Here's the 'core economic flaw' in the AI bull case as stock and bond trades unravel, according to a top economist | Fortune

Torsten Slok, Chief Economist at Apollo, has been vocal this past week about the potential pitfalls of the AI sector, coinciding with a decline in stock prices related to the industry. On Tuesday, he highlighted a slowdown in growth regarding token spending, indicating that the demand for computing power is both flat and highly responsive to price fluctuations.

Slok noted that many businesses investing in AI have yet to realize significant productivity gains. He suggests that AI is increasingly perceived as a cost rather than a catalyst for growth. “The fundamental economic flaw in the AI bullish narrative is that the marginal cost of producing computing power surpasses the marginal benefit it provides to users, which raises concerns about claims of limitless demand for data centers,” he explained. He elaborated that the necessity for computing resources and data centers hinges on the returns that companies can generate, which have been slow to materialize, leading to a hesitance to invest in AI services.

He argued that the key factor in determining data center investments is not merely their physical capacity or power supply, but rather the prices that users are willing to pay for computing resources. Companies are only inclined to pay premium prices if they see a direct correlation with increased revenue, larger profits, or significant cost savings. Currently, there is scant evidence of these factors being realized on a large scale, Slok noted. “Maybe that will come in the future. Maybe not,” he cautioned.

On Thursday, Slok raised concerns about what he termed “circularity” in token spending. He pointed out that AI firms are under scrutiny for financing arrangements that involve lending money or backing debt for other companies that are also their major clients, casting doubt on the sustainability of their growth. He argued that a similar pattern exists between AI startups and the venture capital firms that fund them, with 80% of the leading AI spenders being backed by venture capital. This suggests that startups are using venture capital to cover their computing expenses rather than generating organic income.

By Saturday, Slok was even more critical, stating that AI has “delivered zero financial payoff” for U.S. companies outside of major players in the tech sector, often referred to as the “Magnificent Seven.” While these top firms have seen improvements in profit margins and revenue per employee, the broader S&P 500 has not experienced any similar gains. He reiterated that Corporations in America do not require the latest AI innovations and are reluctant to pay for them. Even as AI technology becomes more affordable and widely available, traditional companies may struggle to leverage it effectively for financial benefit.

“AI may serve as an effective calculator, but in the realm of business, intelligence is primarily about human judgment,” he stated. He emphasized that judgment is context-dependent and can vary significantly based on the individual or model involved. Importantly, Slok pointed out that AI lacks an understanding of accountability, trust, culture, unwritten rules, ethical dilemmas, or fairness—elements that significantly influence reputational risk and, ultimately, revenue and profit margins.

His observations came during a period of significant decline in AI-related stock prices, particularly among chip manufacturers, as investors question the sustainability of ongoing capital expenditures. A few of the major hyperscalers are projected to spend around $800 billion this year, with capital expenditures expected to exceed $1 trillion next year.

However, AI companies might face challenges in raising the necessary capital. Meta, for example, has already accrued substantial debt and recently announced a new class of shares, partly to fund potential taxation liabilities in California. The company is set to spend between $135 billion and $145 billion this year, and the cost of insuring its debt against default reached a record high on Thursday. The five-year credit default swaps widened by 7.6 basis points to an unprecedented 114.79 basis points.

Smaller AI firms, often referred to as “neoclouds,” are also feeling the pressure. For instance, CoreWeave's stock has plummeted over 30% this month, with its five-year credit default swaps rising to 1,145 basis points. Concerns about the financial landscape extend further. Mark Habib, S&P’s Global Head for Digital Infrastructure, pointed out the “interconnected financing structures” between hyperscalers and smaller AI firms or data center projects. He remarked, “It is safe to assume that there is more debt out there that we are not seeing,” suggesting a lack of transparency in reporting from hyperscalers regarding their financial liabilities.