AI Fatigue Is Gripping the Market and Driving Chip Stocks Toward Their Worst Stretch in 16 Months

The year was 2023, and I remember sitting in a cramped Manhattan newsroom, watching a colleague frantically refresh a live stock ticker. Nvidia had just blown past earnings expectations, and the entire room erupted like we'd all personally won the lottery. A senior editor, a grizzled veteran who'd survived the dot-com bust, leaned back in his chair and muttered something I've never forgotten: "Every party has a last guest who doesn't know when to leave."

Chip stocks have been that guest for nearly two years now. But on October 13, 2026, the music started to fade.

The iShares Semiconductor ETF (SOXX) fell for the fifth consecutive trading day, pacing toward its longest losing streak since May 2025. The Philadelphia Semiconductor Index (SOX) shed 2.7% on Tuesday, following a matching decline the prior session. Of the 30 stocks in that index, 22 now sit more than 20% below their record highs, nearly all of which were set back in May or June.

This isn't a garden-variety pullback. This is what happens when a market that has been running on pure adrenaline suddenly remembers it has lungs.

The OpenAI Number That Broke the Spell

Every correction has a trigger, and this one came from a number that, on its face, sounds perfectly respectable.

On October 8, the Financial Times reported that OpenAI's annualized revenue had reached approximately $50 billion. Half a hundred billion dollars. For a company that barely existed three years ago, that's staggering. But here's the rub: the market had been expecting closer to $70 billion.

That $20 billion gap might not sound like much in a world where governments routinely debate trillion-dollar budgets. But in the context of AI infrastructure spending, it's everything. Because if OpenAI, the company that kicked off this entire gold rush with ChatGPT, can't hit its revenue targets, what does that say about the sustainability of the spending spree it inspired?

The market's answer came swiftly and brutally. The SOX fell about 3.4% that day. Intel dropped 5.3%. Micron fell roughly 4.8%. Broadcom slid 4.3%. Nvidia lost about 3%. In a single session, Nvidia's market value evaporated by roughly $169 billion.

But wait, there's a wrinkle. According to analysts at AInvest, the market may have misread the OpenAI figure entirely. The $50 billion annualized revenue versus the $70 billion that had been circulating represents a difference in accounting methodology, not a loss of demand. OpenAI reportedly leaves out sales that flow through its cloud partners, while rivals like Anthropic count them.

That's the thing about market panics. They rarely wait for nuance.

Broadcom's CEO Said the Quiet Part Out Loud

If the OpenAI number lit the match, comments from Broadcom CEO Hock Tan poured gasoline on the fire.

Speaking at an event in Singapore, Tan compared some AI investors to "casino gamblers," questioning whether they fully understood the risks they were taking. He also flagged that capital, not demand, could become the biggest bottleneck in the AI buildout.

Those are startling words from a man whose company is one of the primary beneficiaries of the AI spending boom. When the person selling shovels tells you the gold miners might be gambling, you listen.

Jefferies analyst Jeffrey Favuzza wrote in a Tuesday trading note that one of the biggest questions he's been getting from clients is simply: "why the weakness in AI?" across the board. His answer pointed to a combination of Tan's capital bottleneck warnings and the anticipation surrounding Anthropic's investor day, where investors are reportedly looking for upward of $80 billion in annual recurring revenue.

That's the kind of expectation that makes or breaks a market narrative. And right now, the market is holding its breath.

The Hyperscaler Spending Question Nobody Wants to Ask

Here's where things get a little uncomfortable for anyone who's been cheering on the AI trade.

The entire bull case for chip stocks rests on the assumption that hyperscalers, Microsoft, Amazon, Alphabet, Meta, will keep spending at a pace that boggles the mind. And they have. UBS estimates that hyperscaler capex will rise 76% this year to $673 billion. Goldman Sachs projects US hyperscaler capital expenditure will reach approximately $800 billion in 2026, climbing to $1.1 trillion in 2027.

Those are numbers that don't fit neatly into human comprehension. A trillion dollars. That's more than the GDP of most countries. And it's being spent on data centers, processors, memory chips, electricity, and cooling systems in a race to build computing infrastructure that may or may not generate returns at the scale investors are pricing in.

But here's the part that makes people nervous: UBS expects that growth to slow dramatically. After rising 76% this year, hyperscaler capex is projected to increase by only 25% next year and just 6% in 2028.

Slowdowns aren't inherently bad. A 6% growth rate in a mature industry would be healthy. But semiconductor valuations aren't priced for mature growth. They're priced for the kind of explosive expansion that only happens once. AMD trades at roughly 154 times earnings, according to one analysis. That's not a valuation, that's a prayer.

"When they stop increasing their capital spending, that will certainly be a relief for hyperscalers and a negative signal for the semiconductor sector," said Alexis Bossard, a portfolio manager at Edmond de Rothschild Asset Management, who has already reduced his exposure to chip stocks he considers too expensive relative to expectations.

China's CXMT and the Memory Chip Disruption Nobody Saw Coming

If you're looking for the most underappreciated risk to the AI chip trade, forget about OpenAI's revenue for a moment. Look at ChangXin Memory Technologies.

CXMT, China's largest DRAM maker, announced progress on next-generation DDR5 server-memory modules that use less area per memory cell than conventional designs. The company plans to introduce these modules by the end of 2026.

On paper, this sounds like a technical footnote. In practice, it sent shockwaves through the memory chip market. Samsung Electronics and SK Hynix, the two dominant players in DRAM, saw their shares drop roughly 7% and 10% respectively on Tuesday. Micron, the third major player, has been caught in the crossfire.

The concern isn't that CXMT will immediately capture massive market share. It's that the company's progress changes the narrative. For years, investors assumed that memory chips, particularly the high-bandwidth memory used in AI accelerators, were a cozy oligopoly with high barriers to entry. If a Chinese competitor can achieve technological breakthroughs and scale production, that assumption starts to crumble.

Market participants are worried CXMT "is going to increase capacity of the industry and create a price war," said Hao Hong, chief investment officer at hedge fund Lotus Asset Management.

Of course, not everyone is convinced. Some analysts have listed seven reasons why CXMT's claims don't pose an imminent threat to Micron and SanDisk. But in a market this jittery, even the perception of a threat is enough to trigger selling.

The Global Contagion

What started as a US semiconductor selloff has gone global with a speed that's frankly alarming.

In South Korea, the KOSPI index plunged more than 6% on Tuesday as selling of chipmaker shares entered a fifth straight session, its longest losing streak in more than two years. Samsung Electronics and SK Hynix dropped 10% and 8% respectively on Monday. In Tokyo, memory-maker Kioxia fell 13%.

The KOSPI is now down a quarter from its July highs. Shanghai's tech-focused STAR Market has shed more than a third from its July peak.

Signs of stress are appearing in primary markets too. Firmus, an Nvidia-backed data center company, pulled a planned $5 billion Australian IPO last week after failing to attract sufficient investor demand. That's a data center IPO collapsing in a market where data center demand is supposedly insatiable. The irony isn't lost on anyone.

Even the credit markets are flashing warning signs. Tencent's recent dollar and dim sum bond sale in Hong Kong saw guidance for 10-year notes at about 1.15 percentage points above Treasury yields. Japan's SoftBank sold the biggest high-yield debt pile ever, promising to pay as much as 9.75% on $4.5 billion it raised. When AI-related borrowers are paying near-junk-bond rates, the cost of capital is becoming a real constraint.

"The most important development has been not just bond yields going up, but the volatility for bonds," said Damien Boey, portfolio strategist at Wilson Asset Management in Sydney. "That's starting to work its way out through asset pricing across the board".

What the Bulls Are Saying (Because You Should Hear Both Sides)

Not everyone thinks this is the beginning of the end.

Dan Ives, the outspoken tech analyst at Yorkville Ives & Co., has been characteristically blunt: the semiconductor selloff is "way overdone". He acknowledged there's "some nervousness" around memory stocks but argued the broader reaction doesn't match the fundamentals.

His reasoning centers on supply and demand. Ives has pointed out that the GPU supply-to-demand ratio remains wildly imbalanced, with some estimates suggesting demand outstrips supply by a factor of 12 or more. He's also noted that data center spending is only about 15% complete, suggesting there's still a long runway ahead.

Marc Velan, head of investments at Lucerne Asset Management in Singapore, agrees that the "brutal session" in Korea said more about positioning and the cost of capital than about any fundamental deterioration in AI demand.

And there's something to that argument. The SOX remains up more than 70% for the year even after the recent slide. Micron has tripled in value since the start of 2026. Anyone who bought chip stocks 12 months ago and held through this turbulence is still sitting on enormous gains. The selloff feels painful because it's happening after a period of euphoria, not because the underlying businesses have suddenly stopped growing.

BlackRock, the world's largest asset manager, moved its US equities stance to neutral on Monday, citing "potential volatility around the AI buildout" as the rationale. But it also shifted to a favorable view on short-term Treasuries, a defensive positioning, not a full retreat.

What Retail Investors Are Feeling

There's another dimension to this story that doesn't always make it into the financial pages.

AI fatigue isn't just a market phenomenon. It's a cultural one. A survey from Talker Research found that 54% of Americans are tired of hearing about AI. Two-thirds of knowledge workers reported feeling nostalgic for pre-AI work, according to research from Adaptavist. Nearly 40% said they'd like to remove the technology from their work entirely.

That's the backdrop against which this market drama is playing out. Ordinary people are exhausted by AI. They're tired of the hype, skeptical of the promises, and increasingly doubtful that the technology will deliver on its grandest claims.

When you combine that cultural fatigue with the financial fatigue now showing up in chip stocks, you get a market that's vulnerable to any piece of bad news. The OpenAI revenue miss, the CXMT breakthrough, Broadcom's casino comment, each one lands harder because investors are already primed to believe the AI story might be overblown.

The Anthropic Wildcard

If there's a near-term catalyst that could reverse this slump, it's Anthropic's investor day.

The Claude maker is targeting a pre-Thanksgiving stock market listing, with bankers potentially beginning its IPO roadshow during the week of November 9. It's scheduled to host prospective investors at its San Francisco headquarters on October 14.

Prospective investors have valued the company at between $1.8 trillion and $2 trillion. Those are numbers that make even the most bullish AI optimists do a double-take. And the annual recurring revenue expectations, upward of $80 billion, according to Jefferies, are equally staggering.

Anthropic's investor day is either going to validate the bull case or pour cold water on it. If the company reports strong revenue traction and a credible path to profitability, chip stocks could snap back quickly. If it disappoints, the selloff could accelerate.

Semiconductor investors are clearly on edge heading into the event. You can almost feel the collective breath being held.

The Interest Rate Squeeze

Here's something that doesn't get enough attention in AI stock discussions: the cost of money.

Rising bond yields and financing costs make AI infrastructure investments more expensive, particularly for companies that rely on borrowing or external funding. And Daniel O'Regan, managing director for equity trading at Mizuho, has pointed out something crucial: AI capital spending has outpaced free-cash-flow generation, meaning companies have needed to tap external financing for hardware purchases. If that financing becomes harder to come by, it could dent spending on chips and other forms of AI infrastructure.

In other words, the AI buildout isn't just a technology story. It's a credit story. And credit conditions are tightening.

The International Monetary Fund added its voice to the chorus of caution, warning in its latest financial stability report that "the consequences of an AI correction could be global," and noting there's growing evidence that AI capability investments and economic gains may not be as large as expected.

That's not a prediction of doom. It's a reminder that the AI trade has been priced for perfection, and perfection is a fragile thing.

So Where Does This Leave Us?

Let me be honest: I don't know whether this is a healthy correction in an ongoing bull market or the beginning of something more painful.

Neither does anyone else. The analysts who are pounding the table about AI demand being insatiable might be right. The bears who see a bubble ready to burst might be right. The honest answer is that nobody knows, and anyone who claims otherwise is selling something.

What I do know is that the dynamics have shifted. For the past two years, the market rewarded chip stocks on the promise of AI demand. Now it seems to want receipts. OpenAI's revenue figure, whether fairly interpreted or not, was enough to knock the SOX down 3.4% in a single day. That tells you something about how much faith was baked into these prices.

Twenty-two of the 30 SOX components sitting more than 20% below their highs isn't 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.

And maybe that's healthy. Maybe a little skepticism is exactly what a market that's been running on pure narrative needs. The chip stocks that survive this shakeout will be the ones with real earnings, real competitive moats, and real customer demand, not just a compelling story about the future.

As for that senior editor in the newsroom? He retired a few years ago, right before the AI boom really took off. Sometimes I wonder what he'd make of all this.

I suspect he'd pour himself a bourbon, shake his head, and say something like: "The party's not over. But the smart money's already checking where the exits are."