Photo: Nana Dua / Pexels
AI fatigue drives chip stocks to their longest losing streak since May 2025
A weaker-than-expected OpenAI revenue figure and a blunt warning from Broadcom's CEO have semiconductor investors asking where the money actually comes from
Chip stocks are having a rough October. As of October 13, 2026, the iShares Semiconductor ETF (SOXX) had fallen for five trading days in a row, its longest losing streak since May 2025.
The culprit has a name: AI fatigue. Investors who spent most of the year funding the AI buildout are now asking a simple question. When does all this spending start paying for itself?
A bad week, by the numbers
The damage runs deeper than a single ETF. Of the 30 stocks in the PHLX Semiconductor Index (SOX), 22 now sit more than 20% below the record highs they set in May or June 2026.
The trigger came on October 8. The Financial Times reported that OpenAI’s revenue run rate was approximately $50 billion. The market had been expecting something closer to $70 billion.
A run rate is a snapshot of current revenue projected out over a full year. So the gap suggests OpenAI is earning meaningfully less right now than investors had baked into their models.
The SOX fell about 3.4% that day. Individual names took harder hits:
The pain extended well beyond US megacaps. SK Hynix shares fell for seven straight sessions, shedding approximately 20% over that span. Cerebras dropped around 20% across six days.
The casino comment and other bruises
Investor nerves did not get any help from inside the industry. On October 8, the same day the OpenAI figure landed, Broadcom CEO Hock Tan compared speculation in the AI sector to “casino gambling.”
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Competition added another layer of worry. Advances from CXMT, a Chinese memory chip company, chipped away at confidence in the sector.
Meanwhile, the market is waiting on clearer revenue visibility from other AI developers, including Anthropic.
How we got here
Context matters, because this correction follows a monster run. Even after the recent slide, the SOX is still up more than 70% for the year.
Goldman Sachs projects US hyperscaler capital expenditure will reach approximately $800 billion in 2026, climbing to $1.1 trillion in 2027.
The OpenAI report put a dent in that story. If one of the most prominent AI companies is bringing in roughly $20 billion less on an annualized basis than expected, investors naturally start stress-testing the rest of the chain.
What this means for investors and the AI trade
The core shift is about evidence. For much of the past year, the market rewarded chip stocks on the promise of AI demand. Now it seems to want receipts.
One report about a single company’s run rate was enough to knock the SOX down about 3.4% in a day. With 22 of 30 SOX components more than 20% off their highs, this is not a story about one or two stocks stumbling. It looks more like a broad repricing of how much investors are willing to pay for future AI growth.
The index remains up more than 70% year to date, and Goldman’s capex forecasts point to spending that is still rising, not shrinking.
Revenue disclosures from AI developers such as Anthropic will be closely parsed for signs that demand is catching up with spending. Hyperscaler capex guidance will show whether the biggest buyers are holding firm or tapping the brakes. Competitive pressure from Chinese players like CXMT is another variable that could weigh on pricing expectations for incumbent memory suppliers.
Comments from industry executives will also carry extra weight. Hock Tan’s casino remark showed how quickly a few words from an insider can move sentiment when nerves are already frayed.