Artificial intelligence’s winner landscape is undergoing a structural, large-scale transformation. UBS Group’s research division, Holt, has released a new report stating that “AI infrastructure stocks,” centered on semiconductors and hardware, are poised to significantly surpass the traditional “Magnificent Seven” U.S. tech stocks in their ability to generate cash and economic value. This forecast not only reveals a staggering shift in profits along the AI supply chain but also suggests an unprecedented reshuffling of the leadership in the TMT (Technology, Media, and Telecommunications) sector, long monopolized by software and cloud giants.
According to observations by a team led by UBS Holt analyst John Talbott, the stock price gains of memory chip makers like Samsung Electronics, SK Hynix, and Micron Technology (MU) have recently far outpaced those of hyperscale cloud service providers such as Amazon (AMZN), Alphabet (GOOGL), and Meta (META), indicating a fundamental qualitative change in capital flows and industry profit models.
Memory Industry Leaps from “Value Destroyer” to Global Top Five
The UBS team used Cash Flow Return on Investment (CFROI), a metric that measures a company’s economic return relative to its cost of capital, for its analysis. The data shows that CFROI expectations for AI infrastructure companies are continuously accelerating. In contrast, CFROI expectations for large tech companies have declined by 200 basis points (2 percentage points) over the past two years, coinciding with significantly increased AI capital expenditure commitments.
The “AI infrastructure” group defined by Talbott encompasses U.S. tech hardware companies excluding Apple (AAPL), semiconductors, semiconductor equipment, and global semiconductor firms, with a combined market capitalization exceeding one trillion dollars. The economic profit for this group is projected to surge from approximately $200 billion in 2023 to $1.4 trillion by 2027, a staggering 600% increase.
In comparison, the economic profit for AI hyperscale cloud service providers is expected to increase only from about $200 billion to $400 billion by 2027, a significantly slower growth trajectory. Analysts specifically highlighted that memory stocks are expected to contribute half of the total new economic profit generated by the entire AI infrastructure segment.
Talbott marveled in the report: “The magnitude of this shift is extraordinary. Memory manufacturers were still loss-making in 2023, yet they are now projected to rank among the top five value creators in the global TMT sector in 2025, second only to Nvidia (NVDA).”
TMT Top Five Rankings Face Major Reshuffle
Looking back three years, the top five value creators in the TMT sector, as ranked by Holt’s economic profit framework, were Apple, Microsoft (MSFT), Alphabet, Meta, and Amazon. However, in the projected rankings for 2027, the top five have been completely reshuffled, led by Nvidia, Samsung, SK Hynix, Micron, and Alphabet. Among the traditional “Magnificent Seven,” only Alphabet barely clings to a spot on the list.
Talbott further analyzed that the rarity of a historically highly cyclical and commoditized memory industry leaping from value destroyer to one of the world’s most powerful value-creating sectors in just a few years is almost impossible to overstate. He emphasized: “Equally important, this underscores the rapid pace of this AI cycle. Leadership in value creation within the TMT sector is rotating at an unprecedented speed.”
South Korea’s Stock Market Validates Trend: Semiconductors Dominate, Capital Extremely Concentrated
This memory-driven AI wave is also being validated in South Korea’s stock market. Market sources indicate that the net profit growth rates for Samsung Electronics and SK Hynix this year are expected to reach 570% and 410%, respectively. However, excluding these two companies, the profit growth rate for the remaining South Korean firms is only about 64%. Looking ahead to next year, the profit growth rates for the two semiconductor giants are still expected to remain between 33% and 38%, while other companies are projected to see only an 18% increase.
Institutional analysts note that this extreme profit disparity has led to a high concentration of capital in semiconductor stocks. Between late May and late June, only 105 stocks in South Korea’s KOSPI and KOSDAQ markets advanced, accounting for less than 5% of the total, while a staggering 2,268 stocks declined, with an average drop of 27%. The market capitalization share of semiconductor stocks even briefly surged to 62%, causing the overall market trend to be almost entirely dictated by a single industry.
Analysts warn that while AI infrastructure and supply chain investments can partially offset weak consumer electronics demand, oil price and exchange rate volatility stemming from Middle East geopolitical risks could still pressure global inflation and interest rate policies. Investors optimistic about the AI infrastructure outlook still need to monitor the sustainability of cash flow and profit momentum in a high-interest-rate environment.
As investors grow increasingly concerned about whether the massive capital expenditures by tech giants will yield sufficient returns, ETFs tracking the Magnificent Seven have turned negative this year, with cumulative declines exceeding 1%. UBS’s report undoubtedly provides the market with a new perspective: the golden fruits of AI may be flowing from the “brain” of cloud computing to the “neurons” responsible for memory and transmission—memory and semiconductor infrastructure.