Japan’s Nikkei 225 index sank 4% as shares of memory chipmaker Kioxia halved in value over the course of a month, exposing cracks in the artificial intelligence trade that had driven much of the market’s recent gains, according to market analysts.
Kioxia, which went public earlier this year to strong demand from investors seeking AI-related exposure, saw its stock price decline by approximately 50% from its peak as the company reported weaker-than-expected demand for NAND flash memory used in AI data centers. The sell-off in Kioxia shares rippled through the broader Japanese equity market, dragging down other semiconductor and technology stocks.
The Kioxia downturn highlights the growing divergence between AI hype and real-world revenue generation in the semiconductor space. While companies producing high-end graphics processing units for AI training have seen explosive growth, firms in adjacent segments such as memory and storage have not benefited as much from the AI spending boom.
The Nikkei’s decline also reflects broader concerns about the sustainability of the AI-driven rally in Japanese equities. Foreign investors had poured significant capital into Japanese tech stocks over the past 18 months, betting on the country’s semiconductor supply chain and AI infrastructure plays. The Kioxia sell-off suggests that some of those bets may be unwinding.
Japanese technology stocks had been among the best performers globally in 2025, but the tide has shifted in 2026 as investors reassess valuations and earnings prospects. The Bank of Japan’s gradual monetary tightening has added another layer of pressure on equity valuations by pushing up bond yields and strengthening the yen.
Analysts caution that the AI trade may face further headwinds as companies across the supply chain report earnings that fail to meet inflated expectations. The Kioxia situation serves as a cautionary tale about the risks of assuming that all companies with AI exposure will benefit equally from the technology’s adoption.
This article was adapted from BeInCrypto. Read the original here.
