---
title: "CITIC Securities: The Rigid Narrative and Multidimensional Constraints of AI Development"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/276806091.md"
description: "CITIC Securities released a research report indicating that the AI industry in the United States is in a state of \"rigid bubble,\" with a competition between high valuations and future profit growth. AI has become a core component of the U.S. national strategy, enjoying policy support and backing from major corporations. It is recommended to construct a three-tier dynamic asset allocation portfolio: \"ballast\" opportunities from internet giants, \"shovel-type\" opportunities from computing power chains, and \"contrarian\" opportunities from the software sector. The contribution of the AI industry to U.S. GDP continues to rise, and its economic importance has been deeply politicized"
datetime: "2026-02-25T00:31:03.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/276806091.md)
  - [en](https://longbridge.com/en/news/276806091.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/276806091.md)
---

> Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/276806091.md) | [繁體中文](https://longbridge.com/zh-HK/news/276806091.md)


# CITIC Securities: The Rigid Narrative and Multidimensional Constraints of AI Development

According to the Zhitong Finance APP, CITIC Securities has released a research report stating that the current AI industry in the United States exhibits characteristics of a "rigid bubble." Its "rigidity" stems from AI being deeply embedded in the core of U.S. national strategy and political correctness, enjoying strong policy support and backing from major corporations; while the "bubble" attribute is reflected in valuations that have significantly priced in future growth, with a tense race unfolding between massive capital expenditures and output efficiency, as well as cash flow generation. The short-term strong reality of AI is suppressed by mid-term narratives, and a reversal of the bubble narrative requires advancements in model paradigms, breakthroughs in monetization, and the onset of a rate-cutting cycle. It is recommended to construct a three-tier dynamic asset allocation portfolio when the narrative reversal arrives. The first tier is the "ballast stone" opportunities of internet giants; the second tier is the "shovel-type" opportunities of the computing power chain; the third tier is the "contrarian" opportunities in the software sector.

## The main points of CITIC Securities are as follows:

**The development of AI has become the core of U.S. national strategy and political correctness, receiving strong policy backing and fundamental support from major corporations.**

This constitutes the fundamental source of the current "rigidity" of the AI boom. Economic data clearly shows that the proportion of AI-related industries in the U.S. real GDP has rapidly risen from 5.35% in Q1 2020 to 7.36% in Q3 2025, with the growth rate still accelerating, indicating that it has become an indispensable core growth pillar. Meanwhile, AI giants represented by "MAG 7" continue to occupy over 30% of the market capitalization of the S&P 500 index, contributing the vast majority of market growth momentum, forming a "too big to fail" influence in the financial market.

More critically, this economic importance has been deeply politicized. The "tech right" represented by figures like Musk and Sachs has formed a closed loop of "personnel-policy-interest" through massive political donations and direct participation in politics, leading the current U.S. government to directly bind the maintenance of AI industry prosperity to its own political survival. To this end, the U.S. government has systematically abolished the existing regulatory framework and simplified infrastructure approvals through action plans like the "Winning the AI Race," opened federal lands, and even promoted legislation to suppress state regulations with federal power, aiming to clear all obstacles for AI capital expenditures.

This elevation of AI competition to a national security narrative akin to a "Manhattan Project for the digital age," combined with the urgent pressure of the 2026 midterm elections, means that within the foreseeable policy cycle, the U.S. government's explicit or implicit "backing" of the AI industry has a strong inherent rigidity, significantly reducing market concerns about short-term volatility.

**The current valuation of the AI sector in U.S. stocks is at a high level, engaging in a critical race with future profit growth.**

The market's rise has a rational profit foundation, and typical signals of systemic overheating have not yet appeared. The core observation is that while the valuation level represented by the Nasdaq index is at a historically high percentile, its forward price-to-earnings ratio (Forward PE) has not soared alongside the index reaching new highs, but instead shows a high-level consolidation trend. The fundamental reason lies in the continuous upward revision of corporate profit expectations: the 12-month forward earnings per share (EPS) of the Nasdaq index has strongly risen from about $610 at the end of 2024 to $826, with strong profit growth effectively "digesting" the increase in stock prices. Additionally, a key difference from the internet bubble period is that the revenue growth of current tech giants remains robust, with their core cash cow businesses such as cloud services and digital advertising providing a solid foundation; At the same time, the technology IPO boom that marks the top of the bubble and the large-scale reduction of holdings by insiders have not yet appeared.

However, risks cannot be ignored. Valuation indicators such as price-to-book ratio (PB) and price-to-sales ratio (PS) have reached historical highs, indicating that the market is paying a high price for future growth potential and asset premiums. Therefore, the current health of the market entirely depends on whether the narrative of "growth digesting valuation" can continue to be fulfilled. The outcome of this "race" hinges on the financial reports of the next few quarters; any signs of deterioration in "input-output ratio" could become the trigger that breaks the current fragile balance.

**The continuous expansion of the AI industry is facing quadruple constraints from financial, physical, and competitive aspects.**

**This constitutes a substantial stress test on the "bubble" side of the "rigid bubble."**

The first constraint is reflected in the dangerous "scissors gap" between capital expenditure and output efficiency. Major giants plan to continue expanding capital expenditures, betting on the Scaling law, with the combined capital expenditure guidance of four giants reaching 640.4 billion yuan by 2026, a year-on-year increase of 55%. Their investment intensity has far exceeded current industry revenues, and this high-intensity investment, akin to "eating the grain of the next year," entirely relies on the exponential explosion of future application layer revenues.

The second constraint is the impending cash flow pressure. It is estimated that by 2026, leading companies may face a gap between their free cash flow and the commitment to maintain high shareholder returns after meeting massive investments, forcing the industry to shift from internal self-financing to relying on external financing, as evidenced by the rapid rise to over 150 billion dollars in industry debt.

The third constraint is the unavoidable physical ceiling. Global advanced semiconductor capacity (such as TSMC) is nearing short-term bottlenecks, while the U.S. Energy Information Administration (EIA) predicts that data center electricity demand will account for 9% of total U.S. electricity consumption by 2030. The U.S. power grid may face supply shortages starting in 2026, and the simultaneous shortage of chips and electricity locks in the short-term upper limit of computing power expansion.

The fourth constraint is the unresolved competition in technological pathways. The gap between OpenAI and Google's Gemini has already been closed, while Google's investment in Anthropic has also shone brightly with OpenClaw. If Google wins with its complete ecosystem, it will challenge the current market premium paid for "multiple technological paths running in parallel," potentially triggering a full industry chain value reassessment from chips to cloud services.

**These four constraints are interwoven, and any serious disconnection in one aspect could transform structural risks into systemic pressures.**

**In the narrative-driven "rigid bubble," excess returns come from proactive timing of the narrative cycle and scenario-based layered allocation.**

Given that fundamentals and valuations will be suppressed by grand narratives in the medium term, traditional linear extrapolation investment methods have already failed. Stock price fluctuations will primarily be driven by the intensity of industry narratives, which exhibit cyclical tides with technological breakthroughs, policy signals, financing events, etc. Therefore, effective timing lies in identifying and utilizing marginal changes in narrative intensity for contrarian operations. Investors should consider buying against the trend when the market becomes overly pessimistic due to technological stagnation, downward adjustments in capital expenditure guidance, or crises in landmark companies, and remain vigilant during overly optimistic narratives when model breakthroughs, clear monetization paths, or liquidity improvements occur Based on this, CITIC Securities believes that the best response is to construct a three-layer dynamic asset allocation portfolio when the bubble narrative reversal arrives.

The first layer is "ballast" opportunities, allocating to internet giants with robust cash flows, which can provide downside protection and liquidity support in any scenario. The core of this framework lies in abandoning the gamble on the endgame and instead actively managing uncertainty through strategic layering, thereby effectively controlling its unique narrative and fundamental risks while capturing the long-term benefits of AI.

The second layer is "shovel-type" opportunities. Regardless of which application king emerges, computing infrastructure is essential. On one hand, it benefits from increased AI capital expenditures that enhance EPS, while supply shortages in the industry chain also present opportunities for price increases to boost profits. On the other hand, the explosion of AI applications can alleviate concerns about unsustainable capital expenditures, thereby enhancing valuations.

The third layer is "contrarian" opportunities. The market's extreme concerns about AI Agents essentially stem from excessive optimism about the prospects of AI applications combined with worries about the interests of existing software leaders being harmed. In the medium to long term, it is expected that models will replace some standardized software functions, and software vendors will integrate SOTA models to strengthen customer stickiness, which will run in parallel without fundamentally impacting the industry's core barriers. In this context, it is recommended to seize the layout opportunities after the deep adjustment of the software sector.

**Risk Factors:**

Deterioration of the macro economy leading to a decline in risk appetite; deterioration of the macro economy leading to a decline in users' willingness to pay; Federal Reserve policies being more hawkish than expected; ongoing tightening of regulatory policies in the technology sector; US AI companies' debt accumulation pace exceeding expectations; basic model iterations falling short of expectations; AI applications not landing as expected; escalating global geopolitical risks; and ongoing intensification of industry competition risks, among others

### Related Stocks

- [Global X Artfcl Intlgc & Tech ETF (AIQ.US)](https://longbridge.com/en/quote/AIQ.US.md)
- [ARK Innovation ETF (ARKK.US)](https://longbridge.com/en/quote/ARKK.US.md)
- [ARK Autonomous Technology&Robotics ETF (ARKQ.US)](https://longbridge.com/en/quote/ARKQ.US.md)
- [iShares Future AI & Tech ETF (ARTY.US)](https://longbridge.com/en/quote/ARTY.US.md)

## Related News & Research

- [Google, Accel India accelerator choses 5 startups and none are ‘AI wrappers’](https://longbridge.com/en/news/279178346.md)
- [From AI Fabs to Fintech: Four New Tools for Fast Narratives](https://longbridge.com/en/news/279800895.md)
- [2 Beaten-Down AI Stocks Investors Shouldn’t Dump Yet](https://longbridge.com/en/news/279827679.md)
- [First Verifiable AI Agents Go Live on Mainnet: Cysic AI Launches Autonomous Agent Swarms with Cryptographic Proof](https://longbridge.com/en/news/279479162.md)
- [11:05 ETFaye Launches Fractional Chief AI Officer Services to Provide Strategic AI Leadership for Mid-Market Organizations](https://longbridge.com/en/news/279456596.md)