markets? Are we facing the risk of a new crisis?
Why the AI bubble could burst
Such a scenario could emerge for several reasons. First, the productivity benefits of Artificial Intelligence may have been overestimated. In this case, businesses considering the adoption of advanced AI models could curb their demand for tokens, the fundamental units of data processing utilized by these specific models. Second, the growing availability of open-weight models, which allow users to customize parameters to their own needs, could radically alter the landscape. Users will be able to deploy small language models capable of running on their own infrastructure or even on laptops. In both cases, the revenues investors reap from data centers will fall significantly below initial expectations.
The Magnificent 7 in the crosshairs, stock markets the first victim
The most immediate and visible consequence of a collapse in AI expectations would be a sharp drop in the excessively high market valuations of the so-called "Magnificent 7." These are the seven largest mega-cap technology and related corporations that have played a decisive role in driving the S&P 500 higher. However, a correction in equity markets does not automatically entail a financial crisis or even a recession. If, for instance, Meta's stock declines, Mark Zuckerberg will scale back purchases of castles in Ireland. A slowdown in consumption and capital expenditure will follow, while significant losses will be suffered by investment funds, such as Situational Awareness, specializing in AI-related investments. However, for a stock market correction to evolve into a systemic financial crisis, banks, insurance companies, and other major institutional investors must become entangled. So far, there are few indications that these specific institutions hold substantial direct exposure to AI equity risk.
The real danger lies in the private credit market
Far more alarming are developments in the debt markets. The cost of constructing data centers exceeds the free cash flows of the companies financing them, while revenues from prospective tenants of these infrastructures remain a distant prospect. As a result, a large share of investments is financed through borrowing from so-called private credit funds, which are private investment funds providing loans outside the traditional banking system via private agreements. If demand for data center services proves lower than projected, these specific loans will begin experiencing servicing difficulties. Simply put, borrowers may prove unable to repay their obligations and be driven into defaults. The critical question, therefore, is who has invested in private credit funds.
Pension funds and insurers at the epicenter of risk
Popular perception holds that private credit funds are financed primarily by family offices and high-net-worth individuals, who also invest in venture capital and possess the capacity to absorb occasional losses. The reality, however, is vastly different. The overwhelming majority of private credit capital originates from institutional investors: pension funds, insurance companies, and, among others, university endowments. Furthermore, private credit funds have begun expanding their access to this institutional capital by acquiring insurance companies, whose investment portfolios they can subsequently load up with loans. This practice creates the risk of a crisis in the insurance sector, which could manifest in different ways. A first possibility would be for the insurers' parent companies, namely the private credit funds, to proceed with recapitalizing them, provided they possess the required capital. Alternatively, state guaranty funds, to which healthy insurance companies contribute, could intervene by financing the recapitalization of troubled firms. However, if the entire insurance sector collapses, there will no longer be healthy companies left to fund the rescue of the rest. The third and most extreme option would be intervention by the US federal government via a bailout program, as occurred in 2008 with insurer American International Group (AIG).
Banks are exposed even without direct investments in AI
The risk is not confined to pension funds and insurance companies. Beyond borrowing from institutional investors, private credit firms also raise capital from commercial banks, using their loan portfolios as collateral. In this manner, they increase the leverage of their investments and boost their returns. This means that even if commercial banks do not lend directly to AI companies and data center developers, they may still find themselves exposed to losses. The critical issue is whether regulatory authorities are adequately monitoring these developments and possess a complete picture of the risks accumulating within the financial system.
Slicing and selling loans: A repeat of the 2008 nightmare?
The most alarming element, however, concerns how private credit funds now manage their loans. While they initially operated on a buy-and-hold lending model, an increasing number of funds are now proceeding with the securitization of their loan portfolios. In practice, they pool the loans through special purpose vehicles (SPVs), which issue bonds backed by the cash flows of these specific loans. Subsequently, the bonds are divided into different risk categories, known as tranches. Each tranche carries a different priority in collecting debt-service payments. Certain investors hold priority, while others assume greater risk and are repaid at a later stage. Of particular significance is the so-called "mezzanine tranche," the intermediate tier of medium-risk bonds, which is subsequently sold to other asset managers. To assess how imminent the danger of a new financial crisis is, we would need to know more about both the risk level of these bonds and which investors hold them. Yet, we do not know. If this specific picture recalls the role played by securitizations, special purpose vehicles, the slicing of bonds into risk tranches, and the lack of transparency in the 2007-2008 subprime mortgage crisis, you are not alone.
Lack of transparency could trigger a new global crisis
Just as debt markets fueled the near-collapse of the global financial system following the 2007-2008 crisis, today their exposure to Artificial Intelligence will largely determine future developments. The biggest problem is that we possess minimal publicly available information regarding these specific markets, while regulatory authorities do not facilitate access to such data. In other words, the market is financing a massive investment boom in AI without sufficient transparency regarding where the loans end up, who assumes the ultimate risk, and who will suffer the greatest losses if the bubble bursts. And this is what makes the next crisis potentially far more dangerous: not merely the scale of investments in Artificial Intelligence, but primarily the opacity of the financial system supporting them.
www.bankingnews.gr
Σχόλια αναγνωστών