AI Infrastructure Debt Surges Past $1.5 Trillion

The hidden scale of AI infrastructure financing

Tech companies and their financial partners are deploying unprecedented amounts of capital to build AI infrastructure, but much of the debt fueling this expansion remains obscured from traditional financial statements.

Nvidia has announced partnerships with Apollo, Blackstone, BlackRock, Brookfield, KKR, and Goldman Sachs to mobilize more than $500 billion in third-party capital for AI infrastructure development. CEO Jensen Huang characterized Nvidia's chips as an "investable infrastructure asset," signaling a shift toward treating computing hardware like traditional infrastructure investments.

According to Goldman Sachs analysts, hyperscalers now carry combined lease commitments totaling $1.5 trillion for data centers, research facilities, offices, and equipment—up from roughly $200 billion five years ago. Approximately $1 trillion of these commitments are classified as "uncommenced" leases, meaning they don't yet appear in financial statements but will generate future payment obligations.

Goldman's analysts warned this accounting treatment "can understate leverage and future liquidity needs as these obligations are eventually recognized and contractual payments come due," as first reported by CNBC.

Why it matters

The scale and opacity of AI infrastructure financing creates systemic questions about whether returns will justify the investment—and whether markets can accurately price the risk. With consensus forecasts projecting hyperscaler capital spending alone will exceed $1 trillion annually from 2027 onward, the buildout represents what PIMCO strategist Lotfi Karoui called the largest investment cycle since 19th-century railway construction, adjusted for inflation.

Leverage amplifies volatility

The recent collapse of AI-focused hedge fund Situational Awareness has intensified scrutiny of leverage across the ecosystem. The fund's assets plummeted from $45 billion to approximately $10 billion after it couldn't meet margin calls during a tech sector sell-off. Its concentrated portfolio included positions in SK Hynix and CoreWeave. Citadel later purchased the fund's publicly listed positions at a discount.

JPMorgan CEO Jamie Dimon told CNBC that margin debt levels are "pretty high," increasing the risk of amplified volatility. Hedge funds and other investors are using prime brokerage borrowing and derivatives to magnify their exposure to AI stocks.

Hyperscalers themselves are diversifying their debt issuance beyond U.S. dollar-denominated bonds, tapping euro, sterling, yen, Swiss franc, and Canadian dollar markets. Karoui noted that euro-denominated bonds from Amazon and Alphabet have outperformed their U.S. counterparts, potentially signaling "demand fatigue" in the dollar market.

Concentration risk versus leverage risk

Sahil Mahtani, director of the investment institute at Ninety One, argued that elevated earnings expectations pose a more immediate concern than leverage itself. "That is primarily an expectations problem rather than a leverage problem," he said.

Mahtani pointed to historically high equity concentration in tech-heavy markets, particularly in the U.S., which can amplify market moves when heavily weighted stocks decline—functioning like leverage even without borrowed capital.

The Alternative Investment Management Association maintained that available evidence doesn't support treating hedge fund leverage as an inherent systemic risk, distinguishing the Situational Awareness episode from previous leverage-related crises like Archegos Capital Management's 2021 collapse.

The details were first reported by CNBC.