A prominent financial commentator has warned that emerging financing structures in the artificial intelligence sector are echoing the excesses seen during the dot-com bubble. The veteran observer, who witnessed the market collapse in the early 2000s, expressed a strong desire to avoid a historical repeat. Market attention is currently fixed on a rumored massive guarantee arrangement. Reports indicate that chip giant Nvidia is negotiating to provide up to $250 billion in backstop support for OpenAI, intended to help finance a planned large-scale data center campus in Ohio. Following this news, Nvidia's stock suffered a notable decline, dragging down several other semiconductor stocks. Sources indicate the proposed guarantee would primarily cover lease and construction debt for the project, rather than the chips themselves deployed within the data centers.
These discussions highlight that AI financing is becoming increasingly circular. In recent years, the chip giant has invested in numerous clients and ecosystem partners, who simultaneously serve as major buyers of its chips. This structure evokes comparisons to the telecommunications equipment industry of the late 1990s. Back then, equipment manufacturers helped clients finance large purchases to drive sales. However, once buyers faced cash flow constraints and failed to pay, those transactions spiraled out of control quickly, inflicting heavy losses on suppliers and investors. Historical experience suggests that lending money to companies purchasing your own products carries significant risks.
Despite the warnings, the commentator maintains that Nvidia remains an extremely robust company and is not necessarily predicting a repeat of the dot-com crash. The genuine concern lies in the potential for investors to lose confidence rapidly, particularly when suppliers become overly reliant on customers whose massive spending depends on continuous access to capital. If buyers ultimately possess the capacity to pay for the chips, such as through IPO financing, the suppliers' position will be highly favorable; otherwise, the situation differs. Currently, OpenAI has secretly filed for an initial public offering and is accelerating the expansion of its computing infrastructure to cope with competition.
The risk is not limited to a single enterprise, as an increasing number of companies are betting on the continuous expansion of AI infrastructure and relying on related investments to support their own profitability. If too many companies expect data centers to contribute to earnings, yet the market decides it is no longer willing to pay for more data centers, or if these companies fail to collect payments, the industry could face a scenario similar to that of two decades ago. Even if suppliers possess strong balance sheets, this may not always protect them from cascading shocks when clients over-expand. From a historical perspective, the act of providing guarantees remains risky.
