For the past two years, Nvidia's share price was the market's core gauge for tracking the AI trade. That role has now been handed to the unsecured bonds of a single-B-rated emerging cloud computing company.
This week, the yield on CoreWeave, Inc. (CoreWeave, Inc.) bonds maturing in 2032 briefly broke above 16%, while its five-year credit default swap (CDS) spread spiked to 1,117 basis points, implying a roughly 63% five-year default probability. In a credit trends report published over the weekend, Goldman Sachs stated bluntly that CoreWeave, Inc. has become the "market's beta barometer," with its bonds drawing widespread attention from equity and macro investors, and credit pressure thereby spreading across the entire data center ecosystem.
Meanwhile, UBS desk trader Torsten Sippel independently reached the same conclusion, saying CoreWeave, Inc. has become "an important barometer for the AI trade," and that the market debate is shifting from computing demand itself toward the economics and financing sustainability of AI infrastructure expansion.
This credit turmoil has dealt a direct blow to equity markets. Morgan Stanley's AI tech beneficiaries basket (MSXXAIB) posted its largest single-week drop since its inception in 2023, falling 16.5%, while the S&P 500 fell only about 2% over the same period. Goldman Sachs' emerging cloud computing stock basket (GSXUNEOC) plunged 24.8% in a single week, bringing its cumulative decline since June to 44%. Goldman Sachs has urgently arranged a cross-asset conference call, hosted by someone with the title "Head of Distressed/AI Debt Trading" — a position that did not exist a year ago.
Calm on the Surface, Turbulence Beneath
This week, investment-grade credit indices showed almost no movement — IG cash spreads widened by only about 3 basis points, IG100 widened by about 2 basis points, and IG funds continued to see net inflows of $1.06 billion during the week, bringing year-to-date inflows to $240.6 billion, about $101 billion ahead of the same period last year. Yet the indices masked a structural split.
AI-related high-yield credit spreads widened by 25 basis points over just two days, October 13 and 14, while non-AI credit widened by only 2 basis points. Oracle's long-end bond spreads widened by about 43 basis points to +320, notably underperforming Amazon and Google, with its CDS widening by 30 to 40 basis points; some names in the TMT sector saw intraday spread swings of 20 to 30 basis points. The average price of a sample of high-yield data center project financing bonds fell from 101.5 cents in June to 94.7 cents in October. CCC-rated spreads broke through 1,000 basis points this week, with Morgan Stanley calculating 1,045 basis points, a new high since 2023, while the market saw an "obvious rotation from CCC and stressed names into BB."
Goldman Sachs' credit valuation data reveal the extreme degree of internal divergence: US dollar investment-grade spreads sit at the 9th historical percentile, leveraged loans at the 2nd percentile, while US dollar CCC spreads stand at 1,062 basis points, at the 86th historical percentile. Goldman Sachs' credit desk summed it up this way: "Capital hasn't disappeared, it has just become highly selective."
30.7x Versus 1.8x: Pricing Power of the Marginal Borrower
At the core of this credit divergence lies AI infrastructure expansion's enormous reliance on capital markets. According to Goldman Sachs Global Investment Research (GIR), US dollar investment-grade and high-yield project financing issuance this year has already exceeded $106 billion, nearly triple the $37 billion in 2025, and almost all of it is for data center construction. GIR forecasts that global project financing issuance will reach $300 billion in 2027, with another $420 billion coming directly from hyperscalers.
Morgan Stanley has quantified the overall funding gap. According to its report "Tracking AI's $1.6 Trillion Funding Flow," combined capital expenditure by hyperscalers and emerging cloud companies will reach $1.58 trillion in 2027, of which $195 billion will be financed through leases (which Morgan Stanley calls "synthetic capex"), with another $1.118 trillion in leases signed but not yet commenced. Morgan Stanley expects combined free cash flow for hyperscalers and emerging cloud companies to hit its most negative level in 2027, and "if lease payments are included, the funding gap will be even larger."
Behind these numbers is a stark gap in financing capacity. Morgan Stanley's calculations show that four AA-rated hyperscalers could issue $400 billion in bonds at a 7% interest rate while still maintaining an interest coverage ratio of 11.9x; whereas Oracle and SpaceX, after raising $120 billion at an 8.5% rate, would see their coverage ratio plummet from 4.0x to 1.8x. Emerging cloud companies are even more fragile — Morgan Stanley notes that these companies "rely almost entirely on debt issuance to finance their capex ambitions," with CoreWeave, Inc.'s capital expenditure reaching $47 billion while still facing a massive funding gap.
For this reason, Goldman Sachs' Alisha Pasi framed the core question as: "distinguishing which companies can finance their growth from those whose growth itself depends on the continued openness of the financing window." CoreWeave, Inc. is the most typical representative of the latter, which is the fundamental reason it has become a barometer.
Chain Reaction: From Life Insurers to Hedge Funds
The transmission path of credit pressure has extended to unexpected corners. Goldman Sachs' credit desk noted that "financial and insurance CDS were eagerly bought this week, as the market worried about these entities' concentrated exposure to the AI trade" — pointing to life insurers and investment advisers that significantly increased their holdings of AI data center bonds this year. Oracle's CDS hit a record high of 300 basis points last Saturday, and Goldman Sachs strategist Michael Hartnett has advised clients to stay risk-averse over the next three to four weeks, focusing on price moves in indicators such as Oracle CDS.
At the hedge fund level, Morgan Stanley Prime Brokerage observed a 3-standard-deviation sell-off, but AI themes still account for about 40% of US long-short net exposure, with about 95% of long-side returns in long-short portfolios coming from AI-related trades. UBS shows net exposure at the 30th historical percentile, while gross exposure is above the 90th percentile, "leaving portfolios vulnerable to further deleveraging if P&L pressure persists."
The bond market played a rare hedging role this week. Although core CPI rose 0.33% month-on-month, briefly pushing the 10-year Treasury yield above 5.30%, yields ultimately fell about 15 basis points from the start of the week to close at 5.07%, as "the force of AI-related growth concerns and the stock market decline overwhelmed the inflation shock." The probability of an October rate hike has fallen to the mid-teens percentage range, and Goldman Sachs' economics team expects a pause then, with another hike in December.
The Bull-Bear Debate: Is This De-bubbling or De-risking
Not everyone holds a pessimistic view. Goldman Sachs' Mark Wilson believes the market pendulum has swung too far, and UBS characterized this week as "deleveraging rather than de-risking," similar to how the earlier Situational Awareness event unfolded. Phoenix Capital's Graham Summers argues that if CoreWeave, Inc. eventually goes bankrupt, it is merely "how a construction boom normally works" — physical GPU assets don't just disappear, and hyperscalers "don't need capital at 15%."
However, Goldman Sachs' Pasi may have offered this week's most weighty judgment: "The question is not whether investors are still willing to take on credit risk, but which risks they are willing to hold, and how high a price they are willing to pay for them." Goldman Sachs rates strategist David Curtin also noted that current pressure "does not come from a compression in expected returns on invested capital, but more from financing costs and financing scale themselves."
Structurally, Morgan Stanley expects new issuance in 2027 to shift toward short-duration, chip-collateralized asset-backed securities — precisely the same type of structure that Goldman Sachs and JPMorgan refused to hold at an 11% rate in the Volta chip loan case, providing a footnote on the depth of the current predicament.
What to Watch Next
The market's next batch of key data will arrive thick and fast. Meta will report earnings on October 21, a week early, and UBS believes its subsequent bond issuance may be intended to "get ahead of the (financing competition) crowd." Microsoft will report earnings on October 28, the same day as the Federal Reserve's policy meeting. Goldman Sachs rates strategists note that commodity trading advisers (CTAs) still hold about $164 million in DV01 of Treasury shorts, with $6.5 million to $10 million in DV01 of potential short-covering room over the next month in a strong rally scenario.
The ultimate long-term indicator Hartnett offers points to chip prices: if semiconductor prices resume their climb, it would be a positive for AI credit; once they decline, it would signal "further tightening of AI credit, a peak in AI capex growth, and 2027 nominal GDP below expectations."
When Goldman Sachs calls a junk-rated emerging cloud computing company the "market's beta barometer," the nature of the AI trade has shifted from an equity story to a credit story. Where is the barometer's needle pointing? Until CoreWeave, Inc. can fill a roughly $35 billion funding gap at non-junk rates, every AI rebound may, in credit terms, be nothing more than a bear market rally.
Comments