A warning light has come on for the systematic investing in Nasdaq 100 (QQQ) that had long been the "sure-win formula" for Korean retail investors trading U.S. stocks. As the center of gravity in the AI industry shifts from software to hardware, the big-cap big techs that led the index are losing steam.
According to Investing.com on the 27th, the Nasdaq 100 index has risen only 1.4% this year. The explosive momentum of the past is gone, and it is moving at a snail's pace. The reason is clear. The so-called "software leaders" with heavy weightings in QQQ, such as Microsoft (-3.65%) and Meta (-0.2%), are struggling.
By contrast, semiconductor and memory stocks directly tied to AI infrastructure are strong. iShares Semiconductor ETF (SOXX), a representative U.S. semiconductor industry exchange-traded fund (ETF), rose 14.46% over the same period. SOXX includes Nvidia, Micron, AMD and Qualcomm as core holdings.
iShares Expanded Tech-Software Sector ETF (IGV), which reflects the broader software industry, fell 7.55% over the same period. Even within tech, the return gap is widening.
The analysis that the AI industry's leadership has fully shifted from software to hardware is being borne out by ETF returns. As AI moves from a training focus to an "inference" focus, demand for memory—which determines computing efficiency and data throughput—is surging. In this process, memory manufacturers are enjoying a supplier-advantaged environment, while big techs that must buy in bulk are facing growing expense burdens.
In fact, Microsoft and Google are pouring astronomical sums into securing data centers and power grids to avoid falling behind in the AI race. The Bank for International Settlements (BIS) said they are in a position where they must spend more on building infrastructure than the cash they generate.
According to Hana Securities, this year's capital expenditure (CAPEX) growth rate for Alphabet, Amazon, Microsoft and Meta is estimated at 41%, 21%, 34% and 60%, respectively, from a year earlier. The four corporations' combined capital spending forecast for this year is $497.8 billion, up 36% year over year. That exceeds the 30.6% compound annual growth rate (CAGR) forecast for AI infrastructure investment cited by major research institutions.
By contrast, memory manufacturing corporations are assessed to have structurally entered a beneficiary phase. The spread of AI services is generating unprecedented memory demand, but supply is not keeping up. A structure has formed in which an increase in shipments (Q) translates directly into improved results.
Kim Dong-Won, head of research at KB Securities, said, "With short-term capacity additions in memory supply realistically limited in 2026–2027, the supply shortage is more likely to intensify," adding, "For example, Samsung Electronics' NAND wafer production capacity this year is forecast to decrease 5% from a year earlier, so memory prices are expected to enter an elastic upward phase."
This kind of industrial structure shift is being reflected directly in the Nasdaq 100 index. Big-cap software big techs that dominate market capitalization are weak on concerns that profitability will deteriorate, while semiconductor and infrastructure stocks in the lower ranks by market cap are strong. However, because lower-cap stocks have small weights, their gains are not offsetting the weakness of large caps.
Experts note that mechanically dollar-cost averaging into the Nasdaq 100 as before is unlikely to deliver the same results as in the past. A global strategy manager at an asset manager said, "Considering U.S. Starlink and global sovereign AI projects, capital expenditures centered on data centers are likely to continue expanding through 2032," adding, "In the short term, index gains could concentrate in hardware areas where profitability is becoming visible."
The manager added, "We are taking a more conservative approach than in the past to a strategy of mechanically buying indices with heavy software weightings."
Lee Kyung-min, a researcher at Daishin Securities, said, "It is important to confirm the real demand in the AI industry and the data center investment cycle through major big tech earnings such as Meta, Microsoft and Apple this week," adding, "The impact of last year's surge in memory prices on big tech profitability is also a key variable." Lee added, "Whether AI corporations' expansion in investment expenditure can actually connect to monetization will determine expectations for the industry's long-term growth."