Morgan Stanley's Chief Economist, Xing Ziqiang, is known for his professional insight and high visibility. He accurately identified the September 2024 shift in domestic macroeconomic policy, alerting global institutions early to increased pro-growth measures. His forward-looking analysis of China's recent economic traits, including property adjustment cycles, industrial chain relocation, the 'new three' export boom, and AI computing investment cycles, has significantly influenced institutional asset allocation. On July 30, 2026, at a Morgan Stanley Shanghai briefing, he offered fresh perspectives on second-half policy, the economy's 'K-shaped divergence,' AI competition, and recent global tech stock turbulence.
Regarding the recent violent swings in global tech stocks, Xing Ziqiang attributes the correction to a blend of macroeconomic factors and narratives around micro-level investment returns. He views the volatility in China's A-shares as a microcosm of a global AI-driven tech resonance. The current AI investment cycle, he argues, was sparked by major US tech firms' plans to build computing centers, but the benefits have spread beyond the US to companies in South Korea and China. While South Korean and Taiwanese firms participate more in chips and memory, mainland Chinese companies are involved via optical modules and PCBs. However, all these regional players are tied to the same global investment cycle, leading to synchronized capital market impacts.
The AI narrative has shifted.
Xing further clarifies that the narrative shift may not reflect a change in fundamentals. Major AI players' announced investment plans for this year and next have not been cut, and are even slightly raised, with projected spending around $800 billion this year and $1.2 trillion next year. Supply chain orders show no major issues, yet the capital market has experienced a 'roller-coaster' correction. He attributes this to market awe, as prices may have already priced in much of the good news expected over the next one to two years. Historically, every major tech revolution has involved overinvestment and over-optimism, followed by a 'reckoning' in financial markets between stock prices and financial costs. However, he emphasizes this reckoning does not mean investment is wasted or unhelpful to productivity or the economy. He cites the late 1990s internet investment boom: infrastructure like broadband routers and undersea cables ultimately boosted global productivity, but some 'picks and shovels' companies still faced a price-return reckoning after the dot-com bubble burst. The belief that AI will transform human productivity and that it's a huge tech revolution is not contradictory to the current market volatility.
From 'selling shovels' to 'using shovels.'
Citing views from Morgan Stanley's US, Asia-Pacific, and China strategy teams, Xing suggests AI investment may be entering a 'half-time break' phase, where market focus should shift from 'selling shovels' to 'using shovels.' 'The view needs to be broader, not just on AI semiconductors and computing power, but on companies that can benefit from using AI.' While the market has mainly focused on chips and memory (the 'selling shovels' part), the next stage may gradually shift to which companies can genuinely use these 'shovels' to boost productivity and revenue. He summarizes this as a transition from AI infrastructure providers to AI users, companies that can increase income, improve efficiency, and lower costs through AI.
Can't answer the 'bottom fishing' question.
Xing Ziqiang explicitly stated he cannot answer whether now is the time to 'buy the dip' given the market volatility. However, from a broader global perspective on the next stage of the AI industrial revolution and tech revolution, China holds advantages. The boundaries of the AI ecosystem may also continue to expand. Some energy, resource, and strategic raw material companies, seemingly non-AI tech firms, may enter this system due to AI's growing demand for electricity, energy storage, and scarce resources. Morgan Stanley's strategy team categorizes some of these as 'HALO assets,' which are not easily replaced by AI and may even become scarcer as AI infrastructure expands. Therefore, the AI 'second half' does not mean AI investment is over, but that market attention may broaden from narrow semiconductors and computing power to applications, energy, and a wider industrial ecosystem. Within this broad AI ecosystem, China has a significant advantage in energy transition, offering better green energy tech products to the world. The rise of domestic large models, with cheap token costs, also gives a clear edge if global AI commercializes.
More people are starting to believe in AI.
In Xing's view, the most alarming change in AI trading is not that investors suddenly stop believing in AI, but that more and more people are starting to believe. Six months ago, if someone suggested a 25 basis point US rate hike could impact AI investment, many 'young fund managers' would think such a rate change is insufficient to affect a 'humanity-changing' tech revolution. As this perception spread from the 'young' to the 'old guard' and became a global consensus, market structure became extremely crowded. Crowdedness first manifests as leveraged funds concentrated on AI computing infrastructure. Using the South Korean market as an example, from late last year to early this year, multiple leveraged ETFs with several times leverage appeared. Typically, a large influx of such leveraged funds is a feature of the late stage of a speculative or track-based bull market. When regulatory policy tightens or liquidity or interest rate expectations change, leveraged funds can amplify market volatility. A second pressure comes from primary and bond markets. The massive capital required for AI computing centers means some top tech firms cannot cover all capital expenditure with operating cash flow alone, requiring further financing through IPOs, rights issues, and bond issuance. 'From the first half of this year to the first half of next year, over just one year, these companies need to raise $1 trillion in public markets through bonds and stocks, continuously absorbing liquidity.' This means one side has leveraged funds concentrated on AI, while the other has AI firms absorbing liquidity from stock and bond markets. The market becomes more sensitive to marginal changes in inflation, oil prices, interest rates, and central bank policy. These intertwined factors have triggered some shifts in the AI narrative.
Another route for the AI industry.
Discussing AI competition, Xing believes China is taking a different path, relying more on system integration, power supply, engineering talent dividends, and algorithm optimization to lower AI usage costs. The most impactful data point is token cost. 'The token cost of domestic large models is roughly only one-tenth of that in the US, indeed cheap.' This low cost is gradually showing value on the enterprise side. Some overseas companies previously encouraged employees to use the most advanced models extensively, but as token bills grew, they began to layer AI tasks, assigning complex, long tasks to top models and simple, short tasks to cheaper open-source models. Xing thus asks whether China can replicate its experience from the 2G, 3G, and 4G eras, building AI computing power into a cheap, widely usable digital infrastructure. In the mobile internet era, China lowered the barrier for startups and applications through strong network infrastructure and relatively low tariffs. Similarly, if future computing power can be rented via a national computing network, it could be a Chinese version of a digital infrastructure plan. In chip usage, Xing proposes a 'dual-track computing power system.' On one hand, training the most advanced models still requires the world's most cutting-edge chips. On the other hand, for the reasoning and application phase, the performance requirement for a single chip is relatively lower, allowing domestic chips to gain more application opportunities through state-led computing centers. After more applications in the reasoning phase, domestic chips can accumulate feedback through 'learning by doing,' continuously optimizing systems and efficiency.
A 'September 24' style turning point has not yet arrived.
Returning to China's macro policy, Xing Ziqiang's report title directly states that a 'September 24' style turning point has not arrived. He notes that while some economic indicators weakened in the second quarter, recent policy documents and official statements show policy focus remains on tech, energy, industrial self-sufficiency, and solving 'bottleneck' issues. 'Even with the second-quarter economic slowdown, if we need to find a lever for support, it will likely follow the path of focusing on technology and energy transition. In other words, intervention from the investment side is more likely than from the consumption side.' Additionally, Xing believes the second half of this year is more likely to see the accelerated implementation of fiscal resources already budgeted but not yet deployed, rather than a sudden expansion of the deficit or new large-scale special treasury bonds. He points to over 2 trillion yuan in fiscal capacity available for use in the second half, from central and local government bond balances and previously established new financial instruments. The specific directions are likely investment-side, including grid and energy storage for energy, and AI and computing power for technology.
'Divergence' difficult to resolve in the short term.
Furthermore, Xing believes the economy's 'divergence' situation is difficult to resolve quickly. The upward side includes AI and new energy, while the downward side includes domestic demand, consumption, real estate, and employment. For this structural issue, Xing proposes three suggestions. First, optimize the trade-in policy, redirecting some resources from durable goods like cars and home appliances to service consumption like dining, entertainment, and tourism. Second, consider whether some hard-tech industries already in a global upcycle with high export growth still need strong export tax rebates and fiscal support; resources saved could be used for tax cuts in domestic service and consumer industries. Third, continue to improve the social security system to reduce residents' precautionary savings.
Comments