The current artificial intelligence (AI) boom is largely financed by a "borrowing binge" from tech hyperscalers, leading to hundreds of billions of dollars in debt. This massive influx of capital into the AI sector has prompted some investors to question whether this increased borrowing is contributing to higher US Treasury yields, a trend observed in recent times. The financial underpinning of the AI infrastructure buildout is becoming increasingly scrutinized, especially after the unraveling of the AI-focused hedge fund Situational Awareness.
Nvidia, a key player in the AI industry, is actively partnering with Wall Street firms to raise over $500 billion in third-party capital for AI infrastructure, and is even considering using infrastructure as an investable asset. Hyperscalers, including tech giants like Microsoft, Amazon, Alphabet, Meta Platforms, and Oracle, are leveraging bond markets, joint ventures, and leases to fund this massive expansion. Goldman Sachs analysts estimate that hyperscalers have combined lease commitments of $1.5 trillion for data centers, R&D facilities, offices, and equipment, a significant jump from about $200 billion five years ago. This includes about $1 trillion in "uncommenced" lease commitments, which are not yet reflected in financial statements but represent future payment obligations, potentially understating current leverage and future liquidity needs.
The collapse of Situational Awareness, an AI-focused hedge fund, due to losses on leveraged equity bets, has heightened concerns about the amount of borrowing, its visibility, and potential unwind risks. JPMorgan CEO Jamie Dimon noted that high margin debt increases market volatility. While some strategists, like Sahil Mahtani of Ninety One, believe elevated earnings expectations are a more immediate concern than leverage, the sheer scale of investment is undeniable. Lotfi Karoui, a multi-asset credit strategist at PIMCO, compared the AI capital expenditure cycle to the largest investment cycle since 19th-century railway construction, projecting hyperscaler capital spending alone to exceed $1 trillion per year from 2027 onward. The cost of this buildout is also leading to inflationary pressures, particularly in sectors like electricity, chips, and software, complicating the Federal Reserve's efforts to manage inflation. For instance, electricity prices rose 10% in two years, and the cost of certain computer components like DRAM is expected to increase by 400% by the end of the year compared to 2024, with software and accessories rising 22.4% since July 2024.
