What's Happening to AI Stocks? Profit-Taking Surge Sends Nearly 70% of Tech Stocks into Bear Market

Artificial intelligence (AI) stocks, which have driven U.S. market gains over the past several months, are now beginning to drag down technology sector performance: 69% of the constituents in the S&P 500 Information Technology Index have fallen more than 20% from their 52-week intraday highs, slipping one by one into technical bear markets. Is this merely a short-term correction, or the beginning of a broader reset?

Large semiconductor stocks have been hit particularly hard in this correction. Micron (MU-US) shares have dropped 25% from recent highs, Broadcom (AVGO-US) is down 21%, and Marvell Technology (MRVL-US) has declined by as much as 30%.

Despite the significant pullback, some analysts view this as a natural progression following a strong prior rally: the Philadelphia Semiconductor Index posted a historic gain in Q2, leading investors to take profits at the start of Q3 in July.

William Kerwin, a stock analyst at Morningstar, said: "At the beginning of Q3 in July, there was broad profit-taking across the tech infrastructure sector—for example, chip equipment stocks faced sharp selling on Tuesday."

He points to a recurring pattern over past quarters: tech stocks often come under pressure in the month following earnings announcements, but rebound ahead of the next earnings release.

Memory and storage chip stocks delivered a strong performance in the first half of 2026, but after peaking at the end of June, they have faced severe selling over the past two weeks. Seagate (STX-US), Western Digital (WDC-US), Sandisk (SNDK-US), and Micron have all fallen more than 20% from their highs. This group of historically highly cyclical companies is now at the center of a market debate over whether the AI boom can structurally and permanently boost demand for their products.

Kerwin notes that Samsung Electronics' preliminary earnings released earlier this week acted as a trigger for the recent sell-off. Despite Samsung's operating profit surging to 19 times that of the same period last year, some investors clearly had even higher expectations. "This could signal that memory price increases for players like Micron may slow down."

Although spot prices for DRAM and NAND have recently shown signs of moderating, Evercore analyst Amit Daryanani believes long-term supply agreements and contract pricing with major cloud service providers (hyperscalers) better reflect the true state of the memory market. He firmly believes long-term customer contracts will fundamentally enhance revenue visibility and predictability for these companies.

Daryanani said in a Wednesday report: "We believe memory remains an attractive segment within the tech ecosystem despite short-term volatility. This correction includes a reasonable element of profit-taking, as investors rebalance winning positions ahead of the next earnings cycle to reassess price sustainability and the capital spending pace of cloud giants."