In Brief: Let There Be Compute!

AI infrastructure will become a durable credit opportunity only if investors focus on financing structure, concentration and underwriting discipline rather than making a binary call on the theme.

• Broad credit indices appear stable, but dispersion is rising: software loan spreads widened 338 basis points in 2026 versus 41 basis points for the broader loan index, while the Federal Reserve raised rates by 25 basis points in September and Treasury yields reached multi-year highs.
• AI infrastructure capex through 2030 is estimated at approximately $7.6-8 trillion; KKR estimates public investment-grade bond markets could provide about $1.7 trillion if the largest hyperscalers each reached a 3% index weight, leaving the balance to internal cash flow and wider financing markets.
• AI-linked debt already totals approximately $600 billion, or 6.3% of U.S. investment-grade debt, after issuance reached about $189 billion in 2025 and more than doubled year over year; the financing mix is expanding into private investment grade, asset-based finance, infrastructure debt and structured solutions.
• Diversification is less robust than it appears: 31 issuers account for more than half a trillion dollars of bonds across the hyperscaler and AI complex, five names account for more than half of that amount, and roughly 40% of the S&P 500 is now AI-linked economically.
• Equity volatility can quickly spill into credit: about $96 billion of single-stock leveraged ETF exposure is concentrated in AI, while the JPMorgan U.S. high-yield AI basket moved from 44 basis points tighter than the broad index in late May to 116 basis points wider by the end of August, a 160-basis-point relative move in thirteen weeks.

Takeaway: Investors should select AI-linked credit by dependency, collateral, contractual revenue, structural protection, liquidity and relative value, favouring the stronger parts of the capital structure, pacing exposure and keeping dry powder rather than pursuing an all-or-nothing thematic position.