The AI trade had its roughest week in years. The Philadelphia Semiconductor Index (SOX) fell 3.83% on Friday, October 16, closing at 11,137.29. That made eight losing sessions in a row, the index's longest losing streak since October 2008, and wiped out about 16% of its value since its last gain on October 6. Every one of the index's 30 members ended Friday lower.
The selloff has not undone the year. Chip stocks are still well ahead for 2026. What has changed is the tone: investors are now asking whether the money flowing into AI data centres, much of it borrowed, can earn a return before financing costs bite.
A Broad Rout Across the AI Hardware Stack
Friday's losses were wide and deep. Investing.com reported that Broadcom, Micron, AMD, Intel, Lam Research and KLA each fell about 5%, and Arm Holdings dropped roughly 7%. Figures compiled by 24/7 Wall St. put AMD down 5.95% for the day, Intel down 5.79%, Micron down 5.63%, Marvell down 5.56% and Applied Materials down 4.82%.
The pain spread beyond chips:
- The Nasdaq Composite lost 2.1% and the S&P 500 fell 1.1%, according to Investing.com.
- The Nasdaq 100 dropped 2.5%, and tech stocks lost more than $1 trillion in market value in one day, per 24/7 Wall St.
- The iShares Semiconductor ETF (SOXX) fell 4.14% to $493.39. That leaves it about 16% below its October 6 close and roughly 25% below its 52-week high.
Nvidia, the company most tied to AI capital spending, held up comparatively well with a 1.43% decline. Equipment makers and memory suppliers took the hardest hits.
The streak follows a remarkable run. Before the slide began, the SOX had gained more than 60% in 2026, and SOXX remains up about 64% year to date. AMD shows how stretched some valuations had become. The stock fell more than 18% between October 6 and October 16 but is still up about 147% for the year and trades near 199 times trailing earnings, according to 24/7 Wall St.
Debt, Rates and the CoreWeave Signal
The clearest stress signal came from AI infrastructure credit. CoreWeave, the GPU cloud provider carrying roughly $50 billion in debt, fell 12% on Friday, bringing its five-session loss to about 25%. Investing.com reported that yields on its bonds rose above 15%, a level the bond market usually reserves for distressed borrowers.
That matters because so much of the AI build-out now runs on borrowed money. When financing gets more expensive for companies like CoreWeave, investors start to doubt the order books of the chipmakers that supply them.
Macroeconomic pressure added to the strain. Hotter-than-expected US inflation data raised the risk that interest rates stay high for longer. 24/7 Wall St. noted that West Texas crude reached $101.91 a barrel on October 14 and the 10-year Treasury yield hit 5.10% on October 15. Both make it costlier to fund long-dated, capital-heavy projects. Reports that China's CXMT had made progress on advanced memory chip architecture also weighed on memory names.
Analysts Question the Capex Cycle
The core worry is the size of AI spending commitments. Bank of America strategist Sebastian Raedler, cited by 24/7 Wall St., pointed out that consensus forecasts for hyperscaler capital expenditure have roughly tripled, from about $330 billion in early 2025 to around $1 trillion. He cautioned that "most major capital cycles of the past have ended in disappointment," and argued that equipment makers are the most exposed, since chipmakers can quickly postpone new fabs when customers slow their orders.
Not everyone sees a collapse. Nvidia CEO Jensen Huang has played down the turbulence, saying it will pass. Market structure also points to orderly repositioning rather than panic: the VIX volatility index was 16.46 on October 15, below the level usually linked to market fear. Unlike in 2008, there is no broad credit crisis behind the selling.
Why It Matters
For the AI economy, chip stocks are the most direct public-market gauge of confidence in the build-out. A record losing streak tells you investors are no longer paying any price for AI exposure. They now want proof that spending turns into revenue.
The consequences reach well beyond trading desks:
- Enterprise buyers may see GPU cloud providers tighten pricing or terms if their own borrowing costs keep climbing.
- AI startups that rely on infrastructure partners or venture debt could face tougher financing conditions heading into 2027.
- Hyperscalers face more scrutiny of their capex guidance in the coming earnings season, and any cut to data-centre plans would hit the semiconductor supply chain directly.
The next test is earnings. If the largest cloud and platform companies reaffirm their AI spending plans, this pullback may look like another correction in a volatile year. If they trim budgets, last week's selloff could turn out to be an early warning that the AI capex cycle is peaking.
