At the epicenter is Nvidia, the semiconductor behemoth whose market cap has soared to $5 trillion on the back of its AI GPUs. Nvidia doesn't just sell chips; it invests heavily in its customers to ensure future purchases, exemplifying circular accounting. Take its proposed $100 billion investment in OpenAI: In exchange for funding massive data centers (at least 10 gigawatts), OpenAI commits to deploying millions of Nvidia GPUs. This isn't organic sales growth—it's a premeditated loop where Nvidia's outlay returns as guaranteed revenue. Nvidia CEO Jensen Huang unabashedly touts this as "one of the smartest investments we can possibly imagine," predicting OpenAI could become a "multi-trillion-dollar hyperscale company." But this optimism hinges on the circle: OpenAI's expansion drives more chip buys, boosting Nvidia's stock, which in turn funds more such deals.
The interdependencies extend to hyperscalers and startups, forming a tangled web of mutual back-scratching. OpenAI, despite hemorrhaging $11.5 billion in losses last quarter, secured a $300 billion cloud deal with Oracle for 4.5 gigawatts of capacity by 2027. Oracle, in turn, must purchase $40 billion worth of Nvidia's GB200 chips—400,000 units—to fulfill it. This announcement alone spiked Oracle's stock by 36%, adding $88 billion to founder Larry Ellison's net worth, while Nvidia's shares jumped 4%, tacking on $170 billion to its market cap in a single day.
Meanwhile, OpenAI's pivot to AMD for 6 gigawatts of Instinct GPUs includes a warrant for 160 million AMD shares (a 10% stake), vesting as deployments roll out. AMD executives hail this as delivering "tens of billions" in revenue, but it's classic circularity: OpenAI gets discounted hardware, AMD locks in sales, and the warrant ties their fortunes together, inflating valuations through anticipated synergies rather than proven earnings.
CoreWeave, another AI infrastructure player, epitomizes the loop. Nvidia holds a 5% equity stake and pre-purchased $6.3 billion in cloud services from CoreWeave as a "safety net." This guarantees CoreWeave can afford more Nvidia chips, recycling funds back to Nvidia while propping up CoreWeave's expansion. Microsoft, with its 49% stake in OpenAI from $11 billion in investments, provides Azure cloud but tolerates OpenAI's diversification to Oracle and others—yet this too feeds the circle, as OpenAI's growth necessitates more Microsoft-purchased GPUs. Even government involvement via the CHIPS Act adds a layer: The U.S. took a 9.9% stake in Intel ($8.9 billion) and granted $2.2 billion for domestic fabs, indirectly fueling the AI chip demand without scrutinizing the circular flows.
This circular accounting distorts financial realities, turning potential losses into paper gains. Revenues aren't from broad market adoption but from pre-committed deals among insiders, creating "instantaneous value" that evaporates if one link breaks. OpenAI's non-profit shell running a for-profit arm raises red flags about governance, while the sector's "getting fatter by eating itself" suggests unsustainable bloat. Stock prices reflect hype over fundamentals: Nvidia's valuation assumes perpetual dominance, but rivals like AMD are drawn into the same vortex, their shares buoyed by these pacts rather than independent merit.
Historically, this mirrors the dot-com era's ad-buying frenzy, where companies propped each other up with no viable products. The Bank of England detects a "hint of dotcom bubble 2.0 in AI froth," and experts warn that intertwined balance sheets amplify systemic risk.
Unlike then, AI involves real assets like data centers, but the circularity means overcapacity could lead to a glut, crashing prices. With AI investments staving off U.S. recession, a bust could be catastrophic, setting the industry back decades. Investors chasing multi-trillion valuations ignore the fragility: When the loop unravels—due to regulatory scrutiny, demand shortfalls, or a single default—the bubble bursts, revealing the emperor's new clothes. This isn't innovation; it's a high-stakes Ponzi scheme masquerading as progress, and history suggests the fallout will be severe.
We should learn from the "broadband bubble pop" when the telecommunications crash of the early 2000s, also a period of massive investment, overbuilding of infrastructure, and subsequent market collapse coincided with the broader dot-com bubble burst. This time isn’t different.
This opinion is not stock market advice to buy or sell. Always consult a trusted professional advisor before making a speculative investment base on your risk profile. Editorial comments expressed in this column are the sole opinion of the writer.
