Goldman Sachs Warns AI Debt Bubble Is Straining Bond Markets in 2026

The AI infrastructure build-out of 2026 is no longer just a tech story—it has become a credit market story, and the warnings are coming from the highest levels of finance.

Between Goldman Sachs fixed-income desks, the Bank for International Settlements, and S&P Global Ratings, a consensus has emerged: the scale of debt being taken on by the world's largest technology companies is testing the capacity of the global bond market, creating conditions that several analysts describe as dangerously fragile.

Hyperscaler debt and capital expenditure at record levels

The numbers are staggering. Amazon, Microsoft, Alphabet, and Meta—the four largest hyperscalers—have committed $725 billion in capital expenditure for 2026 alone, a 77% jump from the $410 billion record set in 2025 . Goldman Sachs projects that combined hyperscaler capex could reach $1.4 trillion by 2027 and a cumulative $5.3 trillion by 2030 .

On the debt side, the largest tech companies have already issued more than $170 billion in corporate debt in 2026, exceeding their total for all of 2025 . Morgan Stanley estimates global AI-related debt issuance will top $500 billion annually . Goldman's Max Lukianchikov noted that so far, investors have been focused on yield, but warned that as AI spending continues to rise, "supply could begin to overwhelm demand" .

Brian Garrett's 'brutal' market warning

Goldman Sachs derivatives trader Brian Garrett issued a cross-market warning in July 2026 that has since circulated widely. He flagged that credit spreads on a large basket of technology company bonds widened by 22 basis points in a single week, while equities remain near all-time highs and volatility sits at rock-bottom levels .

Garrett warned that equity markets have historically ignored signals from credit markets "until they suddenly pay attention—and then everything falls apart" .

Goldman's own FICC desk described the situation starkly: "It is hard to remember a larger disparity between price and sentiment within IG [investment-grade] credit." The desk added that "the messaging from credit investors is increasingly clear that it will be very difficult to execute at current levels" .

Perhaps most tellingly, Goldman reported that the market's stress threshold has collapsed: where $75 billion in issuance once pressured the market, now only $25 billion does . This suggests that the bond market's capacity to absorb new AI-related debt is rapidly shrinking.

Bond-equity disconnect widens

There is a growing divergence between credit and equity markets. Credit spreads on tech bonds are widening—MSCI reports that credit spreads in IT and communication-services high-yield sectors have widened sharply in 2026 as fears of AI disruption grow —yet equities remain near all-time highs .

Investment-grade bond markets that had previously been a safe haven during AI-driven equity swings are now showing signs of strain . Goldman's FICC team described this as an unusually large disparity between price action in credit and the sentiment of credit investors . The message: bond investors are getting nervous, while stock investors are not yet pricing in the risk.

S&P downgrades Oracle to one notch above junk

On July 10, 2026, S&P Global Ratings downgraded Oracle's long-term issuer credit rating from BBB to BBB-, placing it just one notch above speculative-grade (junk) status . It was a watershed moment for the AI debt story.

S&P cited Oracle's rapidly rising AI infrastructure spending, negative free cash flow, and heavy customer concentration—especially its reliance on OpenAI, which S&P explicitly flagged as a credit risk. According to S&P, nearly half of Oracle's contractual obligations are tied to OpenAI, and should OpenAI encounter distress, Oracle would face "significant underutilized capacity and financial repercussions" .

Oracle's AI-related capex has surged to $95 billion with prolonged payback periods , and the company plans to raise an additional $40 billion through debt and equity to fund its data center buildout . Despite the downgrade, S&P maintained a stable outlook, and Oracle's stock initially shrugged off the news as investors focused on its $638 billion cloud contract backlog .

The downgrade was not a surprise to credit analysts. By December 2025, Barclays had already downgraded Oracle debt to underweight, projecting a cash exhaustion date of November 2026 based on current capex guidance .

BIS warns of systemic risk from an AI downturn

The Bank for International Settlements (BIS), in its June 2026 annual report (Chapter I: "Progress and peril"), raised direct concerns about the AI investment boom . The BIS noted that the five largest hyperscalers are set to spend over a trillion US dollars on AI-related capex from 2025 through 2026, and that these commitments are outpacing earnings and free cash flow .

The BIS warned that in the near term, the AI investment boom "raises questions about the sustainability of the current economic expansion," and that a downturn in the AI sector could have broad financial stability implications given the scale of debt and capital at risk .

This aligns with related warnings from investment firm Man Group, which described AI credit markets as at risk of a "violent" correction in June 2026, noting that investors face "significant execution risks and delays in development, yet they do not reap any equity benefits if the anticipated boom materializes" .

What this means for investors

The convergence of warnings from Goldman Sachs traders, the BIS, and rating agencies points to a market where unprecedented AI-driven debt issuance is straining credit markets, creating a dangerous disconnect between complacent equity markets and deteriorating credit conditions. Oracle's downgrade serves as the highest-profile corporate casualty so far, but the underlying dynamics suggest more may follow.

As one Goldman desk put it: the market's ability to absorb new supply has collapsed from $75 billion to $25 billion. If that threshold continues to shrink, the next AI-related bond deal could be the one that breaks the market's back.