What explains the nearly $570 billion wave of global AI-related debt issuance expected in 2026, and why have credit markets assigned no risk premium to it despite growing concentration risk and hidden off-balance-sheet liabilities?

The premise that credit markets have assigned no risk premium is increasingly outdated. Here's what's actually happening:

Where spreads remain tight:

• The core reason spreads haven't blown out wider is that the debt is being issued by the highest-credit-quality names in the world — Microsoft, Alphabet, Amazon, Meta — all rated AA or A with fortress balance sheets by traditional metrics. Investors view these as "widow-and-orphan" credits where default risk over 5–10 years is near zero, even if AI returns disappoint . Demand has been strong, particularly for dollar-denominated issuance, though banks note the market may be reaching its absorption limit in some currencies .

Where risk premiums are emerging:

• Credit default swap (CDS) spreads for major AI-linked tech companies have risen sharply through mid-2026, particularly for those with weaker balance sheets .
• Bond coverage ratios for hyperscaler debt are declining, suggesting softening investor demand and the potential for higher yields ahead .
• Goldman Sachs and others note "demand fatigue" in the U.S. dollar market, with euro-denominated Amazon and Alphabet bonds outperforming their U.S. equivalents — a tell that dollar investors are becoming more selective .
• Fitch Ratings has formally warned that an AI market correction could become a meaningful risk to corporate credit markets .

The off-balance-sheet "shadow borrowing" problem

The most concerning structural feature is the explosion of hidden liabilities:

• A Nikkei Asia investigation found that five leading hyperscalers (Alphabet, Microsoft, Amazon, Meta, Oracle) have kept more than $1.65 trillion in debt off their balance sheets, compared with roughly $1.35 trillion in officially disclosed liabilities .
• The Bank for International Settlements has labeled this fast-growing pool of off-balance-sheet private credit "shadow borrowing," tying hyperscalers to nonbank lenders and insurers in ways that were largely absent before 2024 .
• Goldman Sachs estimates that hyperscalers' combined lease commitments for data centers, R&D facilities, and offices total roughly $800 billion, much of it structured through joint ventures and special-purpose vehicles that delay balance-sheet recognition .
• These structures are perfectly legal under SEC rules, but they mean public balance sheets no longer tell the full story of who bears the risk if AI returns are delayed or fail to materialize .

The bottom line

The $570 billion wave reflects a rational response to AI's insatiable capital demands from the world's most creditworthy borrowers. But the notion that markets have assigned "no risk premium" is false by mid-2026 — CDS spreads are rising, demand is softening, and Fitch, the BIS, and the Fed have all flagged systemic concerns. The real risk that remains under-priced is the $1.65 trillion in off-balance-sheet liabilities and the opaque web of private credit, leases, and joint ventures that could turn an AI correction into a credit event that hits insurers, private credit funds, and bond investors in ways that traditional balance-sheet analysis can't capture.