[Financial Planning for the 100-Year Life] Is the AI Feedback Loop the Engine of Growth or a Financial Illusion?

When evaluating the artificial intelligence (AI) investment boom, it is not sufficient to focus solely on the technology’s potential. One must also consider who is providing the capital, where the money is flowing, and ultimately, who bears the final costs. Recently, within the AI industry, a distinct cyclical relationship has emerged in which companies act as both investors and customers, thereby driving each other’s growth and value.

Certain big tech companies and financial investors purchase stakes in AI model companies or extend credit to them, in addition to signing long-term supply contracts. AI companies use these funds to acquire large quantities of graphics processing units (GPUs), cloud services, electricity, servers, and data center capacity. The sales of suppliers increase, and expectations for expanding AI demand boost both share prices and corporate valuations. The higher valuations, in turn, enable further investments and long-term contracts. This forms a structure in which capital, purchases, and revenue continuously reinforce each other.

The act of investing itself is not the problem. Railways, communications networks, the internet, and power grids all required significant upfront investments in their early stages. Likewise, for AI, computing and electricity infrastructure must come first in order for broader industrial productivity gains to follow. However, if the origin and destination of this cycle remain confined within the industry, the story changes. If investment capital is simply converted into the AI company’s purchasing power, which then cycles back as revenue for the investing companies, the resulting increase in reported sales may not necessarily reflect genuine end-user demand.

The key issue is whether external customers will continue to pay costs to AI companies over the long term. Merely noting a rise in GPU and cloud purchases does not mean that businesses and consumers are actually paying more for AI services. Investors should look beyond sales growth rates to examine the quality of those sales. The proportion of revenue from external clients, customer concentration, accounts receivable growth, termination conditions of long-term purchase agreements, and the degree of linkage between investments and supply contracts are all critical factors. It is also important to distinguish whether massive capital expenditures are translating into stable fee income or are being sustained solely through borrowing.

This cyclical structure may become even more vulnerable when coupled with three fissures in the U.S. stock market. The first is overvaluation. The Shiller price-earnings ratio of the S&P 500 stands at 42, more than twice the long-term average of 17. While AI has great potential to enhance productivity and profitability over the long term, the market is now asking, "When will these astronomical infrastructure investments turn into cash flow?" If the pace of investment or earnings falls short of expectations, both earnings forecasts and valuations may be revised downward simultaneously.

The second issue is instability in the Treasury market. With U.S. government debt reaching 40 trillion dollars, the strategy of purchasing long-term bonds while increasing short-term issuances may buy some time, but it does not address the root causes of fiscal deficits and rising interest expenses. Measures intended to suppress long-term interest rates could actually erode confidence in the dollar, spurring greater appetite for alternative assets such as gold, raw materials, and Bitcoin.

The third risk is persistently high oil prices. With ongoing tensions involving Iran and geopolitical risks in the Strait of Hormuz, elevated oil prices could postpone both inflation stabilization and anticipated interest rate cuts. This increases funding costs for AI companies and amplifies the discount rate pressure on overvalued growth stocks. In Korea, this could transmit through channels such as higher import costs, volatility in the KRW-USD exchange rate, and the possible outflow of foreign capital.

There is no reason to doubt the long-term growth prospects of AI. However, in the fourth quarter of this year, the AI industry will face a series of simultaneous tests: the verification of AI earnings, the direction of long-term interest rates and oil prices, and whether liquidity is absorbed by initial public offerings (IPOs) of major AI firms. Considering that the inverse of oil prices and the yen-dollar exchange rate have historically led the S&P 500, one should also prepare for the possibility that a market correction beginning in the fourth quarter could spread into a broader collapse of the AI bubble by 2027. What is needed now is not concentrated investment chasing the AI trend, but diversification into companies outside this sector that generate actual cash flow, as well as into assets that can withstand changes in interest rates and oil prices.

Kim Youngik, Adjunct Professor at Hanyang University’s Institute for Future Talents

This content was produced with the assistance of AI translation services.