#美股AI概念股普涨 Do stock price rises and falls depend on "AI content"? Understanding this logic is more critical

Recently, a fascinating phenomenon has appeared in global stock markets: the economies with stronger AI industry chains tend to have better-performing major stock indices.

Wind data shows that as of May 29, this year, since the conflict between the US and Iran in March, the Philadelphia Semiconductor Index has risen by 58.4%, the Korea Composite Index by 35.7%, Taiwan Weighted Index and the ChiNext Index of A-shares have also performed well, while European and other emerging markets with relatively lower AI content have lagged significantly in gains.

Does this mean that "AI content" is the "code" for stock price movements?
There is a strong positive correlation between the two, but if we broaden our perspective, things are not that simple. Stock performance is influenced by multiple factors, including valuation levels, capital flows, industry cycle stages, corporate profitability, etc., all of which impact the final outcome. "AI content" is very important, but it is not everything.
Nevertheless, it cannot be denied that AI has already become one of the most widely recognized and continuously pursued investment directions in the current global tech industry.
Whether it’s the expansion of computing infrastructure or the gradual implementation of applications, AI is moving from a "concept" to "performance," from "expectation" to "realization." For investors, understanding the development stages and intrinsic logic of the AI industry is more meaningful than simply chasing the "AI content" label.
So, how can we more comprehensively understand the current AI investment opportunities?

01 AI industry chain is a highly certain and prosperous sector
In recent years, the AI industry chain has become a key core in global asset pricing, with related company stock prices soaring. Some worry that the AI sector might be too "crowded."
CICC believes that AI may not have reached a stage where concerns about a bust are warranted. Currently, the global AI industry chain is highly interconnected, and US AI leaders serve as industry barometers. Their trends often propagate through the industry chain and reflect in A-shares. Taking the US AI industry as an example:
First, AI is genuinely boosting productivity.
In the first three quarters of 2025, US GDP grew by 2.51% year-over-year, with the St. Louis Fed estimating that AI contributed about 0.97 percentage points, accounting for roughly 39%, higher than the 28% contribution during the internet revolution in 2000. More importantly, the commercialization of AI agents at the enterprise level is accelerating, with overseas top AI vendors reaching annual revenues in the hundreds of billions of dollars. Computing power investment is shifting from "cost consumption" to "profit-driven."
Second, leading companies have not significantly increased leverage.
Compared to the dot-com bubble period, this round of AI investment relies more on the endogenous cash flow of tech giants. Although capital expenditure is high, the debt-to-equity ratio remains significantly below the average level during the dot-com bubble.
Third, valuation levels are not high.
Wind data shows that as of the end of May 2026, the forward P/E ratio of the S&P 500 Information Technology sector was only 24 times, far below the 55 times valuation during the dot-com bubble in 2000. High growth in performance may already be digesting the valuation.

02 From infrastructure to vertical application realization pace
The current AI industry chain has formed a clear upstream, midstream, and downstream structure, with different maturity levels and investment logic for each segment.
Upstream (infrastructure layer): "Ballast" with high performance certainty

The upstream is the foundation of AI applications, mainly including computing power, data, and cybersecurity.
Computing power and hardware: The focus of AI computing power consumption is shifting from training to inference. Barclays estimates that by 2026, over 70% of generative AI's total computing power demand will come from inference, with inference computing power possibly reaching 4.5 times that of training. Future computing power demand is expected to continue growing. This drives more dispersed, larger-scale computing needs, with higher requirements for energy efficiency and deployment costs. In 2026 and the next 2-3 years, the global and Chinese AI computing power markets are expected to remain in a supply-demand imbalance, making this one of the most prominent bottlenecks.
Data services and cybersecurity: The first step for enterprise AI deployment often involves data governance, integration, and security. Therefore, revenue growth of related companies serves as a leading indicator. Some North American data service and cybersecurity companies have reported multiple consecutive quarters of exceeding expectations, validating the authenticity of enterprise AI investments and providing a clear mapping logic for domestic targets.
Midstream (platform and model layer): The core of value distribution and the reconstructor of business models
The midstream mainly includes large model services (MaaS), vertical domain models, and toolchains.

Token economy has become a core driving force: Tokens have evolved from mere technical parameters to the key production factors and value carriers in the AI era. According to the National Development and Reform Commission, China's daily token call volume has exploded exponentially, surging from about 1 trillion at the beginning of 2024 to 140 trillion in early March 2026, a two-year increase of over a thousand times. This has led to mature business models around tokens, including token aggregation platforms, cloud vendor MaaS services, etc.
Model competition and value backflow: Despite the diversity of applications, the enormous value in the industry chain is accelerating back to the model layer. While AI startups are experiencing rapid revenue growth, most profits are ultimately paid as costs to underlying model vendors and computing power providers. Domestic models have gained significant market share in the global call market, demonstrating strong international competitiveness.
Downstream (application layer): Finding true growth amid differentiation
The downstream applications directly target enterprise and consumer users, representing the final scene of value realization. This segment is currently highly fragmented, with opportunities and risks coexisting.
Enterprise applications: Early commercial validation signals have appeared in some vertical fields, especially in data-intensive and workflow-complex industries such as financial risk control, insurance underwriting, enterprise data integration, and cross-border marketing. AI agents with autonomous planning and tool invocation capabilities are becoming feasible commercial applications, deeply integrating into office automation, customer service, and other scenarios.
Consumer applications: For example, AI short dramas/comics have achieved clear profitability by significantly reducing costs and increasing efficiency. Additionally, AI video generation tools, driven by high demand, have seen rapid price increases. Intelligent assistant applications are showing user stickiness by improving research and learning efficiency, with willingness to pay still being cultivated.

03 "Right-side expectations" for AI applications may already exist, focusing on five major directions
In summary, the development of the global AI industry shows a clear "infrastructure first, applications follow" characteristic. Profit margins for application companies are generally squeezed by upstream model invocation and computing costs, with the industry chain value increasingly concentrated at the model and infrastructure levels.
The fundamentals (profitability) of the AI application sector may still be on the left side, but industry trends and early commercialization validation in some niche tracks have already provided market right-side expectations.
Therefore, it may be difficult for the AI application sector to "bloom everywhere," and those with clear commercial validation signals and initial profitable models in niche tracks may have more opportunities.

Five major directions:
First, large model ecosystems. Large models are the foundational infrastructure of AI applications, with ongoing industry opportunities around training, fine-tuning, and deployment.
Second, AI re-monetization. In the mobile internet era, the fastest monetization was through advertising and e-commerce, and the same may apply in the AI era. For example, in AI e-commerce, users are increasingly accustomed to searching and comparing products via large models, which are beginning to embed transaction modules. Companies providing SaaS systems, data services, and API interfaces for this ecosystem are expected to gain incremental business.
Third, AI programming. According to the "2025 AI Usage Report" jointly released by OpenRouter and a16z (based on 1 quadrillion token call data), token consumption for programming tasks surged from about 11% at the beginning of 2025 to over 50%. Leading overseas vendors have annual revenues reaching hundreds of billions of dollars, with exponential growth, and domestic related deployments are also worth attention.
Fourth, AI for Science. This is one of the most imaginative directions. In fields like pharmaceuticals, new materials, and new energy, traditional R&D requires many scientists to spend years on molecular screening, but AI can significantly shorten this cycle. For example, AI in drug discovery can quickly identify candidate molecules from tens of thousands, greatly improving R&D efficiency. This is not just cost reduction and efficiency increase but a leap in productivity.
Fifth, AI software going overseas. Overseas users have a higher willingness to pay for software than domestic users. Several domestic software companies are already experiencing rapid revenue and profit growth in overseas markets, and those with a high proportion of overseas business may benefit.

The AI industry trend is clear, but the maturity and investment logic of each upstream, midstream, and downstream segment vary. How should investors position themselves?
Digital economy, as the core carrier of AI technology implementation, includes hardware bases like computing centers, chips, and communication equipment, as well as software engines that deeply transform traditional industries such as finance, smart logistics, and telemedicine.
This combination of "hardware and software" gives the digital economy a unique allocation value in the current market. It encompasses both the proven performance of computing power and the underpriced potential of AI applications.