The $570 Billion AI Debt Wave: Hyperscalers Flood Bond Markets

The bank expects the pace of new bond sales to accelerate through the second half of the year . This follows a 2025 in which hyperscaler debt issuance already exceeded $100 billion, and Morgan Stanley's own investment management arm previously projected net debt supply from this group would rise another 30% to 50% in 2026, potentially reaching $130-$150 billion .

Amazon Sets a Canadian Record, Then Secures a $17.5 Billion Loan

No single company illustrates the scale of this borrowing better than Amazon. In back-to-back moves this month, the company demonstrated how aggressively it is tapping global funding sources.

On June 8-9, 2026, Amazon completed a C$14 billion (~$10 billion) investment-grade bond offering in Canadian dollars—the largest corporate bond issuance in the currency's history . The deal shattered the previous record of C$8.5 billion set by Alphabet just one month earlier . The five-part offering, with maturities ranging from three to 30 years, drew massive demand with orders exceeding C$28 billion .

Immediately following the bond sale, Amazon announced a $17.5 billion delayed-draw term loan facility from a consortium of banks including Citigroup, JPMorgan Chase, Wells Fargo, HSBC, and BofA Securities . Unlike a standard bond, this facility gives Amazon the flexibility to draw funds on its own timeline through September 30, 2026, with a three-year repayment window on each draw .

Amazon has guided that it plans to spend $200 billion on AI capital expenditures in 2026, a figure that analysts expect will push its free cash flow deep into negative territory . The company has now reportedly borrowed more than $70 billion across multiple currencies since the start of 2025 .

Oracle: The Cautionary Tale Trading Like Junk

While Amazon comfortably raises record sums, Oracle's experience highlights the growing credit risk embedded in the AI borrowing wave. Despite maintaining an official investment-grade rating, Oracle's bonds have been trading at spreads typical of junk-rated debt since late 2025 .

The numbers are stark. Credit spreads on Oracle's bonds more than doubled from roughly 50 basis points in September 2025 to more than 130 basis points by December 2025, levels not seen since the 2009 financial crisis . The yield on the company's 5.2% notes due in 2035 climbed to 5.9%, surpassing the average yield for upper-tier junk bonds . Oracle's credit default swaps hit their highest level since the Great Financial Crisis, becoming what market observers called a "barometer for AI risk" .

The sell-off was triggered by concerns over Oracle's escalating debt load—which reached $108.1 billion by the end of Q2 FY2026—and investor skepticism about whether its AI investments would generate adequate returns . When Oracle reported fiscal Q2 capital expenditures of $12 billion (well above the expected $8.25 billion) and raised its full-year FY2026 capex guidance by $15 billion to $50 billion, bondholders who had purchased $18 billion in notes just three months earlier saw paper losses of approximately $1.35 billion .

Despite the market turmoil, credit rating agencies have not yet downgraded Oracle to junk status. UBS strategist Matthew Mish said in January 2026 that a downgrade to junk was unlikely in the near term, though a one-notch cut to the lowest investment-grade tier was possible . Fitch affirmed Oracle's BBB rating with a stable outlook in February, and Oracle's co-CEO has publicly emphasized the company's commitment to maintaining investment-grade status .

Investor Caution Grows Even as Issuance Accelerates

The disconnect between Oracle's official ratings and its market pricing reflects a broader wariness creeping into the credit market. Morgan Stanley's report arrives at a moment when some major asset managers are growing more cautious about AI-linked debt.

Fidelity has emerged as a prominent voice of restraint. A late-2025 Fidelity Australia piece noted the firm remained "broadly cautious to downside risks associated with AI capex market euphoria" while still selectively participating in well-priced offerings from fundamentally sound issuers . By mid-2026, however, that caution had deepened: one report indicated Fidelity was actively avoiding bonds from major AI-spending tech firms including Amazon, Alphabet, Meta, Microsoft, and Oracle, citing a flood of new supply and minimal yield pickup .

Salman Ahmed, Fidelity International's global head of macro and strategic asset allocation, had earlier taken a short position against investment-grade corporate debt, arguing that pricing was too rich and an economic downturn could trigger a "proper blowout" . In a Q2 2026 publication, the firm noted that credit spreads "leave a limited margin for error" . A Fidelity Canada piece separately flagged AI as a structural concern for private credit markets, warning that the technology could render traditional software companies obsolete and expose both private credit and broadly syndicated loan markets with their heavy sector exposure .

This caution hasn't stopped the borrowing frenzy. The scale of AI capex is simply too vast for even the world's most cash-rich companies to fund from operations alone. With Morgan Stanley projecting the pace to accelerate and hyperscaler capital spending potentially crossing $1 trillion as soon as 2027, the global bond market's role in financing the AI revolution is only beginning .

The market is now actively pricing in what Mohamed El-Erian described as "disruption risk"—the possibility that the ultimate beneficiaries of AI may not be the companies currently borrowing hundreds of billions to build it. When investment-grade paper trades with high-yield spreads, he noted, it signals that the sector's cost of capital is detaching from current fundamentals and latching onto future uncertainty .