Investors keep selling chip stocks. The iShares Semiconductor ETF (SOXX) fell 11.8% last week. It has now declined for nine straight days, its longest losing streak since 2006.
The damage was worse among the companies that build chipmaking machines and memory. Applied Materials Inc. (AMAT) fell 19.3%, Lam Research Corp. (LRCX) dropped 17.2% and Micron Technology Inc. (MU) lost 14.6%.
The fear behind the selling had a simple shape. The artificial intelligence boom might be running out of buyers.
Online, that fear had already found its own words. Searches for "companies reducing artificial intelligence spending" surged on Google Trends, after a wave of reports on corporations cutting back on AI tools.
If companies spend less on AI, the reasoning goes, chipmakers will sell less.
Yet research notes from Wall Street analysts in the past few days say the opposite.
Bank Of America Sees A Cycle Priced For A Crash
Bank of America analyst Vivek Arya said in a Friday note that the chip sector already prices in a collapse he does not expect.
AI leaders Nvidia Corp. (NVDA), Broadcom Inc. (AVGO) and Micron trade at about 13 times forward earnings on average, according to Arya.
That multiple compares a stock's price with the profit analysts expect over the next 12 months.
It sits within the range Cisco Systems Inc. (CSCO) traded at after the dot-com bubble burst.
"Investors are paying post-crash multiples for a cycle that hasn't crashed," Arya said.
The comparison with 2000 also breaks down on usage.
Back then, telecom companies had laid fiber-optic cable that sat idle. Today's AI infrastructure runs near full utilization, according to Arya.
Prices back that up.
Rental rates for Nvidia's older graphics processors, launched in 2020, 2022 and 2024, are up between 12% and 37% this year.
For Arya, that points to scarcity rather than oversupply.
Bank of America also estimates capital spending by the large cloud providers and specialist AI clouds at about $1.01 trillion in 2026, up 98% from last year. It sees about $1.39 trillion in 2027.
Cantor's C.J. Muse Says Equipment Is Sold Out
On Monday, Cantor Fitzgerald analyst C.J. Muse published his team's takeaways from the SemiconWest trade show in San Francisco.
The first point directly addressed last week's fear.
The analysts said chip equipment "is effectively sold out," with customers giving at least eight quarters of visibility.
ASML Holding NV (ASML) is receiving purchase orders for 2029 and 2030, according to the note.
Applied Materials is tracking 110 new chip plants.
Micron expects memory supply to be even tighter in 2027 than in 2026, with customers holding little to no inventory.
Muse also described frustration among management teams. Most are operating in the strongest backdrop they have ever seen, while their share prices fall, according to the note.
"We continue to recommend remaining Overweight Semis, as we view today's AI Infrastructure buildout as durable and elongated," the analyst said.
Why Companies Reducing AI Spending Is The Wrong Signal
Oxford Economics raised its outlook for global business technology spending growth in 2026 to 14.8%, from 11.4%, in an Oct. 6 report.
Spending on AI-related products is set to top $510 billion this year, up 54%, as firms move from experimentation towards operational deployment, according to the firm.
"Global AI-related spending is projected to reach $3.4 trillion by 2035, or 23% of tech spend," Oxford Economics said.
The Real Risk Ahead Is AI Monetization
None of this makes the AI trade risk-free. Bank of America's equity and quant strategy team led by Savita Subramanian pointed to a different problem in a separate Friday note.
AI capital spending has beaten forecasts by about 50% this year, while AI revenue has beaten them by only about 20%, according to the strategists. Meanwhile, the cost of capital for the largest cloud providers has risen to about 10%, up from 4% in 2023.
In other words, every dollar of data-center spending now has to earn more to pay for itself.
For the strategists, "monetization vs. cost of capital is the only question that matters."
