Main investment themes in U.S. stocks for 2026: Will AI infrastructure or semiconductors lead this round of valuation restructuring?

On June 1, 2026, the S&P 500 closed at 7,580.06 points, and the Nasdaq Composite closed at 26,972.62 points—both at record highs. The S&P 500 has gained a total of 10.7% since the beginning of the year, while the Nasdaq 100 has risen more than 20% over the same period. However, the structure of this upswing is fundamentally different from several previous bull markets. According to calculations by a Citigroup strategist team, the increase in US stock index levels in 2026 is “almost entirely” driven by a small number of mega-cap stocks. The technology sector’s weighting in the S&P 500 has climbed to about 37%; if Alphabet, Meta, Amazon, and Tesla are included, that proportion exceeds 50%. The semiconductor industry’s weighting in the S&P 500 reaches 18%, double that of its peak during the internet bubble.

This extreme concentration is not accidental. Global capital is actively narrowing its choices, concentrating allocation focus on a few companies that have clear-cut AI revenue sources, high gross profit margins, and the ability to make sustainable capital expenditures. For index funds and passive investors, this structure means that the implicit risks in holdings are amplified—when a small number of stocks determine the direction of the entire market, any negative developments targeting these names could trigger a systemic pullback. At the same time, about 5% of the S&P 500 constituents have fallen to new 52-week lows; historically, similar scenarios occurred in July 1929, January 1973, and December 1999, all of them on the eve of major market turning points.

From the standpoint of industry structure changes, the essence of this rally is not a broad-based bull run, but an “AI asset revaluation bull.” Pricing power in the market is shifting from traditional consumer and financial sectors to semiconductor and cloud computing infrastructure. Once this shift in power takes hold, it is difficult to reverse in the short term because it is underpinned by long-term capital expenditure commitments on the scale of hundreds of billions of dollars, rather than short-term theme-driven speculation.

What are the drivers behind storage-chip giants collectively entering the trillion-dollar club?

On May 26, 2026, Micron Technology’s stock surged 19.29% in a single day, closing at $895.88 per share, and its market value surpassed $1 trillion. In the same week, SK Hynix’s market capitalization also crossed the $1 trillion mark, forming, together with Samsung Electronics, the “trillion-dollar three giants” in the storage-chip space. The core significance of this event is not just a single company’s market-cap milestone; it lies in a fundamental shift in the market’s valuation framework for the entire storage industry.

Micron’s revenue in the second quarter of fiscal 2026 reached $23.86 billion, up 196% year over year, while net profit jumped 770.8% year over year. More importantly, UBS sharply raised Micron’s target price from $535 to $1,625 and switched its valuation benchmark from price-to-book to price-to-earnings. This indicates that Wall Street no longer treats Micron as a cyclical supplier of bulk commodities; instead, it has redefined Micron as a strategic asset supplier for the AI era. The traditional valuation logic for storage chips centers on supply-demand cycles and inventory levels, whereas the new logic focuses on irreversible demand growth driven by AI training and inference.

The industry structure is changing: in the past, storage chips were viewed as standardized, interchangeable, low–gross margin hardware; now, as HBM (high-bandwidth memory) becomes the core bottleneck for AI accelerator cards, the pricing power of storage manufacturers is strengthening significantly. Take Micron as an example: in 2026, its HBM product capacity has already been locked in by major AI chipmakers through supply agreements extending to the end of 2027. Such a long-term supply agreement is extremely rare in the history of the storage industry. Looking forward, if AI model parameter counts continue to grow at a rate of 10x per year, demand for storage bandwidth and capacity will far exceed current market expectations—suggesting that the valuation reshaping of storage chips may still be in an early stage.

How will $600 billion in AI capital expenditures reshape the industry chain’s path of earnings growth?

In 2026, the combined capital expenditures of four tech giants—Amazon, Google, Microsoft, and Meta—in the data center and AI chip fields total about $660 billion. Using Morgan Stanley’s definition, AI capital expenditure is expected to surpass $1.1 trillion in 2027. This scale of spending is having a profound impact across the entire industry chain.

Take Amazon as an example: in the first quarter of 2026, AWS’s cloud computing business revenue reached $37.6 billion, up 28% year over year, the fastest growth rate in 15 quarters. AWS’s annualized AI revenue has exceeded $15 billion over the past three years, expanding the scale by nearly 260 times compared with the early stage of AWS. Amazon has received revenue commitments of more than $225 billion from its self-developed chip, Trainium, and its advertising business revenue in the past 12 months also exceeded $70 billion. Driven by this, Amazon’s stock jumped 27% in April alone, recording its best single-month performance since 2007, and its market capitalization neared $3 trillion.

This round of capital expenditures is forming a positive feedback loop: higher capital expenditures lead to more advanced computing infrastructure; the infrastructure advantage translates into stronger AI service capabilities, which then attracts more customers and revenue; revenue growth, in turn, supports the next round of capital expenditures. For upstream suppliers in the industry chain—chip design, wafer manufacturing, advanced packaging, and thermal solution providers—this means demand visibility and durability far surpass any previous hardware cycle. At the same time, it’s important to note that this positive feedback is built on high profit margins and high capital barriers, making it nearly impossible for small and medium-sized enterprises to compete; industry concentration will rise further over the next two years.

Within leading assets, has a divergence already emerged between performance and valuation?

Not all leading technology stocks are on the same growth track. Nvidia reported revenue of $81.6 billion in the first quarter of 2026, and non-GAAP earnings per share of $1.87—both above market expectations. However, after the earnings release, the stock price fell by as much as 1.6% in after-hours trading. The market focus has shifted from “whether it beat expectations” to “whether it can maintain a 75% gross margin on a high base,” as well as how quickly cloud computing giants’ self-developed chips are eroding Nvidia’s market share. This shows that once a company’s market value reaches several trillion dollars, its valuation becomes highly sensitive to signals that fall below extremely optimistic expectations.

Apple, by contrast, shows a different growth rhythm. In the second fiscal quarter of 2026, Apple’s revenue was $111.18 billion, up 16.6% year over year. The iPhone 17 series drove iPhone revenue to a record high. The company also approved a $100 billion share repurchase plan and increased its dividend payouts. Apple’s growth drivers come from the stability of its product ecosystem and the continued growth of service revenue, rather than an AI-driven valuation reconfiguration. Its challenges include supply-chain cost pressure caused by storage-chip shortages and the extension of consumer electronics replacement cycles.

This sector divergence leads to the following assessment: the valuations of AI direct beneficiaries (such as Nvidia and Micron) have already priced in high-growth expectations for the next two to three years, so any marginal slowdown in demand or intensifying competition could trigger valuation contraction. Meanwhile, AI indirectly benefiting stocks (such as Apple and Amazon’s retail business) have valuations that are relatively reasonable, but with less growth elasticity. The market may shift from “rising across the board” to “rotation.” Capital will be redistributed among different sub-sectors rather than simply flowing out of technology stocks.

Market concentration and macro risk: two key variables for the second half of 2026

Market concentration itself is evolving into a risk factor. Goldman Sachs’s strategy team warned that the current rally in US stocks is highly concentrated in a handful of mega-cap technology stocks, and market breadth has fallen to the lowest level since the time of the internet bubble. Citadel Securities pointed out that the fund-driven momentum behind the recent surge in US stocks is gradually weakening, and the risk of a short-term pullback is rising. Goldman Sachs also observed that the US options market’s put/call ratio is in a range near historical extremes, and the asset management size of single-stock leveraged ETFs with 2x leverage or more has risen sharply. When market breadth keeps narrowing and leveraged funds are highly concentrated in a small number of underlying names, once concentrated positions loosen, the speed of a pullback may far exceed that of a normal adjustment.

At the macro level, the Federal Reserve’s interest-rate path remains the core variable. In April, the PCE inflation rate rose 3.8% year over year, reaching a new high since 2023 and far above the Federal Reserve’s 2% policy target. Cleveland Fed Chair Mester has clearly stated that “current inflation risks are significantly greater than employment risks.” Market expectations for rate cuts in 2026 have been substantially lowered: the Dutch bank ING Bank expects the first rate cut to be delayed until October 2026, while the second rate cut would have to wait until January 2027—an obvious contrast with the three to four cuts expected at the beginning of the year. Ongoing geopolitical conflicts in the Middle East continue to push energy prices higher. Brent crude oil futures have remained above 90 dollars, far above the level of about 70 dollars prior to the outbreak of the conflicts.

These two layers of risk compound each other. In a high-interest-rate environment, the discount factor for high-valuation technology stocks increases, putting pressure on their current valuations. Energy prices running at high levels further raise inflation readings and weaken the urgency of Federal Reserve rate cuts. For investors, this means the market path in the second half of 2026 will be more complex than in the first half: the long-term logic of AI has not broken, but short-term volatility could be significantly amplified. This requires more precise management of position concentration and sensitivity to changes in interest rates.

FAQ

Why, despite new highs in US stock indexes in 2026, have most stocks not risen?

The rally in US stocks in 2026 has been highly concentrated in a small number of AI-related mega-cap technology stocks. The technology sector’s weighting in the S&P 500 exceeds 37%, semiconductors account for 18%, and almost the entire index gain comes from these leading companies—leading to about 5% of constituents falling to new 52-week lows, with market breadth at historic lows.

Why has Micron been able to break through a $1 trillion market capitalization in 2026?

Micron has benefited from the explosive AI-driven demand for HBM high-bandwidth memory. In the second quarter of fiscal 2026, revenue rose 196% year over year, net profit surged 770.8%, and Wall Street shifted its valuation framework from cyclical hardware to an AI growth-stock model. UBS also sharply raised its target price to $1,625.

What does $660 billion in AI capital expenditures mean for the industry chain?

The roughly $660 billion in annual AI capital expenditures combined by Amazon, Google, Microsoft, and Meta is forming a positive feedback loop: infrastructure advantages translate into revenue growth, revenue growth supports higher spending, and this provides unprecedented demand visibility to upstream suppliers such as chip design and advanced packaging.

How do Nvidia’s and Apple’s growth logics differ in 2026?

Nvidia is a direct beneficiary of AI; its valuation has already priced in high-growth expectations for the next two to three years, and the market focuses on gross margins and competitive pressure. Apple is an ecosystem- and services-driven growth story; AI affects it indirectly. Its valuation is relatively reasonable but has smaller upside elasticity, and it faces challenges from supply-chain costs and the extension of upgrade cycles.

Why is market concentration considered a risk factor for the second half of 2026?

When an extremely small number of stocks determine the direction of the entire market and leveraged capital is highly concentrated in these underlying names, any negative catalyst can trigger a pullback faster than that of a typical adjustment. The current put/call ratio in options and the structure of leveraged ETF positioning are similar to those seen before the 1929, 1973, and 1999 market turning points.

How does the lowering of rate-cut expectations by the Federal Reserve affect US tech stocks?

In April, the PCE inflation rate reached 3.8%. The market has reduced its expectations for 2026 rate cuts from three to four times down to possibly only once (October). In a high-interest-rate environment, the discount factor for high-valuation tech stocks rises, directly pressuring their current valuations. At the same time, elevated energy prices further limit the space for rate cuts.

What are the main variables to watch for US stocks in the second half of 2026?

There are two key variables: first, whether the degree of market concentration will trigger liquidity-driven pullbacks; second, whether the Federal Reserve can provide clear signals about the rate-cut path while inflation remains above target. The interaction of these factors will determine whether the AI asset revaluation logic can continue to outperform the broader market.

In the long run, is AI-driven valuation reconfiguration sustainable?

From an industry trend perspective, the pattern of AI model parameter counts doubling every 10 months has not slowed, and demand for storage bandwidth and computing power continues to grow exponentially. This means the fundamental support for the revaluation of hardware assets remains, but short-term valuations have already priced in part of those expectations. Sustained performance will be required to absorb the remainder.