The recent volatility in Chinese A-shares is a microcosm of a global AI technology resonance, rather than being triggered by a sudden inflection point in China's own economy or domestic fundamentals. This view was shared by Morgan Stanley's Chief China Economist, Robin Xing, during a media briefing on July 30.

Looking ahead, Xing and his strategy team believe that AI investment is entering a "half-time break," requiring investors to broaden their horizons. The primary theme for the second half will shift from solely focusing on AI semiconductors and the "pick-and-shovel sellers" of underlying computing power, toward companies that can apply AI to reduce costs, improve efficiency, and boost their own revenues (AI adopters). Furthermore, broader ecosystem sectors characterized as "HALO assets," such as resources and energy security closely tied to computing power, also face strategic opportunities for value reassessment.

The recent global stock market turbulence, particularly in the AI and computing power sectors, has been dramatic. Xing attributed the fluctuations in Chinese A-shares to this global AI resonance rather than any sudden domestic changes. The initial shock in this adjustment cycle began in the South Korean market, where previously strong memory chip stocks faltered, before rapidly spreading to other major global markets. This phenomenon highlights the deep synergy and interconnectedness of the global AI supply chain. In this AI cycle, Chinese mainland companies primarily participated in mid-stream supporting roles, such as optical modules and PCB panels. While benefiting from the cycle, they were inevitably affected by the volatility that spread from South Korea to the US and then to the Chinese mainland.

Xing believes the global resonance was caused by a complex interplay of macroeconomic factors and micro-level AI investment narratives. On the macro side, geopolitical conflicts in the Middle East have kept oil prices high, raising concerns about inflation stickiness. Meanwhile, uncertainty persists regarding the US Federal Reserve's policy direction and the outlook for rate cuts. Although economists see a minimal probability of a rate hike within the year, the market remains cautious. The Morgan Stanley US economics team has consistently argued that a rate hike this year is unlikely. While Kevin Warsh is often categorized as a hawk due to his speaking style, Xing noted that his core stance may not be. Warsh has repeatedly laid the groundwork for AI's potential deflationary impact on employment, leading the institution to view him as "outwardly hawkish but inwardly dovish." Crucially, even a 25-basis-point rate hike by the Fed, Xing argued, should not severely impact the AI revolution. Theoretically, it shouldn't, as this is a landmark investment in social productivity progress, and few would be deterred by a minor rate increase.

Analyzing capital expenditure from the real economy, Xing pointed out that major AI companies' investment plans for this year and next have not been cut, but have even been slightly increased. This year's planned spending is around $800 billion, with projections for next year reaching $1.2 trillion. Order books for companies involved in both core computing power and supporting supply chains remain robust, with no cliff-edge decline. Yet the capital market has moved in the opposite direction, a divergence that requires investor vigilance. Xing admitted, "We must respect the market—some companies may have already priced in expectations for the next one to two years."

If fundamentals and macro rates are not enough to derail the industry trend, what is the real cause of the significant volatility in the tech sector? Xing identified the core issue as an overly strong consensus among global investors, leading to extremely crowded trades. On one hand, heavily leveraged positions in AI computing infrastructure were seen globally, notably in South Korea. When excessive leverage is concentrated in a single sector, the market becomes highly vulnerable to marginal liquidity tightening. On the other hand, from a financing perspective, leading AI companies and internet giants are absorbing vast amounts of market capital to fund massive data center investments. According to statistics, from the first half of this year to the first half of next, top AI firms plan to raise a total of approximately $1 trillion through IPOs, rights issues, and bond offerings in public markets. This extreme "capital absorption" effect has made the market hypersensitive to any hint of rising inflation or rate changes, culminating in a roller-coaster correction.

Viewing the current AI narrative from a broader history of technology, Xing noted that every profound tech revolution typically undergoes a cycle of over-investment, exuberant frenzy, and eventual local correction. This "correction" does not signify the failure of the technology but rather a bursting of valuation bubbles and a revision of individual company financial expectations. The internet revolution of the late 1990s serves as an example. The market was extremely enthusiastic, believing investment in broadband, routers, and undersea cables would expand indefinitely. The bursting of the dot-com bubble in 2000-2001 led to a devastating crash in the stock prices of related equipment suppliers. However, in hindsight, the undersea cables and network routing infrastructure laid down during that period were not wasted; they formed the solid foundation for the subsequent mobile internet boom. Based on this, Xing firmly believes that "AI will profoundly change productivity, and this is not contradictory to market volatility."

Regarding how to build a distinctive computing power foundation amidst the accelerating AI industry, Xing believes that leveraging the open-source ecosystem and extreme cost-effectiveness, combined with the coordinated planning of a national computing network, could incubate a highly competitive, Chinese-style digital infrastructure solution. Currently, Chinese domestic large language models (LLMs) exhibit distinct differentiation, typically using open-source models and continuously reducing operating costs through dual optimization of computing power calls and underlying algorithms. Morgan Stanley estimates that the inference cost of Chinese LLMs is currently about one-tenth of that in the US, a significant price advantage. This extreme cost control holds immense appeal for commercial deployment and long-tail application scenarios. Xing pointed out that "open-source, more cost-effective Chinese models, if combined with policy support for computing power rental from a national computing network, could form a unique competitive advantage for a Chinese digital infrastructure solution."

Xing suggested learning from the mobile internet era and applying it to the AI computing power field. By having public finance and state-owned entities lead the construction of large, intensive computing power centers and offering rental services at lower costs to domestic LLM developers and various tech startups, several benefits could emerge. This would not only alleviate the pain point of high computing costs for individual companies but also stimulate a new wave of business model innovation and entrepreneurial activity across society. Overall, Xing identified several strategic advantages for China. These include a huge advantage in power supply and energy transition closely related to AI, while also being able to supply green energy products globally; the potential for highly cost-effective domestic LLMs to capture significant share in the vast global long-tail market during the commercialization phase; and the advancement of localized AI computing power and national computing network infrastructure providing a more solid development base for China's tech ecosystem. In the broad context of the next-stage AI-led tech revolution, China's systemic advantages remain worthy of serious investor attention.