CITIC SEC: The Rigidity Narrative and Multi-dimensional Constraints of AI Development

CITIC SEC released a research report stating that the current AI industry in the United States exhibits characteristics of a "rigid bubble." The "rigidity" comes from AI being deeply embedded in the core of U.S. national strategy and political correctness, enjoying strong policy support and giant profit backing; while the "bubble" attribute reflects a significant pricing of future growth in valuations, with a tense race unfolding between massive capital expenditure and output efficiency, cash flow creation. The short-term strong reality of AI is being suppressed by a medium-term narrative, and the reversal of the bubble narrative requires advancements in model paradigms, monetary breakthroughs, and the beginning of an interest rate reduction cycle. It is recommended to construct a three-tier dynamic asset allocation portfolio when the narrative reversal comes. The first tier is the "ballast stone" opportunity of internet giants; the second tier is the "shovel-type" opportunity of computing power chain; the third tier is the "reverse-type" opportunity of the software sector. Key points of CITIC SEC: 1. Developing AI has become a core element of U.S. national strategy and political correctness, receiving strong policy support and fundamental support from giants. 2. This constitutes the fundamental source of the current "rigidity" of AI prosperity. Economic data show clearly that the share of AI-related industries in the actual GDP of the United States has rapidly increased from 5.35% in Q1 2020 to 7.36% in Q3 2025, and the growth rate is still accelerating, indicating that it has become an indispensable core growth pillar. At the same time, AI giants represented by "MAG 7" occupy more than 30% of the market value of the S&P 500 index, contributing the vast majority of the market's growth momentum, forming a "too big to fail" financial market influence. 3. More crucially, this economic importance has been deeply politicized. The "Tech Right" represented by Musk and Sachs has formed a closed loop of "Personnel-Policy-Interests" through huge political donations and direct participation in politics, linking the prosperity of the AI industry with their own political lives directly. For this reason, the U.S. government has systematically abolished the original regulatory framework and simplified infrastructure approval through actions like the "Winning the AI Competition" plan, opening up federal land, and even pushing legislation to suppress state regulations with federal power, aiming to clear all obstacles for AI capital expenditures. 4. This narrative of raising AI competition to the level of the "Digital Age Manhattan Project" and the urgent pressure of the midterm elections in 2026 mean that within the foreseeable policy cycle, the U.S. government has extremely strong internal rigidity in explicitly or implicitly supporting the AI industry, significantly reducing market concerns about short-term fluctuations. 5. The current valuation of the U.S. stock AI sector is at a high level and is in a crucial race with future profit growth. 6. The market's rise has a rational profit basis and has not yet shown typical signals of systemic overheating. The key observation is that although the valuation levels represented by the Nasdaq index are at historically high percentiles, the forward price-earnings ratio (Forward PE) has not soared with the index hitting new highs. The fundamental reason is the continuous upward revision of earnings expectations: the 12-month forward EPS of the Nasdaq index has surged from around $610 at the end of 2024 to $826, effectively "digesting" the rise in stock prices. In addition, a key difference from the Internet bubble period is that the revenue growth of tech giants remains robust, with their core cash cow businesses such as cloud services and digital advertising providing a solid base; at the same time, the tech IPO hype that marks the top of a bubble and the large-scale insider selling phenomenon have not appeared. 7. However, risks should not be ignored. Asset 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 expensive premiums for future growth potential and asset pricing. Therefore, the health of the current market entirely depends on whether the narrative of "growth digesting valuation" can continue to be realized. The victory or defeat in this "race" lies in the financial reports of the next few quarters, and any signs of deteriorating "investment-output ratio" could become the fuse that breaks the current fragile balance. 8. The continued expansion of the AI industry faces four constraints from the financial, physical, and competitive levels, constituting a substantial pressure test for the "bubble" side of the "rigid bubble." 9. In the narrative-driven "rigid bubble," excess returns come from active timing of the narrative cycle and scenario-based layered allocation. 10. Given that fundamental and valuation will be suppressed by grand narratives in the medium term, traditional linear extrapolation investment methods are no longer effective. Stock price fluctuations will be mainly driven by the intensity of industry narratives, which will exhibit cyclical tides with technological breakthroughs, policy signals, financing events, etc. Therefore, effective timing lies in identifying and utilizing the marginal changes in narrative intensity to engage in reverse operations. Investors may consider counter-buying when the market becomes overly pessimistic due to events such as technological stagnation, downward revisions in capital expenditure guidance, or crises of iconic companies, and remain vigilant when the narrative becomes overly optimistic due to breakthroughs in models, clear paths for monetization, or improvements in liquidity. Based on this, CITIC SEC believes that the best response is to construct a three-tier dynamic asset allocation portfolio when the bubble narrative reverses. 11. The first tier is the "ballast stone" opportunity, allocating to internet giants with stable cash flows that can provide downside protection and liquidity support in any scenario. The core of this framework lies in abandoning bets on the endgame and instead actively managing uncertainty through strategic layering, effectively controlling the unique risks of narrative and fundamentals while capturing the long-term dividends of AI. 12. The second tier is the "shovel-type" opportunity. Regardless of whose application dominates, computing power infrastructure is indispensable. On one hand, it benefits from the increase in AI capital expenditure to boost EPS. The industry's supply chain shortage also brings opportunities for price increases and profit improvement, while the explosion of AI applications can also reduce concerns about unsustainable capital expenditure and improve valuations. 13. The third tier is the "reverse-type" opportunity. The extreme concerns about AI agents in the market stem from an excessively optimistic view of the prospects for AI applications combined with concerns about the interests of existing software leaders being harmed. However, in the medium to long term, the expectation is that the replacement of some standardized software functions by models and the strengthening of customer stickiness by software vendors accessing SOTA models will coexist in the long run, with the industry's core competitive barriers not fundamentally impacted. In this context, it is recommended to seize the opportunity for deep adjustments in the software sector after the configuration. Risk factors: 1. Deterioration of the macro economy leading to a decrease in risk appetite. 2. Deterioration of the macro economy leading to a decrease in willingness to pay for users. 3. Federal Reserve's policy being more hawkish than expected. 4. Continued tightening of policy regulation in the technology sector. 5. AI companies in the United States borrowing at a pace faster than expected. 6. Iteration of basic models falling short of expectations. 7. AI applications not meeting expectations. 8. Aggravation of global geopolitical risks. 9. Continued intensification of industry competition risks, etc.