The AI revolution is devouring capital at an unprecedented pace, with U.S. secured debt for data centers surging 112 percent to $25.4 billion in 2025, up from $12 billion in 2024, as hyperscalers like Microsoft, Google, and Oracle race to erect GPU-packed facilities amid soaring demand for generative models. This borrowing bonanza, which has ballooned 1,854 percent since 2022, reflects the voracious appetite for infrastructure: AI capex now consumes 94 percent of hyperscalers’ operating cash flow minus dividends and buybacks, up from 76 percent last year, forcing a “borrow-to-build” scramble that taps every debt corner from high-yield junk bonds to private credit markets churning $50 billion quarterly. JPMorgan Chase strategists warn of a $5 trillion global data center spend over five years—potentially $7 trillion in the bull case—requiring $1.5 trillion in investment-grade bonds alone, alongside $150 billion from leveraged finance and up to $40 billion annually in securitizations. Yet with annual depreciation on new builds hitting $40 billion against $15-20 billion in revenues, per industry estimates, this debt-fueled exuberance risks systemic tremors if AI monetization falters.
Hyperscalers’ dominance underscores the scale: global capex rocketed 72 percent year-over-year to $127 billion in Q2 2025, per Synergy Research Group, with generative AI adding $50 billion in quarterly sales to digital services revenue growing 18 percent annually. Oracle’s pivot exemplifies the frenzy: in September, it inked a staggering $300 billion, five-year deal with OpenAI for 4.5 gigawatts of capacity under the “Stargate” initiative, fueling a 359 percent surge in future contract revenue and a 54 percent stock rally. This pact, part of OpenAI’s $500 billion infrastructure push with SoftBank, positions Oracle as a compute kingpin, but it has spiked credit default swaps to five-year highs, reflecting investor hedges against ballooning leverage—hyperscalers’ bond spreads widened to 0.78 percentage points over Treasuries, the broadest since April’s tariff shocks.
Smaller players face steeper perils. Bitcoin miners, squeezed by the 2024 halving that halved block rewards to 3.125 coins and cratered hashprices from $0.12 to under $0.05 per terahash, are pivoting aggressively to AI data centers—repurposing power-hungry rigs for high-performance computing that yields 25 times higher revenue per megawatt. Firms like Core Scientific, Iris Energy, and Hut 8 have raised over $4.6 billion in debt and convertibles since late 2024, with Q3 issuances hitting $6 billion, betting on AI’s stability to offset mining’s volatility where electricity devours 90 percent of costs. A stark example: Core Scientific’s 2022 bankruptcy amid a 70 percent Bitcoin plunge saw lenders seize collateral rigs, a fate echoed in 2025’s $8-11 million per megawatt retrofit costs that strain cash flows and invite customer concentration risks from bursty AI workloads. States like Texas and Georgia dangle tax credits for such shifts, but execution falters: profitability timelines stretch amid power constraints, with miners’ valuations now hinging on hybrid models yet vulnerable to dilution and defaults if AI demand wanes.
Morgan Stanley pegs total data center spending at nearly $3 trillion through 2028, with hyperscalers funding $1.4 trillion from cash flows, leaving the rest to debt that could ripple globally if defaults mount. Menlo Ventures notes only 3 percent of consumers pay for AI services—$12 billion annually—exposing the chasm between capex and returns. Private credit eyes $800 billion of the gap, but hotspots like CoreWeave’s CDS at 505 basis points signal distress.
Investors must navigate nimbly. Limit exposure to 10 percent in smaller pivots like miners, favoring diversified hyperscalers. Track CDS via Bloomberg terminals for Oracle spikes, setting alerts at 100 basis point jumps to flag leverage risks. Diversify bonds across maturities—short-term for liquidity, long-term for yields—and stress-test portfolios against 20 percent AI revenue shortfalls using Monte Carlo simulations. Quarterly audits of debt covenants, per Skadden’s guidance, mitigate off-balance-sheet vehicles’ opacity. Hedge with puts on high-yield ETFs amid junk bond spreads at 400 basis points.
The $25.4 billion debt wave isn’t euphoria—it’s a high-stakes bet on AI’s $46.9 billion crypto-adjacent market by 2034. Smaller players’ risks loom largest, but opportunities abound for the vigilant.
Portfolio managers, the frenzy accelerates. Rebalance toward resilient hyperscalers today, monitor CDS religiously, and audit exposures now. In AI’s debt inferno, prudence forges alpha—act before defaults ignite the next crisis.
