CoreWeave's "double whammy in stocks and bonds"the "eye of the storm" in this round of the AI crash

This infrastructure company focused on AI cloud computing is becoming the most concentrated embodiment of the risks building up in this round of artificial intelligence investment enthusiasm. This week, CoreWeave's stock plunged 25% in a single week, its worst performance since November 2025; at the same time, the yield on its junk-rated bonds soared to 15%, typically seen as a critical signal of debt distress, and credit default swap pricing indicates a probability of default exceeding 60% within five years. Although Nvidia CEO Jensen Huang publicly expressed confidence in CoreWeave on Thursday, and Deutsche Bank characterized the selloff as a "buying opportunity," the stock rebounded 5.9% that day, but market unease did not dissipate, and on Friday CoreWeave's stock fell another 12%, ending the week with a 25% decline. The ripples set off by this "double whammy in stocks and bonds" have spread outward: the Philadelphia Semiconductor Index fell 11% this week, down about 24% cumulatively from its June 2025 high; the Nasdaq 100 recorded its worst weekly performance since April 2025, falling 5.5%; the average yield on high-yield bonds rose to 8.41% on Wednesday, a nearly 18-month high, with the risk premium widening to its broadest level since March. Bond market alarm: from an individual stock crisis to an industry signal In this round of turmoil, moves in the bond market are seen by some market participants as a more far-reaching signal than stock prices. CoreWeave's bonds, issued in June 2025 with a yield of 9.675%, saw their yield surge to 15% this week, touching a level typically deemed "distressed." At the same time, the cost of debt default protection rose during Thursday's session to the highest level on record. Max Frumes, head of financing business at credit data and analytics firm 9fin, said: CoreWeave is absolutely the 'canary in the coal mine' for this market segment, and this will become a turning point for the pricing, accessibility and structure of credit underpinning the AI boom. Pricing will rise, scrutiny will be stricter, and investors may show. Jim Reid, head of global macro and thematic research at Deutsche Bank, noted that the market is entering a new phase in which debt financing is growing in importance. He wrote in his latest research report: As we enter a capital expenditure cycle driven by debt rather than free cash flow, we must increasingly focus on the trajectory of credit prices. The cost of AI financing is generally not a problem, but worrying signs are appearing at the margins. From a broader perspective, the U.S. bond market has issued $106 billion of investment-grade and high-yield bonds for project financing so far this year, equivalent to nearly three times the $37 billion for all of 2025, with a large share of the funds flowing to AI investments such as data centers. According to estimates by Goldman Sachs strategists, global project financing bond issuance could exceed $300 billion next year. Junk bond market under pressure, cloud computing companies hit by financing chill CoreWeave's debt distress has already had a clear on the entire AI infrastructure financing ecosystem. According to Bloomberg, bonds related to data centers including Tract Capital Management, Prime Data Centers and Cipher Digital all fell this week to their lowest levels since issuance. The average dollar price of high-yield data center bonds slid from around $101 in June 2025 to about $95. Goldman Sachs's U.S. AI Data Center basket index (GSUCADSP) experienced a sharp spike early this week. CoreWeave's co-founder said this week that the company is exploring financing in the investment-grade debt market, including the vast pool of insurance institution capital, a move that sparked concerns the company will borrow heavily on a large scale, further intensifying selling pressure. Grant Nachman, founder and chief investment officer of Shorecliff Asset Management, said: At current market price levels, large-scale unsecured financing for new cloud computing companies is either prohibitively expensive, simply impossible to complete, or both. He believes that new debt issuance, if it is to attract buyers, may have to be secured by assets or contracted cash flows and offer stronger creditor protections. On Thursday, JPMorgan and Goldman Sachs increased the concessions offered to investors (sweetened terms) on a $5 billion leveraged loan they underwrote for AI cloud computing startup Volta Infrastructure Holdings. The loan ultimately priced at an actual yield above the 11% initially discussed, matching the highest interest rate record in AI infrastructure financing to date. Amanda Lynam, chief credit strategist at Goldman Sachs, had previously warned that the surge in corporate issuance in the U.S. high-yield bond market is starting to overwhelm debt investors. John Yovanovic, co-head of leveraged finance at MetLife Investment Management, pointed directly to the core of the problem: The problem is not that credit conditions have changed over the past year, but that a massive amount of supply is still coming over the next year and a half. Stock market level: valuation re-rating of the high-leverage model In the stock market, CoreWeave's highly leveraged business model is facing a harsher re-pricing. Since its IPO at $40 in March 2025, CoreWeave's stock once soared more than 360% on the explosion in AI demand. But a sharp pullback followed, and it has now fallen about 67% cumulatively from its peak to around $60, less than $25 away from its $40 IPO price. It is worth noting that the company's initial public offering itself was rather tortuous and required Nvidia's "backstop" support. As for this selloff, optimistic analysts believe the market reaction has been excessive. Deutsche Bank analyst Brad Zelnick maintains a buy rating and noted in a Thursday research report that CoreWeave has "a strong customer base, a proven execution track record, backlog conversion capability and leading infrastructure performance," and that the fundamental picture of AI computing demand continuing to exceed supply has not changed. However, changes in the macro environment make it difficult for optimism to fully take hold. Thomas DiFazio, ETF strategist at Roundhill Financial, said: The macro backdrop has created a 'sell first, ask questions later' market. In the short term, it is hard to say everything has turned around. Robert Schiffman, an analyst at Bloomberg Intelligence, noted that CoreWeave's close relationship with Nvidia could in theory provide a source of capital when market financing channels tighten. But Huang explicitly denied this possibility while attending an event in Stockholm, saying: CoreWeave does not need a bailout. Their strength is growing every day. I am optimistic. Their cost of capital is actually falling. Market divergence: short-term panic or a turning point In this turmoil, market judgments on CoreWeave and the broader AI investment logic show clear divergence. The pessimistic side focuses on: rising U.S. Treasury yields and persistently high inflation pushing up financing costs; the long cycle from construction to returns for AI data centers; the risk that GPU equipment may face obsolescence before new facilities begin operating; and rising political controversy over data center construction permits and electricity pricing. Larry Holzenthaler, senior portfolio manager at fixed-income asset management firm Catalyst Funds, said: A few months ago, I would have said people were too optimistic about AI. He believes that the risks ignored or downplayed by the market in June this yearhuge financing needs, high leverage, and reliance on external financingare now being re-priced in a more violent way. The optimistic side believes that the AI capital expenditure cycle is still in its early stages and that the long-term logic on the demand side has not wavered. Matt Bartolini, head of global research strategists at State Street Investment Management, said: We are still at a very early stage in the AI capital expenditure cycle. Interest rates will not become an insurmountable obstacle, but they do change the calculus behind continued debt issuance and whether the financing of that debt can match credible future growth. The long-term demand story is what fundamentally drives the capital expenditure cycle, and I think demand will persist regardless of where interest rates go. John Han, portfolio manager at fixed-income asset management firm F/m Investments, said his team is currently "re-examining our overall exposure in AI infrastructure, especially where risk is concentrated and whether we are being adequately compensated for it." At the heart of this debate is a historic moment of valuation coordinate reset for AI investors: as the era of free money ends, the sustainability of the business models of debt-driven AI infrastructure companies is undergoing an unprecedented market test. This article is reprinted from Wall Street See and Hear, authored by Bao Yilong; GMTEight editor: Wenwen.