#美股AI概念股普涨 Historical Review: Significant changes in macro expectations have a notable impact on the short-term performance of U.S. stocks in the semiconductor hardware sector.

Our review finds that since 2024, the Philadelphia Semiconductor Index (SOX) has experienced four clear pullbacks, each lasting relatively short periods (between half a month and two months), including April 2024 (-15.2%, decline), July 2024 (-31.1%), February 2025 (-49.1%), and March 2026 (-18.6%). The primary trigger for these declines has been changes in macro expectations, with market concerns sequentially reflecting: stagflation, recession, recession, and stagflation.
Regarding the logical transmission between macro factors and U.S. semiconductor hardware, we believe one explanation is more plausible: since 2023, AI computing power has been the core driving force behind the stock price movements of U.S. semiconductor hardware companies, with major U.S. tech giants being the primary sources of funding for massive AI computing investments (accounting for over 50% of total). Currently, AI remains in its early development stage, with tech giants mainly relying on traditional businesses (internet, software, etc.) to provide ongoing financing support for AI investments.
In simple terms, the prosperity of AI computing power (i.e., U.S. semiconductor hardware) depends not only on technological and application progress within the AI industry itself but also on stable and ideal macroeconomic conditions as support. Recent strong employment data and tense Middle East conflicts are also significantly influencing market expectations for FED monetary policy, which in turn fuels concerns about the sustainability of AI capital expenditure (CAPEX).

AI Progress: The short-term industry narrative is nearly perfect, but there is still a noticeable distance from the claim of a “completely closed-loop commercialization.”
Since early 2026, aided by rapid penetration of AI Agents, exponential growth in Anthropic’s ARR (annual recurring revenue), and persistent tight supply and demand for AI computing power, the AI industry appears to be thriving. However, we also note that the current high prosperity of the AI industry is supported by factors such as: short-term curiosity among enterprises and individuals, exponential growth in AI token consumption driven by chatbot to AI Agent evolution, and price increases and business model shifts to token-based billing due to tight supply and demand for computing power.
In the short term, from upstream (semiconductor hardware) to downstream (cloud providers, model vendors, etc.), the entire industry chain is significantly affected by inflation caused by tight supply and demand for computing power. Some segments, such as storage chips, are generating excess profits that are difficult to explain with basic economic principles.
From a mid-term perspective, we still need to explore more high-value monetization scenarios beyond existing AI coding applications to match the massive upstream AI CAPEX investments. Economically, the token-based billing model is more of a transitional arrangement; ultimately, pricing should be linked to actual commercial output and utility.

Future Outlook: Expect continued high volatility, with close attention to the risk of phase mismatches between investment and output.
Since the dot-com bubble in 2000, over the past 20+ years, the rise and fall of global technological waves have been primarily driven by industry trends, but macro factors also play an important role.
In the short term, we believe the market will remain highly volatile due to: 1) persistently high yields on long-term U.S. bonds, which make the U.S. stock market itself highly unstable and significantly suppress risk appetite; 2) benefiting from tight supply and demand for computing power in the short term, the AI industry’s micro-level fundamentals are relatively flawless, and industry logic continues to reinforce itself positively. However, the current ultra-high gross margins of U.S. semiconductor and hardware companies depend on the sustained investment in global AI CAPEX.
We estimate that North America’s four major cloud providers will spend about $710 billion on CAPEX in 2026, roughly matching their operating cash flow, and they are raising more funds through debt and equity issuance. Such behavior may lead Wall Street to become more short-sighted and demanding. From a top-down perspective, the AI industry has very little room for error in the coming quarters.
While the long-term trend and commercial value of AI are almost unquestioned, the current crowded market, massive AI CAPEX, and financial pressures on tech giants pose risks of phase adjustments due to mismatched investment and output rhythms. Indicators such as token price data and tech giants’ bond CDS are currently favored for short-term monitoring.

Risk Factors:
Sticky inflation and runaway risks; AI technological progress falling short of expectations; risks of uncontrolled AI development; tech giants’ capital expenditure slowdown and contraction risks; geopolitical conflicts causing global supply chain disruptions and blockages; policy expectation chaos ahead of U.S. midterm elections, etc.

Investment Strategy:
Short-term adjustments in U.S. tech stocks are mainly driven by revisions in monetary policy expectations, previously crowded markets, and individual company noise.
In the short term, the optimistic logic of continuous self-reinforcement in the AI industry is unlikely to reverse, but the industry still has a clear distance from “fully closed-loop commercialization,” and more high-value monetization scenarios need to be developed.
Meanwhile, rising long-term bond yields and extremely limited market tolerance imply continued high volatility. It is crucial to closely monitor the phase mismatch risks between AI investments and outputs, with high-frequency indicators as a practical approach at this stage.