AI

Big Tech's AI Debt Binge Shows Warning Signs, Morgan Stanley Says

Amazon's surprise $25 billion bond offering met tepid demand as spreads widen across tech giants borrowing heavily for AI infrastructure.

Omega Editorial· July 8, 2026· 3 min read

Corporate debt markets are flashing warning signals about Big Tech's aggressive borrowing to fund artificial intelligence infrastructure, according to a senior Morgan Stanley portfolio manager.

Vishal Khanduja, who leads broad markets fixed income at Morgan Stanley, told Bloomberg TV this week that credit risk among major technology companies appears undervalued as they raise billions through bond issuances. His concerns center on the possibility that these firms are taking on excessive debt relative to investor appetite.

The alarm bells grew louder when Amazon announced plans to sell $25 billion in bonds—a move Khanduja called a "surprise" since many market participants believed the company had finished its borrowing for 2026. More troubling, investor demand for this latest offering proved notably weak, according to the report from Business Insider.

Spreads widen as supply surges

Beyond tepid demand, bond spreads for Big Tech companies have widened in recent weeks. These spreads—the additional yield investors require above benchmark Treasury rates—signal growing concern about the volume of debt flooding the market.

"It's supply, overall," Khanduja said, identifying the sheer quantity of new issuances as the core issue.

The numbers support his assessment. During the first five months of 2026, four major players in the AI infrastructure race—Amazon, Meta, Alphabet, and Oracle—collectively issued $159 billion in debt, according to analysis from the Kobeissi Letter. By the end of May, Big Tech represented more than 8% of the entire US corporate bond market, a record concentration per Bank of America data.

The $7 trillion question

The trajectory appears unsustainable to some observers. SemiAnalysis, an independent research publication, projects the AI-linked debt market could balloon to $7 trillion by 2029. If that estimate proves accurate, JPMorgan strategists noted this week, AI-related debt would become the second-largest asset-backed securities category in the United States—trailing only mortgage bonds.

That comparison carries weight. The mortgage-backed securities market's explosive growth preceded the 2008 financial crisis, though the underlying risk profiles differ substantially.

Why it matters

The scale of Big Tech's AI borrowing represents a fundamental bet that massive capital expenditures today will generate sufficient returns tomorrow. If investor appetite wanes while companies continue issuing debt to fund data centers, chip purchases, and AI model training, borrowing costs could rise sharply—potentially forcing firms to scale back infrastructure plans or accept lower margins. For enterprise technology buyers, any slowdown in Big Tech's AI buildout could delay access to more powerful models and services. The widening spreads and weak demand suggest the market is beginning to question whether current spending levels are justified by realistic revenue projections.

These details were first reported by Business Insider.

#corporate debt#big tech#ai infrastructure#bond markets#capital expenditure#morgan stanley

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in AI

AI· 3 min read

AI Data Center Spending May Require $2.5-10 Trillion in Revenue

Two independent analyses suggest the industry's capital expenditure dramatically outpaces any plausible near-term monetization path.

Via AI Watch · Aug 22, 2026
AI· 3 min read

How Uber's Algorithm-Based Pricing Drove Fares Up 83% in Four Years

The ride-hailing giant replaced predictable meter-style rates with real-time algorithmic pricing that critics say maximizes revenue extraction.

Via AI Watch · Aug 21, 2026
AI· 3 min read

AI Agent Performance Depends More on Harness Than Model Choice

Nvidia research shows custom scaffolding and supervisor components can boost benchmark scores from 30% to 100% using the same underlying model.

Via AI Watch · Aug 21, 2026