AI money pit: Amazon issues $25 billion in bonds, AI-themed bonds sold off across the board.

Author: Dong Jing, Wall Street Insight

The AI infrastructure arms race is pushing the balance sheets of tech giants to historic limits, triggering alarm bells in the bond market.

Amazon launched a new round of $25 billion USD bond issuance on Tuesday. The news triggered a massive sell-off in AI-related bonds, with spreads on bonds from hyperscalers like Meta and Amazon widening sharply in a single day, far exceeding the typical daily fluctuation range of the investment-grade market.

According to Bloomberg data, if this issuance is completed successfully, it will push global AI-related bond sales this year to a record $335 billion, more than doubling the total for the full year 2025. Meanwhile, Goldman Sachs warned clients over the weekend, advising proactive hedging against the risk of a debt bubble among hyperscalers.

The core concern of the market is: The "bottomless pit" nature of AI capital expenditure is eroding the credit quality of tech giants at an unprecedented pace. According to Morgan Stanley data, the overall leverage ratio of hyperscalers has surged from 0.9x in Q3 2025 to the current 1.8x, doubling in just two quarters, and has already surpassed the leverage level of the entire energy sector, growing at a rate of about 0.3x per quarter.

John Lloyd, Head of Global Multi-Sector Credit at Janus Henderson, pointed out that portfolio holdings of hyperscalers are already quite concentrated, and new bond issuances must offer sufficient price concessions to attract investors to continue adding positions.

Amazon Returns to Bond Market Again, Single Issuance Hits Yearly High

Amazon's bond issuance this time is divided into up to eight maturity tranches, ranging from 3 to 40 years, with the longest being bonds maturing in 2066. The initial price guidance is approximately 145 basis points over comparable U.S. Treasuries. The underwriters are Barclays, Goldman Sachs, JPMorgan Chase, and Morgan Stanley. The proceeds will be used for general corporate purposes, including debt repayment, mergers and acquisitions, and capital expenditures.

This is another massive fundraising by Amazon this year. In March, the company issued $37 billion in USD bonds, setting the fourth-largest single issuance record in U.S. corporate bond history. During the same period, it also issued €14.5 billion in EUR bonds, and subsequently completed issuances in Swiss franc and Canadian dollar markets. According to Bloomberg data, since last year, Amazon has raised approximately $82 billion through multi-currency high-grade bonds, while the company's estimated capital expenditure this year is close to $200 billion, leaving a significant funding gap.

This issuance will complete Amazon's USD funding plan for 2026, with any subsequent USD bond issuance considered opportunistic. However, analysts generally believe that this does not mean Amazon's overall bond issuance is coming to an end.

AI Bond Spreads Widen Sharply in a Single Day, Market Selling Pressure Intensifies

Following the news of Amazon's new bond issuance, the secondary market for AI-related bonds came under pressure.

According to MarketAxess data, the spread on Meta's 6.3% coupon bonds maturing in 2056 widened by about 8 basis points in a single day; Amazon's 5.95% coupon bonds maturing in 2066 widened by about 14 basis points. In comparison, the spread fluctuation for high-grade corporate bonds on a normal trading day is typically only a few basis points.

John Lloyd commented that the widespread weakness in hyperscaler bonds that day mainly reflects investors shifting funds to participate in Amazon's new bond subscription, rather than any substantive change in credit fundamentals. However, he also noted that investors are increasingly questioning how much additional debt the industry will need to issue in the coming years to finance AI infrastructure.

SpaceX's $25 billion bond issuance completed in June was also not spared, with related bonds swept into this sell-off. According to BondCliQ statistics, six major participants in the AI race—Amazon, Microsoft, Alphabet, Meta, Oracle, and Nvidia—have cumulatively issued over $460 billion in bonds in recent years, with approximately $100 billion of that occurring this year.

Echoing the volatility in the bond market, the credit default swap (CDS) market has also shown notable movement. Oracle's CDS spread is rapidly widening, and CDS for Coreweave and SpaceX have shown similar trends. The credit market is beginning to reassess the risk pricing of AI-related names.

Oracle's stock price has fallen nearly 50% from its high on June 2, dropping to its lowest level since 2026, seen as a typical case of high-leverage AI concept stocks under pressure.

Morgan Stanley data shows that the speed at which the overall leverage ratio of hyperscalers is rising has alarmed the market—from 0.9x to 1.8x in just two quarters, with no sign of slowing down.

This trend is highly similar to two episodes of credit tightening that occurred at the end of 2025 and in March 2026, when CDS for tech names surged sharply, temporarily halting the debt-driven wave of capital expenditure.

Goldman Sachs Issues Warning, Advises Clients to Actively Hedge Against Bubble Risk

Facing the accumulation of these risks, Goldman Sachs trader Adam Crook sent a report to clients over the weekend, advising them to proactively position hedging strategies to counter the potential market impact from the continued debt expansion of hyperscalers.

In the report, Crook pointed out a scenario: The stock market continues to hit new all-time highs, while the credit market struggles to rise further from historically tight valuation levels—massive supply pressure will continue to suppress credit spreads, and higher interest rates also pose resistance to refinancing for existing issuers, especially those with lower ratings.

Crook emphasized that if a macro shock occurs—such as significant new information challenging the AI bull market thesis or capital expenditure plans—these parts of the credit market, currently at historically low spread levels, would provide asymmetric returns and convexity for macro or multi-asset portfolios as a risk hedge.

It is worth noting that data from the public bond market may only represent the tip of the iceberg of the AI capital expenditure wave. According to analysis cited by ZeroHedge, in addition to disclosed capital expenditures and public debt, there is an off-balance-sheet exposure as large as $1.8 trillion, posing potential systemic risk concerns.

Currently, spreads on AI-related bonds have been wider than the overall U.S. investment-grade corporate bond market for several months. As hyperscalers' financing needs continue to push the limits, bond investors face increasing uncertainty about how to price this "bottomless pit" of an AI arms race.