Is the AI debt bubble about to burst? Goldman Sachs publicly admits for the first time: Market absorption capacity is nearing its limit.

```

The unprecedented AI borrowing frenzy by tech giants is pushing the global investment-grade bond market to its limits. Goldman Sachs credit strategists and traders have issued warnings, admitting that the massive bond supply from mega-cap cloud computing companies has caused severe indigestion in the market, and if this crack widens, it will not just be a problem for the bond market.

Jeffrey Papai, Goldman Sachs investment-grade bond trader, stated bluntly in a report released last week that the $75 billion in AI-related bond issuance over the past month has left the market "struggling to digest," with Goldman Sachs's AI bond basket spreads widening by 22 basis points in a single week. More critically, he points out that the market’s capacity to absorb is rapidly shrinking—where previously over $75 billion of supply would be needed to pressure the market, now just $25 billion can leave the market "on the back foot."

Brian Garrett, Goldman Sachs derivatives trading head, also confirmed that the credit trading desk has described the situation as "carnage," while during the same period, the volatility of the S&P 500 index has fluctuated within a range of only 30 basis points, raising concerns about the divergence between stocks and bonds.

The market impact of this warning is already visible. According to MarketAxess data, last week the spread of Alphabet's 10-year bond widened by 0.12 percentage points, Meta's 10-year bond spread increased by 0.16 percentage points, while Bloomberg data shows the average spread of investment-grade bonds as a whole only rose by 0.02 percentage points in the same period. The mega-cap computing sector is rapidly diverging from the traditional investment-grade market.

Borrowing Scale: From Margin to Systemic Risk

The pace of AI-related bond issuance has far exceeded the market's previous expectations.

According to the Wall Street Journal, Alphabet, Amazon, Meta, Oracle, Nvidia, and SpaceX—six tech companies classified as "mega-cap AI computing providers" by the market—have cumulatively issued about $244 billion in bonds in the global bond market this year, more than double last year's $108 billion, and over 14 times the $17 billion in 2024.

Morgan Stanley data shows that the overall leverage ratio of mega-cap computing providers has soared from 0.9 times in the third quarter of 2025 to the current 1.8 times, doubling in just over two quarters, and has already surpassed the leverage level of the entire energy sector, with the ratio continuing to climb by about 0.3 times per quarter.

In terms of market share, Amazon, Meta, Google, Microsoft, and Oracle currently account for 4% of the U.S. dollar investment-grade bond index. Using duration-weighted (DV01) metrics, AI-related issuers, including the six mega-cap computing companies and three major chip companies, already account for 9% of the index, with the proportion still rising.

Goldman Sachs Warning: The Market Can No Longer Absorb the Next Round of Supply in the Usual Way

Amanda Lyman, Goldman Sachs credit strategy chief, pointed out in a report released last Thursday that if the debt scale of mega-cap computing providers is raised to the level of large U.S. banks, theoretically, only about $510 billion of incremental debt capacity could be released. Meanwhile, Goldman Sachs' equity research team expects that the five big mega-cap computing providers' combined AI capital expenditures for the 2025-2030 fiscal years will reach $5.8 trillion, and this year’s estimated capital expenditures already consume most of these companies’ operating cash flows, making debt financing an indispensable core approach.

In the report, Jeffrey Papai further points out that the proportion of AI-related bond supply in the overall investment-grade market has jumped from about 1% in 2024 to about 7% in 2025, and to about 18% in the first half of 2026. For issuances with maturities over 15 years, the proportion is as high as 42%.

Papai clearly states, "The incremental absorption capacity from investors has severely shrunk," and lists several prerequisites for the market to digest the next round of mega-cap computing capital expenditures: issuing companies must slow the pace and provide clearer forward guidance; change the issuance structure, including currency and maturity diversification to avoid over-concentration in a single investor group; banks’ direct lending, private credit, and alternative asset managers must absorb more share; and bond prices must fall further (i.e., spreads widen further) to attract new investors.

However, Papai also admits that the "price falls further" route itself contains an internal contradiction—if a full-scale sell-off of AI bonds unfolds, it will have a direct impact on the stock market, possibly spreading to the broader investment-grade market.

Cash Flow Gap: The Fundamental Driver of Debt Expansion

Mega-cap computing providers continue large-scale borrowing because their capital expenditures have far exceeded the coverage capacity of their own cash flows.

Apollo chief economist Torsten Slok recently warned, that so far corporate profit margins outside the tech sector have yet to show improvement, and AI companies’ valuations are entirely based on the assumption that "the profit margin of S&P 493 components will eventually see a sharp increase." He noted that if productivity gains from AI take five years rather than five months to materialize, the stock market, which is priced for immediate earnings growth, will face a painful repricing.

Slok further pointed out that token prices continue to slide, China’s models have already surpassed their U.S. counterparts in both usage and market share among the most common global models, putting additional pressure on the free cash flow outlook of mega-cap computing providers. He warned that if the market consensus for AI payback timing proves overly optimistic, three knock-on effects will follow: cash flow and earnings will miss expectations, the "Magnificent 7" stocks will be sold off and drag down the broader market, and mega-cap computing providers’ balance sheets will continue to deteriorate, risking credit rating downgrades.

BofA chief investment officer Michael Hartnett also lists "AI capital expenditure cuts" as the largest tail risk in his latest report, pointing out the critical path: the bond market "vigilantes" cut off liquidity supply to mega-cap computing providers, forcing them to turn to equity financing and cut jobs—Meta, Microsoft, and Amazon have already laid off 13%, 10%, and 9% of their workforce respectively.

Bubble Narrative Evolution: From "Generative AI" to the "Agent AI" Relay

This is not the first time the AI debt market has approached a critical point.

In early May, the Financial Times reported that banks such as JPMorgan, Morgan Stanley and SMBC were actively seeking ways to transfer data center-related debt risk to a broader group of investors, as the core buyers were already showing resistance to the continued influx of AI bond supply. At that time, the AI debt bubble seemed on the verge of collapse.

However, just about 48 hours after that report was published, Goldman Sachs released a research note titled "Decoding the Agent Economy," predicting that Agent AI would drive LLM profit margins sharply higher, providing bondholders with ample principal and interest repayment protection. This narrative shift quickly calmed bond market tensions and spawned a new wave of larger bond issuances—Amazon completed a $37 billion bond issuance in March this year (the fourth largest corporate bond issuance ever), and issued another $25 billion in July; SpaceX, just days after its IPO, entered the bond market with a $25 billion bond issuance.

Now, Goldman Sachs credit trading desk's "carnage" warning marks a weakening of support for this narrative. Papai, in the conclusion of his report, writes that in the short term, the AI bond complex may catch its breath due to some adjustment in spreads and the issuance window approaching the earnings blackout period, but "in the medium to long term, given the supply outlook, it is expected to further underperform the broader market," and suggests investors hold credit volatility positions to hedge against the possibility of "contamination" of the wider market.

Risk Warning and DisclaimerThe market has risks, and investment should be cautious. This article does not constitute personal investment advice, nor does it take into account individual users' specific investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment based on this article is at your own risk. ```