U.S. stock market fragility: AI semiconductor bubble and economic structural risk

In the first half of 2026, the U.S. stock market continued its strong performance under Trump's second term, with the S&P 500 posting a total return of nearly 30% since the 2024 election and the Dow briefly breaking through the 50k-point mark. However, beneath this surface lies deep structural fragility: the economy is highly dependent on consumption driven by financial asset appreciation among high-net-worth groups, the semiconductor sector—especially AI-related chips—holds excessive weight in the index, and there is a potential gap in the realization of AI investment returns. The sharp volatility of South Korea's KOSPI index, price corrections in stocks like Micron, Meta's strategic pivot, and the weak June nonfarm payrolls data all point to a core issue—the risk of decoupling between the current stock market and the real economy is accumulating.

Wealth Concentration and the Fragile Balance of the Consumption Engine

The U.S. economy is notably consumption-driven, and consumption is highly concentrated. According to relevant data, the top 10% of households by income contribute about half of consumer spending, and households earning over $125k annually are the main drivers of current consumption growth. In these households' wealth structures, financial assets, especially stocks and related investments, account for a very high proportion. Since 2026, the financial assets of the top 1% have expanded significantly alongside the stock market rally, becoming a key driver of discretionary consumption.

The Trump administration has emphasized stock market performance and 401(k) returns, and the market has indeed posted considerable gains. But this implicit assumption of "everlasting asset price appreciation," once subject to correction, would directly transmit to consumption and overall economic growth. Historical experience shows that the reversal of wealth effects often accelerates economic downside pressure. The current economy is described as a "zombie economy" sustained by rich people's spending—exaggerated as it may be, it reflects the challenges to the real purchasing power of the broad middle- and low-income groups eroded by inflation. Official inflation data stands above 4%, while wage growth lags, putting pressure on real living standards.

Extreme Concentration in the Semiconductor Sector: Higher Weight Than the 2000 Tech Bubble

The semiconductor industry now accounts for about 20% of the S&P 500, higher than the tech sector's share at the peak of the 2000 internet bubble. This concentration is alarming, as the entire sector is highly tied to the AI narrative. Once demand expectations adjust, systemic risk will be significantly magnified.

The South Korean market, a global hub for memory chips, serves as a forward-looking signal with its volatility. In 2026, the KOSPI surged dramatically on the AI chip frenzy, but in June it saw a single-day crash of over 6%-10%, triggering program trading halts and evaporating massive market capitalization in a short time. Heavyweights like SK Hynix and Samsung led the decline, directly affecting U.S.-listed semiconductor companies. Investors locked in profits of 200% levels and exited, highlighting concerns over valuation bubbles.

Micron Technology serves as a typical case: its stock price fell nearly 16% within a few trading days recently, though it still posted substantial gains for the year. The company's fundamentals face a turning point: although demand for HBM (high-bandwidth memory) remains strong, price pressures and potential oversupply risks are emerging. Trump publicly praised Micron and mentioned the company's $50k investment initiatives as a boost signal, but market reaction was limited, showing that narrative-driven support alone is difficult to sustain valuations over the long term.

Easing AI Demand Narrative: Hyperscaler Awakening and Oversupply Risks

The core assumption behind the AI investment frenzy is that companies will achieve productivity leaps and revenue growth through massive capital expenditures. But in reality, several large tech companies are beginning to scrutinize returns. Meta Platforms announced plans to launch a cloud business, selling AI computing power to external customers—a strategic pivot seen as a signal of internal overcapacity. The company had previously invested heavily in building data centers and is now seeking to monetize them, reflecting a reassessment of returns on AI infrastructure investments.

Institutions like Goldman Sachs have warned that semiconductor valuations face three core risks: first, potential oversupply of HBM chips—rising prices stimulate more capacity, eventually leading to oversupply and price corrections; second, hyperscalers waking up from their "AI spending slumber," seeking actual returns rather than continuous expansion; and third, broader geopolitical and competitive pressures.

China's progress in open-source models, cheap energy, and low-cost computing further compresses the pricing power and market share of U.S. chip companies. Enterprise customers increasingly seek control over computing resources, models, and data, rather than relying on third-party "token" services—a trend accelerating the commoditization of AI.

Cracks in the Labor Market: AI Fails to Deliver Broader Dividends

In June 2026, U.S. nonfarm payrolls added only 57,000 jobs, far below expectations and revised prior data, while the unemployment rate edged down to 4.2%, mainly due to a decline in the labor force participation rate. Wage growth lags behind inflation, and traditional industries face high cost pressures. Industry leaders like Nvidia's CEO have noted that U.S. competitiveness in traditional sectors like automotive and manufacturing has declined, with AI infrastructure becoming the remaining growth engine. But AI spending currently mainly boosts profit margins for tech giants, with limited impact on employment and productivity in the broader economy.

Corporate AI investments often increase costs rather than significantly reduce expenses, with profit margins in most non-tech sectors remaining in the 9%-12% range for years. The 50%-60% gap between optimistic AI expectations and actual delivery, if it continues to widen, could force a reversal of the capital expenditure cycle, dragging down related industries.

Policy Narrative vs. Reality: Reindustrialization vs. AI Dependence

The Trump administration promotes reindustrialization and reshoring manufacturing, but high cost structures make traditional industries uncompetitive globally. Policy focus has shifted toward AI and technology, yet AI has failed to deliver the widely promised wage increases and job growth. Any correction in highly concentrated sectors like semiconductors could trigger a chain reaction, affecting AI infrastructure construction related to 2-3% of GDP.

Markets expect AI to drive significant corporate earnings growth, but the current path relies on sustained capital expenditures and narrative support. Once hyperscalers collectively shift to selling rather than consuming computing resources, excess capacity will lead to price collapse, marginal incentives disappearing, and transmission throughout the ecosystem.

Outlook: Seeking a Sustainable Growth Path

In 2026, the U.S. economy faces multiple challenges: dependence on asset prices, sector concentration risk, and broad-based growth weakness. In the short term, policy support and corporate earnings may sustain stock market resilience, but long-term sustainability depends on whether AI can truly translate into productivity dividends rather than remaining a cycle of capital expenditure. The warning signals from South Korea's market, Meta's pivot, and weak employment data all remind investors to be wary of valuation re-rating risks.

Policymakers and market participants need to focus on supply-side reforms, cost control, and diversified growth engines, rather than relying solely on asset bubbles and single-technology narratives. Only by achieving more inclusive economic growth can a foundation for long-term prosperity be built.