Nvidia is collaborating with prominent Wall Street firms to spearhead the mobilization of over $500 billion in third-party capital aimed at enhancing AI infrastructure. As this financing landscape grows increasingly intricate, analysts are scrutinizing the mechanisms driving the current growth surge. Tech giants and their financial allies are exploring various funding methods, including bond markets, joint ventures, leases, and more, to support an expansive infrastructure development effort. Concurrently, the rising leverage is amplifying investors' exposure to AI, with hedge funds and other investment entities utilizing prime brokerage borrowing and derivatives to maximize their returns as they capitalize on this boom. This heightened borrowing amid AI-focused strategies is drawing attention, especially following the collapse of hedge fund Situational Awareness, which faced significant losses due to its leveraged equity strategies. This incident has raised alarms over the extent of borrowed capital, its visibility, and the potential for rapid unwinding, as experts debate whether the generated revenues can justify the recent high expenditures.
The scale of investment in AI infrastructure is noteworthy. Nvidia’s collaboration with firms like Apollo, Blackstone, BlackRock, Brookfield, KKR, and Goldman Sachs may involve developing private asset structures and asset-based financing solutions. Jensen Huang, Nvidia’s CEO, indicated on CNBC that the company's chips have now evolved into what he calls "investable infrastructure assets." In a similar vein, several major tech companies are leveraging joint ventures and leasing strategies to finance their AI data center developments, delaying the appearance of this debt on their financial statements until leasing contracts take effect. According to Goldman Sachs analysts, hyperscalers now hold lease commitments amounting to $1.5 trillion for data centers, research and development facilities, offices, and equipment, a steep increase from around $200 billion just five years ago. This total includes roughly $1 trillion of “uncommenced” leases yet to be disclosed in financial reports, which will necessitate future payments. Analysts from Goldman caution that this may understate both leverage and future liquidity needs as these obligations are acknowledged and contractual payments come due. The recent uptick in debt issuance, alongside less apparent leverage forms, intensifies the examination of whether the returns from AI infrastructure justifies the tremendous amounts being invested. Lotfi Karoui, a multi-asset credit strategist at PIMCO, noted in August that the current cycle of AI capital expenditure, adjusted for inflation, is likely to be the most significant investment wave since the 19th-century railway era. However, he emphasized that the extent of this infrastructure expansion remains "deeply uncertain," with projections suggesting that capital outlays from hyperscalers could exceed $1 trillion annually from 2027 onward, showing no signs of slowing down.
The scale of debt accumulation among hyperscalers is significant enough that they are diversifying beyond dollar-denominated issues, venturing into euro, sterling, yen, Swiss franc, and Canadian dollar markets. Karoui pointed out that the superior performance of euro-denominated bond spreads from companies like Amazon and Alphabet compared to their U.S. equivalents might indicate "demand fatigue" in the dollar market. This impending wave of AI-related bond issuance in the U.S. could lead to higher spreads due to the underperformance of a select few larger AI-exposed issuers.
The downfall of Situational Awareness has illuminated the fragility of concentrated, leveraged AI investments during sharp market shifts. The fund struggled to address several margin calls after its focused portfolio, which included assets like SK Hynix and CoreWeave, suffered significant declines, with its value plummeting from $45 billion to approximately $10 billion. Citadel, a larger multi-strategy hedge fund led by Ken Griffin, subsequently acquired Situational Awareness's publicly traded positions at discounted rates. Following this, both SK Hynix and CoreWeave have seen recoveries. Jamie Dimon, CEO of JPMorgan, recently expressed on CNBC that margin debt levels are "pretty high," noting that this contributes to an increased risk of volatility. Meanwhile, Sahil Mahtani, director of the investment institute at Ninety One, highlighted that the immediate concern lies more with elevated earnings expectations than with the existing leverage levels.
