Bulls vs Bears on the AI Chip Selloff: A Normal Pullback or the Peak?

The Philadelphia Semiconductor Index sank 7.2% on Friday, NBC News and Reuters reported, after Intel (INTC) forecast weak sales and blamed a shortage of memory chips, and cash burn at Amazon (AMZN) revived worries about AI spending. The Nasdaq Composite had its biggest one-day drop since 5 June, Reuters reported.

The argument now is whether that was a normal pullback in a boom with years left to run, or the start of a reckoning for the spending behind it. At the center is one number: Goldman Sachs expects the five biggest US hyperscalers, the companies that run giant data centers, to put $1.2 trillion into AI infrastructure in 2027, up from about $800 billion this year.

The bull case

Keith Lerner, chief investment officer, Truist

Tech’s drop is a normal pullback in a bull market. “Tech’s current 10% decline does not appear abnormal relative to prior pullbacks during this bull market,” Lerner wrote in a note quoted on Yahoo Finance’s live blog on Friday, adding that the longer-term uptrend in tech remains intact. He didn’t call the bottom: Truist’s model points to roughly 2% to 8% more downside for tech from current levels.

David Goeckeler, chief executive, Sandisk

Memory demand will grow in PCs, phones and data centers next year. On Sandisk’s call on Thursday, the same evening Intel warned about PCs, Goeckeler said “PCs, smartphones, data center will all show positive growth next year,” because each device keeps getting more memory, according to The Motley Fool’s transcript and Yahoo Finance. Sandisk’s own data-center revenue rose 61% from the previous quarter, to $4.80 billion. Its stock plunged on Friday anyway.

Ryan Hammond’s team, Goldman Sachs

AI spending will rise by about half next year. A Goldman Sachs team led by Ryan Hammond forecast in September that the five biggest US hyperscalers will put about $1.2 trillion into AI infrastructure in 2027, above the roughly $1.1 trillion Wall Street expected, Bloomberg reported on 25 September, according to PYMNTS. The companies’ own words since then point the same way: Anat Ashkenazi, chief financial officer of Alphabet (GOOGL), said on 28 October that she expects its capital spending “to increase significantly in 2027,” as quoted by IBTimes.

The bear case

Jay Goldberg, analyst, Seaport Research Partners

Intel’s collapse is about trust in its management. Intel blamed a memory shortage for its weak forecast, but Goldberg told Reuters on Friday that the drop reflects “a loss of faith in management’s credibility,” and that bad planning was the cause. Intel plunged 28.6%, its worst day since 1974, and lost more than $146 billion of market value, Reuters reported.

Lee Min-hee, analyst, BNK Investment & Securities

The AI building boom may already have peaked. Lee told the Korea JoongAng Daily on 29 October that “the pace of AI infrastructure investment may have peaked in the third quarter of this year,” and that Big Tech will face weaker free cash flow and funding pressure. The filings show the squeeze: Alphabet’s free cash flow was negative $9.07 billion in the third quarter, and Meta (META) came in at negative $2.12 billion.

Ludovic Subran, chief investment officer, Allianz

The debt behind AI data centers was priced too tight for the risk. Subran told the Financial Times in mid-October that “data center bonds are having their Darwin moment, where the weakest credits struggle to survive,” according to EnterpriseAM’s report on the FT story. One CoreWeave (CRWV) unsecured bond fell below 80 cents on the dollar, pushing its yield toward 15%, the FT reported.

Where they agree

Expectations were extremely high, and Friday’s selling didn’t discriminate. “It’s more a question of what is priced into a stock and how much investors are expecting,” Art Hogan, chief market strategist at B. Riley Wealth, told Reuters, as did Matt Britzman, senior equity analyst at Hargreaves Lansdown: “The market is in a sell first, ask questions later mood right now.”

The number that decides it

The number is next year’s AI spending, and whether the companies paying for it can fund it from the cash they earn. The first read comes on Tuesday 3 November, when AMD (AMD) reports after the close, with a call at 5 p.m. ET; its forecast for data-center chips will show whether AI orders are still growing. Arm (ARM) and Qualcomm (QCOM) report on Wednesday 4 November, and both depend on phones, where pricier memory is squeezing phone makers.

Our read

We think Friday looks more like Lerner’s pullback than Lee’s peak, for now. The plans behind AI chip demand still point up, and several of the stocks hit hardest, AMD, Arm and Qualcomm among them, had no news of their own.

The bears win the argument about cash, though: Alphabet and Meta both spent more on capital projects than their operations brought in last quarter, so 2027’s budgets will lean harder on balance sheets and debt markets, where Subran sees the cracks. AMD’s report on Tuesday is the first test.