Last week’s market downturn, particularly the sharp sell-off in AI and semiconductor stocks, is seen by the author as an overreaction driven by positional unwinds rather than a fundamental collapse of the AI narrative. The author draws parallels to past “vendor finance” schemes but argues that Nvidia’s sustained technological advantage and the intrinsic value of its chips differentiate the current situation.
Despite concerns raised by CoreWeave’s debt yields, the author maintains that the AI supercycle is far from over and predicts continued growth in data center infrastructure.
READ MORE FROM CNBC
The market experienced an "obliteration" last week, with significant drops in major indices and particularly in the semiconductor sector. The author argues that the market's reaction, especially on Friday, was an overreaction to the perceived end of the "data center story" or AI supercycle, a narrative that has driven a substantial portion of market capitalization.
The article posits that last week's sell-off was not a reflection of the cause but rather an effect of positional unwinds by certain firms and the collapse of undisclosed hedge fund leverage. This is compared to the "vendor finance" schemes of the dot-com era, specifically citing Lucent Technologies, where companies financed clients to mask weak underlying business. The author likens Nvidia CEO Jensen Huang's funding of companies like CoreWeave to this practice, suggesting it's a "debt-fueled hyperscale-hoax."
However, the author then presents a counter-narrative, emphasizing Jensen Huang's credibility and the enduring value of Nvidia's chips. Unlike the dot-com era, where the end product was often unnecessary, Nvidia's GPUs are argued to have lasting utility and value, potentially holding their worth for a decade. This resilience, the author suggests, allows for innovative financing structures, such as Nvidia lending against its own chips, creating a sustainable ecosystem.
The author contrasts this with the situation at CoreWeave, whose debt yields surged to over 15%, indicating a mismatch between spending and return. While bears see this as the end of the AI build-out, the author believes this is a temporary "skirmish" in a larger, ongoing "war for or against the data center." The article concludes that the market is taking a "well-deserved breather" rather than signaling the end of the AI supercycle, which is projected to reach $10-$12 trillion. The author expresses confidence that companies like OpenAI and Anthropic will eventually go public and that the demand for data centers will continue to grow, ultimately proving the bears wrong.
