The memory sector is cleaning up.
The first half of 2026 for U.S. stocks officially concluded on June 30. Looking at the performance of technology stocks, the biggest winners are no longer the 'Magnificent Seven' like Microsoft, Apple, and Nvidia.
Data shows the Philadelphia Semiconductor Index (SOX) surged over 80% in the second quarter, accumulating gains of more than 100% for the first half, with numerous chip and infrastructure companies reaching record highs. In the same period, the Nasdaq Composite Index rose approximately 12.8%, and the S&P 500 gained about 9.55%. Some tech giants that once represented the AI trend have begun to slow down noticeably, lagging behind the broader semiconductor sector.
The logic behind the market's rise has not strayed from AI. In recent years, capital was more concerned with 'who possesses the most advanced large model, holds the greatest value in the AI era.' However, starting this year, the market is increasingly focusing on another question: who is actually making money from AI? Investors are shifting from large-cap tech stocks towards memory chips, semiconductor equipment, advanced manufacturing, and AI infrastructure suppliers.
Memory Sector's Dominant Performance
Judging by gains, the biggest winners among U.S. tech stocks in the first half of the year almost all came from the semiconductor supply chain, especially the memory segment.
As of June 30, nearly all of the top ten performers in the S&P 500 by gain were from the information technology sector. Among them, Sandisk's stock price rose over 850% in six months, becoming the best-performing stock in the S&P 500. Micron Technology (NASDAQ: MU) surged over 300%, its market value exceeding $1 trillion for the first time and surpassing Berkshire Hathaway to enter the top ten U.S. stocks by market cap. Western Digital, Intel, Seagate Technology, and Marvell Technology, which was mentioned by Jensen Huang, all saw gains exceeding 200%.
It's not just memory. Taiwan Semiconductor Manufacturing Company (TSMC) rose about 57%, while lithography giant ASML Holding (ASML) gained approximately 86%. The entire AI industrial chain—from GPUs, HBM (High Bandwidth Memory), and advanced packaging, to wafer fabrication and equipment manufacturers—is beginning to share in the AI dividend.
In contrast, past AI star companies showed tepid performance. Although Nvidia continued to hit record highs, its cumulative gain for the first half was only about 7%, ranking at the bottom among Philadelphia Semiconductor Index components. Gains for Qualcomm and Broadcom lagged behind the overall sector.
This round of gains stems from the further explosion in demand for AI infrastructure.
Over the past year, large model parameter sizes have continued to grow, AI agents have begun entering enterprises, applications like video generation, robot training, and multimodal reasoning are rapidly being deployed, and the computing power required for training and inference continues to climb. GPUs remain the most expensive component of the entire system but are no longer the sole beneficiary.
To allow GPUs to perform at their full potential, AI servers require more HBM, higher-capacity DDR, faster SSDs, and more complex advanced packaging. The value chain of AI servers is extending outward, leading to simultaneous prosperity across the entire supply chain.
Meanwhile, the memory sector, a previously unremarkable niche, has suddenly stepped into the spotlight. For a long time, memory chips have been viewed as a typical cyclical industry. With intense price volatility, an industry supply prone to imbalance, and profitability highly dependent on the business cycle, memory companies historically struggled to achieve the high valuations seen in software or internet platforms.
But AI has brought change. Demand for HBM, high-capacity DRAM, and enterprise SSDs from AI servers is growing almost exponentially, while supply struggles to expand rapidly in the short term. This supply-demand imbalance is driving sustained price increases for memory, directly leading to significant improvements in corporate profitability.
This year, Micron has repeatedly raised its performance guidance, with enterprise storage products in short supply. Micron executives anticipate that tight memory supply conditions will persist beyond 2027, with industry supply gradually improving only in 2028. Companies like Sandisk and Western Digital are also benefiting from surging demand from enterprise data centers, with profit expectations being continuously revised upward. Analysts widely believe the memory industry is gradually evolving from a past cyclical stock into a crucial infrastructure asset for the AI era.
Looking at the overall picture, regardless of whether Microsoft, Amazon, Google, or Meta ultimately wins the AI race, building data centers, procuring servers, upgrading networks, and expanding storage are inevitable. Companies providing 'picks and shovels'—such as those in GPUs, HBM, wafer fabrication, advanced packaging, and the server supply chain—have become the most direct and certain beneficiaries in the AI investment cycle.
This repricing has become the most significant change in U.S. tech stocks during the first half of 2026.
AI Enters the 'Accounting' Phase
If the upstream supply chain is experiencing a value reassessment, then on the other side, past AI beneficiaries—tech giants like Microsoft, Meta, and Alphabet—are now facing a new test from the capital markets: as AI investment enters the 'cash-burning' phase, when will it truly start generating profits?
Judging by first-half performance, among the Magnificent Seven, Microsoft led the declines, falling over 22%. Tesla and Meta also saw declines to varying degrees. On the gainers' side, Google rose 14% cumulatively, followed by Nvidia, Apple, and Amazon.
Particularly entering June, the seven giants collectively lost about $2.3 trillion in market value in a single month, marking their worst monthly performance in a year.
According to the latest capital expenditure plans from these companies, Microsoft, Alphabet (Google's parent), Amazon, and Meta will continue to maintain historically high levels of AI investment this year, with combined annual capital expenditure expected to exceed $700 billion, the vast majority directed towards data centers, AI chips, and cloud infrastructure.
This round of AI investment scale surpasses any previous cloud computing infrastructure expansion cycle, also meaning large tech companies' free cash flow will remain under pressure for some time.
Taking Microsoft, the leader in declines, as an example, the company briefly entered the $4 trillion club in late October last year. However, as of June 30 this year, Microsoft's market value was only $2.77 trillion, representing a significant contraction, with its stock price falling nearly 20% in one month. The market is beginning to worry that rising data center investments, chip procurement, and power costs may mean the timeline for AI commercialization will be longer than expected.
Previously, Microsoft's CFO forecast the company's full-year 2026 capital expenditure would reach $190 billion, a 61% increase compared to 2025, with expected impacts of $25 billion from component price increases. The market is concerned whether growth in Microsoft's AI and cloud computing businesses can offset the rising infrastructure costs.
Among the tech giants, Apple faces different pressures. Although iPhone sales remain stable and the services business continues to contribute steady profits, it has yet to introduce an AI product that can truly alter its growth trajectory. As memory chip prices continue to rise, component costs for products like Mac and iPad are increasing simultaneously, making Apple a passive responder. The company stated it has 'never seen component prices rise with such magnitude and speed,' and has been forced to raise prices. Similarly, Microsoft's Xbox recently announced price increases.
In fact, the gains in tech stocks this first half are not simply a shift from internet platforms to semiconductors. Instead, the new profits created by AI are being redistributed upstream along the supply chain. In the past, the first to receive valuation premiums were internet giants with models, platforms, and user access points. This year, the first to deliver on performance are infrastructure suppliers like memory chips, advanced packaging, wafer fabrication, networking equipment, and servers. As AI data center construction continues to heat up, these 'picks and shovels' companies are the first to convert orders into revenue and profit, thereby receiving higher valuation premiums from the capital markets.
An industry analyst noted that under the dominance of the major AI computing power cycle, a structural reshaping of industry and capital pricing has begun. The demand from AI for computing power, storage, and software platforms is sustainable, and technological iteration will continue to open up growth space for the industry.
For the second half of the year, the market's focus is not only on how much the performance of the next generation of large models improves and the scale of larger capital expenditures, but also on who can first prove that AI investments can be consistently converted into revenue, profit, and free cash flow. The upcoming second-quarter earnings season will be a key test for this AI investment thesis.
Comments