Morning Edition · Saturday, October 17, 2026Published at 1:06 AM EDT · New York
CoreWeave fell nearly 12% and AMD, Intel and Micron each lost more than 5%, and the S&P 500's 2% weekly decline erased the previous week's gains to a record high.
On Friday the Nasdaq Composite fell 2.14% to 26,140.03, the S&P 500 lost 1.09% to 7,659.44 and the Dow Jones Industrial Average slipped 0.34% to 52,002.41. CoreWeave, an AI cloud operator financed heavily with debt, fell nearly 12%. MarketWatch noted that software and semiconductor stocks had seldom moved together this year but fell together on Friday as fears about AI and interest rates hit the market. AMD fell 5.95%, Intel 5.79%, Micron 5.63% and Broadcom 4.52%. Energy shares gained 2.52% as crude rose.
For the week, the S&P 500 lost about 2%, erasing the previous week's gains to an all-time high. The Nasdaq posted its worst week since June. Investors are questioning the debt behind AI data centers just as 10-year borrowing costs stay near 5%.
Gary Marcus, the AI researcher and longtime critic of large language models, gave the bearish case in comments reported by the Economic Times. He questioned OpenAI's ability to meet its financial commitments and warned that the failure of leading AI companies could cause losses across the financial system on the scale of the 2008 banking crisis. Marcus has made this argument for years, and the market's own pricing is a better indicator. Friday's sharpest falls hit the most leveraged infrastructure names, not the companies with the largest cash reserves.
The retail investors who supported the rally have not left. Israeli business daily Globes reported from a Robinhood conference for high-risk traders that "everyone has a story about how they got rich." The paper described ordinary Americans as one reason the stock market keeps setting records.
Part of a tracked trend
AI Build-Out Meets the Credit Cycle
While long-term interest rates stay above 5%, the debt-financed part of the AI infrastructure build-out will keep facing rising funding costs, so stress at the weakest borrowers will keep spreading to their lenders, their suppliers and AI-concentrated equity markets.
What this means
The AI build-out was financed on the assumption of cheap, plentiful credit. With 10-year yields above 5% and the Fed still leaning toward hikes, the most exposed companies are those that borrowed against future AI revenue: neocloud operators (cloud providers built specifically to lease out AI computing capacity), chip suppliers whose order books depend on those borrowers, and the private-credit funds that lent to them. The cash-rich hyperscalers face lower valuations but no funding gap. What is already priced is a slower pace of AI spending. What is not fully priced is refinancing risk at the leveraged end of the chain. The closest precedent is the 2000-2002 telecom bust, where the carriers that borrowed to build fiber failed while the technology itself survived. If yields stay near current levels, the likelier sequence over the next few months is wider credit spreads for AI infrastructure borrowers and more equity issuance, before any cut in hyperscaler capital budgets. The read would be proved wrong if strong cloud revenue in the coming earnings season shows customers paying fast enough to service the debt.
What to watch
Observations to monitor, not financial advice.
Synthesized from: MarketWatch · Economic Times · Globes (Hebrew) · TradingKey
Start a discussion in Townsquare.
More from this edition
Comments
1Oct 17, 5:06 AM
CoreWeave's 12% drop was about double the losses at AMD, Intel and Micron, which suggests the selloff priced financing risk at debt funded operators more heavily than chip demand.