In response to recent market volatility, Xing Ziqiang, Chief Economist for Morgan Stanley China, made a clear judgment at a media briefing on July 30: this round of A-share adjustment is actually a microcosm of the global AI technology sector's resonance, not because the domestic economic fundamentals suddenly deteriorated.

AI investment is entering a "half-time rest" shift period. In the second half, investors need to broaden their perspective, and the core investment logic will undergo a shift.

|A-share volatility is a microcosm of global AI resonance, and fundamentals have not deteriorated

This round of A-share volatility is actually a microcosm of the global AI technology resonance, not due to a turning point in the A-share market itself or the domestic economic fundamentals.

1. Where does the impact come from? The global AI industry chain is highly intertwined

Looking back at the transmission chain of this round of adjustments, the initial shock began in the Korean market, where previously outstanding storage stocks were the first to suffer, then quickly spread to the United States, and ultimately to Chinese mainland.

In this AI cycle, Chinese mainland companies mainly participate in midstream supporting segments such as optical modules and PCB panels. Although people have shared in the dividends, due to the high degree of coordination and deep binding of the global AI industry chain, it is naturally impossible to remain unaffected by this global upheaval.

2. Why the drop? Macro concerns intertwined with trading crowding

Xing Ziqiang believes that the reason behind this global resonance lies in the complex intertwining of macro factors and micro AI investment narratives:

Macro concerns: Middle Eastern geopolitical conflicts have kept oil prices high, sparking market concerns about inflation; At the same time, the market is full of uncertainty about the Fed's policy style and rate cut outlook, with overall sentiment walking on thin ice.

Micro-level overdraft: The capital market has seen a market trend completely opposite to that of the real economy. Some companies may have already overloaded their expectations for the next year or two, resulting in overly high valuations.

3. Will the Federal Reserve raise interest rates? The impact on AI is limited

In response to market concerns about Federal Reserve policy, Xing Ziqiang gave a clear assessment:

Extremely low probability of rate hikes: The Morgan Stanley team believes the likelihood of the Fed raising rates this year is low. The public often categorizes Fed Chairman Walsh as a "hawk" due to his speech style, but he has a forward-looking understanding of the deflationary effects AI brings, and institutions tend to view it as "hawkish on the outside, dovish on the inside."

利率难挡AI革命：更为关键的是，即便美联储真的加息25个基点，也不会对AI革命产生巨大冲击。AI是关乎社会生产力进步的划时代投资，很少有人会仅仅因为利率的小幅抬升就动摇。

4. 实体产业真相：订单饱满，资本开支未减

从实体产业的资本开支来看，AI大厂的投入不仅没有削减，反而略有上调。今年的投资计划约为8000亿美元，明年更是预计高达1.2万亿美元。无论是算力核心环节还是配套产业链，相关企业的在手订单依然饱满，并没有出现断崖式下跌。

｜剧烈波动是挤泡沫而非技术破产

既然基本面和宏观利率都不足以动摇AI的产业趋势，那么近期科技板块为何会经历“过山车”般的剧烈波动？摩根士丹利中国首席经济学家邢自强指出，核心症结在于全球资金的一致预期过于强烈，导致交易格局极度拥挤。

1. 暴跌真相：杠杆资金脆弱与万亿“抽血”效应

杠杆交易极度脆弱：全球资金都在重仓押注AI算力基础设施（如韩国市场的典型杠杆交易）。当海量杠杆资金集中在单一板块时，市场对任何流动性收紧的风吹草动都会变得极其敏感。

巨头融资“抽血”：为了支撑算力中心的巨额投资，AI头部企业和互联网大厂正在海量抽取市场资金。据统计，从今年上半年到明年上半年，头部AI企业计划通过IPO、发债等工具融资合计约1万亿美元。在这种极致的“抽血”效应下，市场草木皆兵，最终引发了剧烈调整。

2. 历史镜像：这是“局部清算”，而非“技术破产”

From the grand perspective of technological history, every profound technological revolution goes through a cycle of "overinvestment—frenzied prosperity—partial liquidation." The current adjustment is essentially about squeezing out valuation bubbles and revising the financial expectations of individual companies, rather than declaring the bankruptcy of AI technology routes.

3. Lessons from the Internet bubble: When the bubble bursts, the base endures

The internet revolution of the late 1990s is the most vivid example. At that time, the market was extremely enthusiastic, with the belief that investment in broadband and routers would expand endlessly. After the tech bubble burst in 2000, related equipment manufacturers' stock prices suffered a devastating blow. But history proved that the submarine cables and network routing facilities accumulated at that time were not wasted; they formed a solid foundation for the later prosperity of mobile internet.

| How is China building its own computing power foundation?

Currently, domestic large models have forged a differentiated competitive path. The industry generally adopts an open-source model and deeply optimizes algorithms and computing power calls to minimize operating costs.

According to Morgan Stanley's estimates, the current inference cost of large models in China is only about one-tenth that of the United States. This significant price advantage makes domestic models highly attractive for commercial implementation and long-tail scenario applications.

Xing Ziqiang suggested drawing on the successful experiences of the mobile internet era and replicating them in the field of AI computing power. The specific approach is: under the leadership of the state (public finance and state-owned power), large-scale, intensive computing power centers are built, and then "computing power rental" services are provided at lower costs to domestic large model manufacturers and tech startups.

Kingtech Perspective | Embracing the "Second Half," Mining AI Applications and HALO Assets

Morgan Stanley clearly pointed out that the current market shock is due to "partial liquidation" caused by trading crowding and leveraged capital clearing, not a collapse of AI industry logic. This means that in the second half of AI investment, the market will completely bid farewell to the frenzy of "mindless buying and selling shovels (hardware)." Funding will become more selective, shifting from simply hyping underlying computing power to seeking "AI Adaptor" companies that can truly leverage AI to reduce costs, increase efficiency, and boost revenue.

Focusing on the dual driving forces of "AI application ends" and "HALO assets."

In the investment logic for the second half of AI, investors are advised to focus on two main directions:

First, "AI application ends" with real real-world scenarios, especially software and internet platforms that can use AI technology to form business closed loops and enhance their core competitiveness;

Second, "HALO Assets" (Heavy Assets Low Obsolescence), which is the resource and energy security sector deeply linked to computing power. With the large-scale construction of AI data centers, the value of underlying physical foundations such as power, energy storage, and strategic minerals will be revalued, as these assets possess strong inflation resistance and long-term allocation value.

Seize the differentiated dividends of China's "high cost-performance computing infrastructure."

The inference cost of large models in China is only one-tenth that of the United States, and the country is leading the construction of intensive computing power centers. This "open-source model + national-level computing power network" model is a unique moat for China's AI sector.

In terms of investment, it is recommended to focus on core infrastructure companies deeply involved in the construction of the national computing power network, as well as AI application companies that can rapidly iterate and capture the global long-tail market by leveraging low-cost domestic computing power. In today's macro environment full of uncertainty, this Chinese version of digital infrastructure, which offers "extreme cost-performance" and "policy support," is the optimal solution to weather cycles.