Chip Stocks Fall 8 Straight Days as AI Doubts Meet Rate Fears

Semiconductor stocks just logged their worst stretch in more than a decade, with the iShares Semiconductor ETF (SOXX) falling for an eighth consecutive session as doubts about AI spending collided with growing expectations of Federal Reserve rate hikes. SOXX closed at $514.71 on Thursday, down roughly 12.6% across the eight sessions, its longest losing streak since September 2015. Nvidia, AMD, Intel, memory makers and AI cloud providers have all been marked down along the way.

Eight straight sessions, the longest streak since 2015

SOXX is a broad gate into the industry rather than a bet on one chip designer. The iShares fund tracks the NYSE Semiconductor Index, holds about $44.5 billion in net assets, charges a 0.33% expense ratio and splits exposure roughly 83% across chip makers and 17% across semiconductor equipment names. That structure is why one theme, the price of capital, can drag the entire basket lower at the same time.

One more red day would match a record set in 2006.

The scale reads better in index terms. The Philadelphia Semiconductor Index, the benchmark most traders watch for the sector, was down 3.3% on Thursday, adding to a market-value loss of about $1.64 trillion over six sessions. The 20 largest AI and semiconductor companies by market cap shed more than $1 trillion across four trading days, and memory makers accounted for roughly 40% of that global loss.

SOXX had still returned about 81% year to date through October 13, so this is a giveback from a very large run rather than the end of the trade. Holders are left well above the $260.44 low set in November 2025 and well below the $655.95 high reached in June 2026. One mechanical wrinkle matters for anyone reading levels: iShares has a forward split with a November 3 record date, and split-adjusted trading begins November 5, which resets the quoted price without changing value.

Is the AI bubble actually cracking?

Not proven. What cracked so far is the assumption that AI spending keeps compounding regardless of financing cost.

The cracks trace back to specific data points rather than a single headline. A Financial Times report said OpenAI told investors its annualized revenue was approaching $50 billion at the end of September, roughly $20 billion below the $70 billion the company had previously projected, and analysts still saw chip stocks sell off as the figure was debated. In September, an essay from Anthropic CEO Dario Amodei warning that the most advanced systems could slip beyond human control sent the semiconductor index down as much as 5.9% in one session.

Not everyone reads it that way. In September, Bank of America semiconductor analyst Vivek Arya called the reaction background noise against a secular market where AI capital expenditure could grow threefold to more than $3 trillion by the end of the decade. Even after this year's gains, the chip index trades near 26 times forward earnings, a fraction of the roughly 150 times seen at the dot-com peak.

Financing structure is the argument now, not the multiple.

Why do rate fears matter for chip stocks?

Because the AI buildout is being financed with borrowed money, so higher rates raise the cost of every data center, every GPU cluster and every long contract priced off cheap capital. US consumer prices rose faster than economists expected in September, and traders responded by pricing the possibility of Federal Reserve rate hikes, with the odds of a hike reportedly climbing as high as 32% during October.

The heaviest balance sheets fell hardest. Intel and AI cloud provider CoreWeave each lost more than 20% over two weeks.

Nvidia fell about 1.8% on Thursday to roughly $217.23, less than 9% below levels above $237 earlier in the month, a display of relative strength inside a broken sector. CoreWeave's bonds pushed yields higher as investors priced the risk of a company whose revenue depends on borrowing to build capacity its customers have not yet paid for.

On Thursday, Nvidia CEO Jensen Huang said the company would not provide financial guarantees to CoreWeave, which eased one specific worry. CoreWeave rebounded 5.9% in the following session before slipping back after hours. Two buckets are forming: companies generating cash from AI revenue today, and companies pledging future revenue against debt raised today.

Where memory makers and South Korea stand

South Korea is where the pain has been most violent. The KOSPI fell 7.6% on Thursday, triggering a circuit breaker that halted trading, then rebounded 3.4% on Friday with Samsung Electronics up a similar margin. Over the seven sessions from October 2 to October 15 the index dropped 20.70%, from 7003.74 to 5554.22, its worst run since the COVID shock of March 2020, and it still sat about 19% below where it closed on October 2.

SK Hynix lost 11% in a single session and Samsung dropped 9.5% as fears that higher rates could curb AI expansion spread through the memory complex. Micron traded near $931, well below the $1,050 it touched earlier in the month. Three names, Samsung, SK Hynix and Micron, account for roughly 40% of the market value destroyed across global AI and semiconductor companies, which tells you the selling has been concentrated in the most cyclical corner of the trade.

Margin debt makes it worse. KOSPI and KOSDAQ credit loan balances stood at 32.75 trillion won on October 15, 13.2% above the end of July level.

CNBC's Jim Cramer, writing on X, called SK Hynix the epicenter of fear precisely because the KOSPI remains more than 50% above its low, leaving plenty of holders underwater even after the plunge. Forced unwinds of borrowed money have a way of extending a decline past the point where the underlying business changed.

What does this mean for bitcoin miners and crypto?

For crypto, the transmission channel runs through equity correlation rather than fundamentals. A July 2026 report from 10x Research found that bitcoin mining stocks had decoupled from bitcoin's price and grown far more synchronized with AI and semiconductor narratives, with Riot Platforms showing a rising correlation to the SOX ETF since April 2026. Both the mining names and the semiconductor index have since pulled back from their highs together, which is the pattern traders are watching.

Bitcoin itself has no exposure to memory pricing or data center debt.

The rate path still matters on its own terms, since higher borrowing costs tighten global liquidity and raise the cost of carry across speculative assets. The exposure that matters more is structural: miners that redirected capital budgets toward AI hosting now carry the same financing and depreciation questions as the AI cloud providers being punished this month.

A chip rebound would lift AI-pivoted miners with it.

What are the levels and dates to watch next?

Three numbers frame the near term. The first is whether SOXX can hold its recent lows, with the range bottoming near $498.55 against the $655.95 high set in June. The second is the nine-session record from 2006, which one more down day would tie. The third is the KOSPI, where the 5554.22 close on October 15 marks how much of the 20.70% drop still has to be recovered before the rebound counts as anything more than a bounce.

Three dates: the next US inflation print, the Fed meeting after it, and Nvidia's quarterly report.

Inflation data is what shifted rate expectations and with them this entire move, and Nvidia's results, which in past years landed in the third week of November, now carry outsized weight as the clearest read on order books and customer financing. November 5 belongs on any chart too, since SOXX starts trading on a split-adjusted basis that day. The burden of proof sits with the bulls: cash flow from AI deployments has to appear in reported numbers soon enough to justify the debt raised to build it.

Until that shows up, every quarterly report and every inflation print gets read twice, once for the data and once for the financing behind it.