AI Infrastructure Debt Nears $570B as Bond Demand Softens

The infrastructure boom powering artificial intelligence is leaving a significant mark on credit markets, with global AI-related debt issuance projected to reach $570 billion in 2026, according to Morgan Stanley forecasts first reported by Forbes contributor Robert J. Szczerba.

By the end of May, approximately $236 billion had already been issued—four times the pace from the previous year. The financing wave extends beyond major cloud providers to encompass a broad range of AI-linked issuers and structures, from data center developers to utilities expanding capacity to meet surging power demands.

Why it matters

The shift toward credit-financed AI infrastructure isn't a cash crisis, but it does redistribute risk across the financial system in ways that are increasingly difficult to track. As more financing moves into private credit and off-balance-sheet vehicles, executives face growing exposure through cloud vendors, utility partners, and data center operators—even if they never directly purchase AI-related securities. Understanding where this risk sits matters for strategic planning and counterparty assessment.

Bond market signals early strain

A key indicator of changing market dynamics comes from coverage ratios—the measure of investor demand relative to bonds offered. Apollo reports that orders for hyperscaler bonds covered nearly five times the amount offered in February. By July, that coverage had fallen below two times.

While not indicative of a market breakdown, the declining ratio suggests bond investors may soon demand higher yields to absorb the increasing supply. Deal composition could account for some variation, but the directional trend warrants attention as an early signal of shifting investor appetite.

Strategic borrowing, not financial distress

A common misreading of the data suggests major technology companies are running out of cash. The reality is more nuanced. Goldman Sachs analysts note that consensus estimates place hyperscaler capital expenditure near 100% of operating cash flow—a heavy commitment, but not a liquidity crisis.

Morgan Stanley characterizes these companies' starting financial position as exceptionally strong. The decision to borrow reflects capital allocation strategy: preserving flexibility for share buybacks, acquisitions, and maintaining liquidity buffers while continuing aggressive infrastructure investment.

The frequently cited $1.5 trillion figure represents a funding gap in Morgan Stanley's analysis—the difference between projected global data center investment through 2028 and what Big Tech can fund from cash flow. This gap can be filled through various instruments: bonds, private credit, asset-backed securities, mortgages, or equity. Morgan Stanley assigns roughly $200 billion to related corporate debt issuance through 2028.

Risk migrates to less transparent structures

Public bond markets remain relatively stable, with J.P. Morgan noting that spreads sit near cycle lows. These markets offer transparency and established pricing mechanisms.

The concern lies elsewhere. Morgan Stanley projects an approximately $800 billion private credit opportunity in data center financing through 2028. Much of this financing can reside in off-balance-sheet vehicles that the Bank for International Settlements describes as shadow borrowing.

Meta's $27 billion Hyperion joint venture illustrates the structure: Blue Owl-managed funds own 80%, Meta holds 20%, and some of Blue Owl's funding comes from debt sold to PIMCO and other investors. The arrangement keeps substantial project debt off Meta's public balance sheet.

Utilities are also borrowing heavily to serve data center demand, with U.S. investment-grade utility issuance around $135 billion in 2025 and projected to reach $145 billion in 2026. Disclosure standards and creditor protections vary widely across these financing structures, making potential losses harder to identify if utilization or pricing falls short of projections.

The indirect exposure question

For executives outside the hyperscaler ecosystem, the relevant question isn't whether AI will generate sufficient returns to justify the spending. It's who finances the timeline to profitability, and who absorbs the impact if returns arrive late or disappoint.

These dependencies run through cloud service agreements, utility partnerships, and data center capacity contracts. Exposure may surface as higher prices, capacity constraints, or counterparty risk—even for organizations that never directly invest in AI infrastructure.

The details were first reported by Robert J. Szczerba in Forbes.