Morgan Stanley: A-Share Volatility Is a 'Half-Time Break' in Global AI Resonance; Fed Unlikely to Hike This Year; China Can Leapfrog in AI Applications with 'These Two Elements'

Morgan Stanley (hereinafter 'Morgan Stanley') Chief China Economist Xing Ziqiang stated on Thursday (30th) during a media briefing that the recent synchronized pullback in A-shares, Korean equities, and the U.S. AI hardware supply chain is not due to a sudden deterioration in China's fundamentals, but rather a 'narrative correction' driven by global AI technological resonance.

Although the major cycle of AI capital expenditure has not ended, trading has become excessively crowded, and the market is now entering a 'half-time break'.

He also pointed out that the likelihood of a rate hike by the U.S. Federal Reserve (Fed) this year is low, and the new chair, Walsh, exhibits a 'hawkish exterior, dovish interior' stance. He advised China to replicate its mobile internet-era success by adopting an 'open-source models + national computing grid' strategy to achieve a 'leapfrog' in AI applications.

Xing reviewed the origin of the current sell-off: South Korean memory stocks collapsed first, followed by declines in U.S. tech giants, which then transmitted to A-share optical modules, PCBs, and other midstream components.

He noted that AI computing centers operate within a globally distributed supply chain: the U.S. provides the architecture, Japan and South Korea supply memory chips, and China integrates the midstream components. As a result, upstream and downstream sectors are all within the same cycle. Thus, while markets from Korea to the U.S. and A-shares have been impacted, fundamentals have not deteriorated. In fact, major firms' capital expenditure plans for the next two years are not decreasing but increasing—projected at approximately $870 billion this year and $1.2 trillion next year. Orders remain secured; the volatility stems from some companies having prematurely priced in one to two years of future expectations, not from disappearing demand.

Xing drew a parallel with the 2000 dot-com bubble, noting that while overinvestment in undersea cables and routers triggered financial corrections, the infrastructure hardware was not wasted—it instead supported the subsequent internet boom. Similarly, this round of AI 'reckoning' does not invalidate the productivity narrative.

Therefore, Morgan Stanley's strategy team recommends shifting investment focus from 'shovel sellers'—AI computing semiconductors—to 'shovel users'—application companies that leverage AI to reduce costs, increase revenue, and improve efficiency—as well as sectors with 'HALO assets', such as resources and energy.

Additionally, Xing noted that the Fed's decision to hold rates steady early Thursday morning, combined with Walsh's ambiguous forward guidance, has failed to ease market concerns. However, Morgan Stanley's U.S. team maintains that the Fed will not hike rates this year. While Walsh's rhetoric appears hawkish, his core message repeatedly emphasizes the deflationary logic that 'AI boosts productivity and suppresses employment and inflation'—a stance Xing characterizes as 'hawkish exterior, dovish interior'.

He believes that even a single rate hike by the Fed would not significantly undermine confidence in transformative technology investments. The real issue is that market expectations have become too uniformly bullish. In the first half of this year, leveraged funds in Korea were heavily concentrated in AI infrastructure. Moreover, major AI and internet firms are expected to raise approximately $1 trillion in equity and debt financing over the next year, making the market extremely sensitive to oil prices, inflation, and hawkish Fed signals.

Regarding the path to leapfrogging in the AI industry, Xing calculated that China's large model inference cost is about one-tenth that of the U.S., primarily due to its open-source approach and dual algorithm optimization. While overseas tech giants announce massive capital expenditures, they face constraints from power grids, community opposition, and hard power supply limits.

Xing believes China can leverage public fiscal funding to build large-scale computing centers and rent computing power at low prices to large model and tech companies, transforming the AI computing grid into a digital public infrastructure akin to 3G/4G networks. This 'digital infrastructure model'—'compensating for weaknesses in advanced computing power and nurturing an ecosystem for application and startup innovation'—is precisely the most underappreciated expectation gap where China holds a relative advantage in the second half of the AI race.