On November 8, news emerged that with the Democratic Party controlling Congress, an AI super-ban proposal triggered panic selling in the market, causing violent swings in overseas US markets. US stocks plunged on Thursday, with the Philadelphia Semiconductor Index crashing 12%, before staging a retaliatory rebound on Friday, though the Nasdaq gained only 1% and the semiconductor index rose just 4%, showing a clearly insufficient recovery.
However, Morgan Stanley, known as the stock market's "whistleblower," said the US stock selloff is nearing its end and that the coming two to three weeks are a good time to add positions. At the same time, news about the sharp shrinkage in market value of the two major US AI large-model companies continues to flood screens. Anthropic's valuation is approaching $1 trillion, nearly halved from October, and OpenAI's new funding round valuation is also shrinking.
Beyond the US "Blue Wave," a wave of protests against data center expansion is sweeping across multiple European countries. The protests stretch from Scotland to Spain, with 2,000 people gathering in Dublin and 1,300 in Helsinki on the same day, while Amsterdam saw clashes and arrests. The core conflict: Ireland's data centers already consume 23% of the nation's electricity, and a Google project in Finland faces criminal investigation for illegal logging of forest land. Protesters point directly at "Big Tech kidnapping democracy" and demand a full pause on construction before public democratic votes are held.
Nomura Securities released a research report stating: Nomura says the "Blue Wave" is hitting US data centers, while China AI actually benefits; for China, the most direct benefit of slower US data center construction is reduced difficulty in obtaining AI chips and eased price pressure. The firm believes Chinese AI companies, cloud service providers, and research institutions will be able to obtain relevant GPUs, HBM modules, and advanced packaging materials more conveniently and at lower cost through intermediary trade hubs.
In fact, last Friday everyone may have had a feeling that the national team did not step in, but retail investors spontaneously bought the dip to defend the market, and the overall decline in A-shares was not large, with the bubble already squeezed out in advance. Compared with the Philadelphia Semiconductor Index, the STAR 50 fell only 2.6%.
Of course, within the tech small-caps, A-shares showed clear divergence. On one hand, optical communications and PCBs kept falling, especially optical modules, with Zhongji Innolight plunging nearly 10% on Friday, Dongshan Precision and Cambridge Technology hitting the limit down, and New Easy Electronics, Tianfu Communication, and Yuanjie Technology all falling sharply.
According to statistics, last week Zhongji continued to crash nearly 19%, Shengyi Technology fell nearly 19%, Jingwang Electronics and Tianfu Communication dropped more than 17%, Cambridge Technology, WUS Printed Circuit, and Pengding fell more than 16%, Changxin Bochuang and O-Net Technologies dropped more than 15%, and Lianxun Instruments, Shennan Circuits, T&S Communications, Yuanjie Technology, New Easy Electronics, and Victory Giant all fell more than 10%.
On the other hand, AI applications, photovoltaics, semiconductors, biotech, and related industries showed clear gains. Especially in AI applications, Mango Excellent Media surged 56%, photovoltaic player Maxwell and energy storage player Haibo Si Chuang both rose more than 30%, and in semiconductors Hua Hong Semiconductor, Shanghai Silicon Industry, and Jiangfeng Electronics all rose more than 20%.
This means capital is gradually shifting toward AI applications, and the semiconductor track within tech small-caps is once again drawing fund attention. For the outlook, CITIC Securities said the market adjustment has already been brought forward to this year, and structural turning points are increasing next year. In the strongest phase of the AI bull market, non-AI sectors have already generally entered a downtrend ahead of time and undergone deep valuation adjustments. But precisely because of this, if AI investment slows its growth pace in the future, there is no reason to remain pessimistic on non-AI sectors. If AI enters a consolidation period, non-AI sectors may usher in a turnaround, as falling interest rates, declining material prices, and policy rebalancing all favor a more balanced market. "K-shaped convergence and broader market breadth" may be the most important market feature of 2027. We expect divergence within the tech sector, with broad structural repair opportunities in non-AI sectors, and the main process of valuation contraction already completed in 2026. Clearing on the funding side takes time, and investors will gradually return to reasonable return expectations. Stable-return strategies represented by PB-ROE are relatively more advantageous than boom strategies, and the style may undergo a major shift. In allocation, moderate diversification is needed, responding along four lines: stable returns, excellent overseas manufacturing, domestic demand repair, and new AI technology themes.
In addition, CITIC Securities believes Hong Kong stocks will see moderate profit growth in 2027, but financing unlocks and a slowdown in southbound flows will keep liquidity tight in the first half of 2027. Currently, the forward PE of the Hang Seng Index and Hang Seng Tech are 9.67 and 13.10 times respectively, at historical lows, and policy support and connectivity reform are expected to underpin a recovery. In allocation, it recommends balancing growth and dividends: for growth, position in four directions—biotech, AI applications, intelligent driving, and embodied intelligence; for dividends, use bond-like assets as ballast while also considering dividend growth.
Well-known economist Chen Li said we may hold a cautious attitude toward bottom-fishing the Nasdaq index. Over the next 12 months, US AI infrastructure investment growth faces downward pressure, and earnings expectations for upstream hardware companies still carry adjustment risk. An index rebound may occur, but I would rather place new investment opportunities in downstream applications, especially To C AI Agents that can use cost-effective models to lower service costs.
GF Strategy said the speed and magnitude of this round of decline in AI-related indices exceeded most historical samples; but after the end of the fourth month from the peak, rebounds are the norm; statistics show further downside within one year is limited, and during that period there is more than a 90% probability of a rebound of over 30%, but returning to previous highs requires earnings to accelerate again, and the current baseline scenario should still be treated as a rebound.
Sinolink Strategy: considering that the endogenous stability of Chinese assets remains solid, combined with the recent strengthening of stabilization policies, the appeal of Chinese assets is increasing. Under an AI slowdown, the path to lower capital costs is already clear. The first half of the shock comes from a simultaneous decline in capital returns, while the second half will bring global opportunities. Some signs of earnings resilience and marginal improvement have already appeared, and we recommend seeking defensive allocation clues from both macro and industry lines.
Soochow Strategy's Chen Gang believes that against the backdrop of strengthening headwinds in the AI upstream narrative, the current period can be seen as the starting point for investment in the AI midstream and downstream. Similar to how the expected headwinds for hardware are "unfalsifiable," the development prospects of AI midstream and downstream currently also have an "hard-to-falsify" quality, and compared with the upstream infrastructure chain, application sectors also have an advantage at the positioning level, deserving greater attention. Specific directions to watch: 1) AI film and television and content generation-related directions, where after a supply explosion the scarcity of IP and traffic distribution increases, with attention on high-quality IP copyright holders (including some publishing companies) and channel distribution platforms; 2) AI4S, where the core is still AI pharma, with attention on wet-lab supporting links (gene synthesis, recombinant proteins, antibody expression, etc.), integrated dry-wet AIDD drug discovery platforms, and AI-enabled drug discovery CROs; 3) Physical AI, mainly focusing on data collection-related software and hardware, followed by OpenAI, T-chain components, and domestic robot bodies; 4) On the edge side, the market has already priced in "AI computing power price increases squeezing traditional consumer electronics demand" relatively fully, and the subsequent game point lies in the increment brought by physical AI and consumer-grade edge new products, with attention on valuation repair in links such as SoC.
Guotai Haitong believes the market adjustment this year has already been relatively sufficient, with no excessively low expectations; but conditions for a strong rebound still require timing and a better trigger. In the current environment of relatively high expectations and high positioning but increased uncertainty, a strong and sustainable rebound still depends on subsequent substantive improvement. We still maintain that the timing is approaching, and better conditions will emerge around the end of the year and the beginning of next year. Currently, the forward PE of the CSI 300 (excluding electronics and communications) remains persistently low, while trading heat is also at historical lows. The steady cash flow generation ability of large-cap blue chips, the value of leading companies' moats, and stable dividend return attributes deserve attention. In addition, the AI industry cycle is not over. China's narrative of strategic catch-up as a major power, its centralized and coordinated governance structure, and lower interest rates and more accommodative financial conditions mean China's AI growth and investment are expected to develop an independent narrative.
Shenwan Hongyuan believes we should first do a good job with the oversold rebound within a major wave adjustment: short-term historically extremely high cost-performance plus efforts to stabilize capital market expectations mean A-share rebounds have their own patterns. But the fundamentals of the tech track remain headwinds, and short-term positioning structure is under pressure, neither of which supports a major wave rebound. Historical experience to reference: the magnitude and duration of an upward rebound may be limited, but the time during which new lows are not made may be longer. Structurally, thematic trends and independent-logic industries are advantageous, while tech track assets heavily held by institutions may underperform. Among clues for theme rotation and independent logic, directions with relatively high rankings in industry comparison: 1. The effectiveness of fundamental research is recovering, with attention on directions where fundamental expectations exceed mid-2026 and 2027 expectations are better than 2026, with focus on domestic semiconductors and PCBs. 2. Future directions with high attention in AI application-side industry and primary markets are embodied intelligence, medical AI, and AI security. 3. The continuation of the Democratic sweep trade is also a signal validating the strength of thematic trends. New energy and semiconductors outperforming plus the computing power chain adjusting is itself part of the market's structural opportunities. Among these three structures, there are still directions maintaining strength, which is also evidence that the rebound phase is still ongoing. Conversely, if all three structures weaken, subsequent rotating themes may become even weaker, constituting a signal that the rebound is nearing its end.
Comments