Is the AI Investment Trend Shifting? The 'Buy Chips, Sell Software' Strategy Begins to Unravel

The artificial intelligence (AI) investment frenzy has driven a sharply divided trading strategy in technology stocks this year, but signs of a shift are now emerging. As the market begins to reassess whether AI capital expenditures will deliver tangible returns and whether software companies will truly be disrupted by AI, the once-dominant 'buy chips, sell software' trade is cooling. Recently, software stocks have stabilized and rebounded, while semiconductor stocks have pulled back from record highs, suggesting investors are rebalancing their AI-related portfolios.

So far this year, the Philadelphia Semiconductor Index (PHLX Semiconductor Index) (SOX-US) has surged 78%, achieving its best quarterly performance in history, fueled by strong AI-driven demand. However, since July, the index has declined by 12%. In contrast, the index tracking software stocks rose 2.2% this month. Although it remains down 12% year-to-date, the pace of decline has clearly narrowed.

Bob Doll, Chief Investment Officer at Crossmark Global Investments, said that when market trends become overly extreme, it's time to think contrarily. He has recently begun gradually increasing exposure to software stocks while reducing semiconductor holdings. He believes the earlier market narrative—that software companies would be rendered obsolete by AI while chipmakers had struck a 'new gold mine'—had become excessively one-sided.

Market sentiment toward the software sector is turning more optimistic. Last week, Guggenheim upgraded its investment ratings for three software firms: Salesforce (CRM-US), ServiceNow (NOW-US), and Check Point Software (CHKP-US). The firm argued that while AI may indeed transform the industry, the widespread pessimism about software companies facing existential threats resembles more of a 'hallucination'.

Shortly after, HSBC raised its rating on Adobe (ADBE-US) from 'Hold' to 'Buy', asserting that the market had overestimated the negative impact of AI-powered design tools on Adobe's core business.

Bloomberg data shows the performance gap between semiconductor and software stocks has reached historically rare levels. Last month, semiconductor stocks recorded both the largest single-day outperformance and the largest single-day underperformance relative to software stocks. Additionally, the 40-day return correlation coefficient between the two sectors turned negative for the first time, indicating capital is flowing in opposite directions.

Meanwhile, the semiconductor sector is facing headwinds. Despite Samsung Electronics reporting strong earnings, its stock failed to rally. Moreover, a Reuters report revealed that Chinese AI startup DeepSeek is developing its own AI chips, sparking investor concerns over increased competition and potential oversupply in future AI chip demand, causing the semiconductor index to plunge 4.7% in a single session.

Market skepticism is also growing over whether AI infrastructure spending will continue expanding as expected. Earlier, Bloomberg reported that Meta (META-US) plans to rent out excess AI computing capacity—an announcement interpreted by investors as a warning sign of potential data center overcapacity, leading to a sell-off in AI infrastructure-related stocks.

Nat Schindler, an analyst at Scotiabank, noted that if major cloud providers continue investing in cutting-edge AI models while simultaneously monetizing surplus computing power, the market will inevitably reassess capital expenditure forecasts for the coming years. However, he emphasized that he does not believe AI capex has peaked, nor is there evidence that Meta is stepping back from the AI race.

These concerns have attracted prominent short-sellers. Michael Burry, known as 'the Big Short', recently established short positions in NVIDIA (NVDA-US), Applied Materials (AMAT-US), and the iShares Semiconductor ETF (SOXX-US), arguing that semiconductor ETFs have become rare, easily identifiable overvalued assets in the market.

Additionally, OpenAI, which had planned an IPO this fall, is reportedly considering delaying its public listing to 2025 due to heightened volatility in tech stocks.

Nonetheless, long-term confidence in the semiconductor industry's fundamentals remains strong. Bloomberg Intelligence forecasts that semiconductor industry profits will grow 47% by 2027, with valuation estimates being revised upward recently. In contrast, software and services sector earnings are projected to grow 16.5% this year, but market expectations continue to decline.

Market observers believe AI is redefining the competitive landscape for software companies, challenging traditional advantages like high margins and recurring revenue. Yet, what ultimately drives stock performance is whether firms can transform AI into new growth engines. Similarly, while the semiconductor industry remains fundamentally robust, the next major test will be whether current valuations have already fully priced in future growth.