Last week’s market crash, driven by heavy selling in AI and data center stocks, triggered fears that the ‘AI super-cycle’ was over. However, this article argues the plunge was a misinterpretation, not a fundamental flaw, largely caused by hedge fund position unwinds rather than a broken paradigm. The author asserts Nvidia’s role in funding AI startups like CoreWeave is legitimate, contrasting it with past ‘vendor finance’ failures, emphasizing the enduring value of Nvidia’s chips and the long-term potential of a $10-$12 trillion AI build-out.
Last week’s market performance was nothing short of a nightmare for many, marked by significant drops across the tech sector, particularly in semiconductors. The rapid decline, culminating in Friday’s devastation, led many to believe that the monumental data center and AI “super-cycle” – a term often associated with eventual collapse – was finally over. The prevailing sentiment suggested a retreat to safer, more traditional stocks like Colgate-Palmolive, Coca-Cola, Procter & Gamble, and Kimberly-Clark, abandoning the promise of burgeoning AI factories.
The speed of this market reversal prompted questions about its cause and the whereabouts of buyers. My analysis suggests this wasn't a fundamental flaw in the AI thesis, but rather a disproportionate reaction. The selling, especially on Friday, concentrated heavily on the components driving the data center/AI narrative, which, while representing a substantial portion of the market’s capitalization (around 25%), is not its entire absolute percentage.
I believe the recent market turmoil stemmed from an unwind of positions by certain firms, potentially coupled with the collapse of a lesser-known hedge fund that had borrowed heavily. This aligns with the argument, championed by my friend Herb Greenberg, that this marks the end of a 'debt-fueled hyperscale-hoax' – a 'trillion-dollar con' orchestrated by Nvidia CEO Jensen Huang, who, according to this view, encouraged spending on GPUs to fuel the cycle. The narrative suggests that CoreWeave’s corporate bond yields spiking signaled the bond market squeezing out marginal buyers.
However, it's premature to judge Jensen Huang. We live with two competing investment narratives that demand critical analysis. Jensen Huang, a long-time Club name, foresaw the potential of accelerated computing as the catalyst for instant pattern recognition, enabling machines to 'think.' His collaboration with OpenAI’s Sam Altman highlighted the profit potential for those needing artificial intelligence. While many large tech companies focused on cloud infrastructure, only a few, like OpenAI, Anthropic, and to some extent Alphabet’s Google Gemini, have truly capitalized on AI at scale. Nvidia’s pivotal role in seeding these ventures allowed generative AI and agentic AI to flourish, driven by Huang's desire for Nvidia, its shareholders, and the U.S. to lead.
Critics often liken Jensen’s funding activities to Lucent’s 'vendor finance' model from the turn of the century, where the telecom equipment company financed its clients, masking underlying weaknesses until the business became unsustainable without it. This 'Lazy Susan' analogy suggests money is merely cycled back to Nvidia for product purchases. Having observed such dynamics in the past, my initial instinct was to expose what seemed like a dubious scheme.
But today is different. The vendor finance of the early 2000s, like Lucent’s, provided products that the market ultimately didn’t need. The core of the current argument against Nvidia is a similar accusation: that it's merely a modern Lucent, and Jensen Huang a charlatan akin to Rich McGinn, who artificially boosted sales with fictitious clients. If you subscribe to this 'Lucent paradigm,' then the entire AI infrastructure is overvalued or worthless. It's an elegant, almost too-perfect narrative for short-sellers.
The build-out, significantly propelled by Michael Intrator, CoreWeave’s CEO (a former bitcoin miner), involved billions from Nvidia in exchange for vast chip purchases to erect data centers for Microsoft and other hyperscalers. Bears contend this circular arrangement required CoreWeave to continuously con public debt investors. This 'poetry in motion' supposedly stopped when CoreWeave’s debt yields surged past 15%, making further builds economically unviable and, by extension, halting Nvidia orders – signaling 'game over.' This storyline persisted last week, even wiping out Thursday’s bounce.
Yet, I offer an alternative. Thursday’s bounce coincided with Jensen Huang’s simple message regarding CoreWeave: no intervention needed; things were better than anticipated. This was a brilliant move, avoiding any perception that Nvidia was 'backstopping' like a central bank. Nvidia possesses what neither Jim Jones nor the current bears have: credibility. While the public debt market for CoreWeave might be paused, Jensen knows Nvidia’s chips retain their value for up to 10 years, lending against them for five. If prices hold, he can resell or lease them, a fact largely ignored by critics. There will soon be a two-way market for these compute assets, similar to mortgage-backed bonds, a reality the loan market already acknowledges, allowing Intrator to borrow at lower rates than public markets offer.
I believe the public markets are misjudging CoreWeave, which is a linchpin in the build-out thesis. It's not a sham. CoreWeave isn't Lucent, nor is Nvidia Nortel. This is merely a skirmish in a larger war for data centers. The broader battle might materialize during midterm elections if data centers become politically unpopular. Last week’s market action seemed to suggest market exhaustion and political headwinds for data centers; I contend neither is true. The market is pausing, and many localities will still welcome data centers. The ultimate build-out could reach $10-$12 trillion, as projected by Morgan Stanley CEO Ted Pick. I anticipate both Anthropic and OpenAI will go public, and states that ban data centers will regret missing out on the coming influx of capital.
My optimistic view has many adherents, perhaps too many, some of whom overleveraged their bets without fully grasping the complex dynamics. These 'broken hedge funds,' not 'broken paradigms,' are the problem. If I'm right, the market clears quickly, revealing the larger hedge funds whose positions were liquidated. CoreWeave’s debt prices will stabilize, and even Oracle’s stock could rise. While bears will seek to make their year, remember the longevity of Nvidia’s chips, the imminent public market for compute (driven by BlackRock), and the inevitable triumph of generative AI. History will note those who either embraced the fourth industrial revolution too aggressively or lacked sufficient capital to endure the journey.
