CICC: How should we view the trends in US equities and the AI hardware industry chain?

CICC released a research report stating that, in terms of computing power chips and complete server systems, demand driven by domestic data centers remains robust, with the short-term bottleneck still on the supply side. A near-term adjustment in the pace of overseas development provides a time window for domestic computing hardware to catch up, and related companies' earnings are expected to continue to be released. On production factors, the core medium- to long-term localization logic for wafer fabs, semiconductor equipment and components remains unchanged. This year, affected by AI sentiment, the large manufacturing sector's share prices have corrected significantly in the short term, and valuations have become attractive. CICC remains bullish on the vigorous development of the domestic computing power industry chain during the 15th Five-Year Plan period. Overseas strategy: How should we view the US stock market selloff? On US Eastern Time November 5, the three major US indicesthe S&P 500, Nasdaq and Dow Jonesfell 4.4%, 5.5% and 3.4%, respectively, all posting their largest single-day declines since the April 2025 tariff shock. The AI industry chain led the declines, with the Nasdaq 100 falling more than 6% and the Philadelphia Semiconductor Index plunging about 12%, both marking their largest single-day drops since March 2020. Chart: The three major US indicesthe S&P 500, Nasdaq and Dow Jonesfell 4.4%, 5.5% and 3.4%, respectively Source: FactSet, CICC Research Department. How to view the US stock market selloff and its subsequent trajectory. CICC's brief commentary is as follows: 1. Characteristics of this selloff: significant declines, broad-based selling, signs of panic and stampede 1) Significant declines: Both the Philadelphia Semiconductor Index and the Nasdaq 100 posted their largest single-day declines since March 2020, falling 12% and 6%, respectively. 2) Broad-based selling: Unlike previous AI cooling episodes when safe-haven sectors rose modestly, all 11 US sectors fell yesterday, including safe-haven consumer staples and utilities, a sign that a stampede may be occurring. Chart: All 11 US sectors fell yesterday Source: FactSet, CICC Research Department 3) Rising volatility and panic: The VIX rose to 28 (above 25 is a dangerous level, second only to the Iran situation earlier this year). The gap between VIX spot and three-month futures widened to the largest since April 2025, which is often a direct reflection of market panic and even a precursor to further pressure. Chart: VIX rose to 28, second only to the level during the Iran situation earlier this year Source: Bloomberg, CICC Research Department. 2. Causes of this selloff: AI regulation amplified by sentiment The direct trigger for this decline remains market panic that Democrats, after winning the midterm elections, may increase AI regulation and halt data center construction. Newly elected Democratic Texas Governor Gina Hinojosa said in a media interview on Thursday that "Texas will enact laws to protect citizens from the impact of data centers"[1]; at the federal level, Sanders also renewed calls for a nationwide moratorium on data center construction [https://www.cnbc.com/2026/11/05/democrats-data-center-restrictions-midterms.html]. Policy proposals at both the state and federal levels fermented together, intensifying market concerns about the pace of AI infrastructure buildout and the sustainability of investment. The market has swung back and forth over the past two dayspre-market futures fell sharply on voting day on the 3rd, then rebounded and recovered, with global markets rallying, before today's sharp decline. As CICC noted in "Asset Implications of a Democratic 'Sweep'," many Democratic agenda items may have completely opposite effects on the same asset, and with uncertainty over which can be advanced first and whether they can be advanced at all, global markets "choose" a direction to tradefor example, both rallies and selloffs can find plenty of explanatory reasons. Essentially, the core impact of the midterm elections on asset pricing lies in whether the market's dominant force is checks and balances aimed at "course correction" (inflation, Iran situation, Fed, tariffs) or disruption oriented toward "opposing for the sake of opposing" (AI, fiscal policy). The day before yesterday, the market traded the former; yesterday, it traded the latterbut this does not mean the market can judge in advance which direction will develop first. Take AI regulation concerns: local-level pressure indeed warrants attention, but Democrats' Senate seat advantage is still insufficient to push federal-level AI regulation legislation through. 3. How to view the market at this level? Focus on liquidity in the short term; stabilization requires clearing and catalysts; the market is focusing on gold and the Hang Seng Tech Index Combined with the selloff characteristics analyzed above, regardless of the cause and catalyst, and whether subsequent concerns materialize or not, short-term price action is no longer driven by fundamental logic. Whether the emergence of panic and stampede will lead to further forced liquidations and liquidity risks will take time to manifestfor example, CICC has already seen liquidation announcements from some hedge funds. US equity leverage had previously risen to a record high, with the implied leverage level from FINRA margin balances rising to a record high of 3.4, and the CBOE put/call ratio on open interest also rising to 1.33, a new high since end-2018. This also means the market is more sensitive to price and volatility changes; once a rapid decline triggers margin calls, passive deleveraging and derivatives hedging adjustments, selling pressure may spread from the AI sector to other holdings, creating a mutual reinforcement between falling prices and deleveraging. Chart: Implied leverage from FINRA margin balances rose to a record high of 3.4 Source: Haver, CICC Research Department. Chart: The CBOE put/call ratio on open interest also rose to 1.33 Source: Haver, CICC Research Department For now, the stress remains confined to the stock market. If both US Treasury yields and the dollar are high, that would warrant greater vigilance, as it would mean the market is selling safe assets including Treasuries to meet funding gaps, and liquidity risk would rise sharply. AI is currently in a transitional lull, and the key problem is the lack of new narratives and catalysts from demand and models, which also amplifies credit risk and regulatory pressure risks. Combined with high-intensity capital expenditure continuously squeezing cloud providers' free cash flow, concerns about the pace of AI revenue realization and return on investment are mounting. Market stabilization requires natural clearing or a powerful catalyst to reverse pessimistic expectationseither way, patience is needed. Chart: Overall US equity valuations are not currently at extreme highs Source: Bloomberg, CICC Research Department On timing, the first half of next year is a key verification period; the market needs to see something like Anthropic's better-than-expected results and actual business breakthroughs in Q1 this year to boost confidence. Otherwise, as market patience runs out, the outlook for 2028 capital expenditure and returns will further dim, and the market's main trading theme may shift to recession trading. Chart: Segments where growth expectations were previously more fully priced in have corrected more, while cloud providers have been relatively resilient Source: FactSet, CICC Research Department On structure, AI may remain the main theme, but within it, greater emphasis should be placed on segments with more sustained and certain profitability. Overall US equity valuations are not currently at extreme highs, but there is clear divergence within the AI industry chain. In this AI rally, hardware infrastructure segments such as network interconnect/optical modules, semiconductor manufacturing, and liquid cooling/thermal management have priced in higher earnings growth expectations, while cloud providers are relatively reasonably priced. Recent market performance also reflects this divergence: segments where growth expectations were previously more fully priced in have corrected more, while cloud providers have been relatively resilient. This also means cloud providers' pricing still leaves room for future earnings growth, but if earnings in segments such as optical modules cannot sustain high growth, valuations face further downside risk. Observing the market's performance over the two days around the midterm elections, the path of least resistance in market trading has gradually converged on the Hang Seng Tech Index and gold. 1) Hang Seng Tech has less AI exposure, so there is less need to worry about AI risk, and expectations of easing geopolitical risk leading to lower oil prices and inflation could also reduce rate-hike pressure. 2) Gold benefits both from rising political gaming risk and from lower oil-price inflation reducing rate-hike pressure and weaker dollar and Treasury yields. However, the Hang Seng Tech Index's original problem lies not externally but internally; its "advantage" is that it is sufficiently low and has small AI exposure, but a rebound relying solely on external environment improvement and capital reallocation faces resistance from fundamental pressure and re-escalation of geopolitical risk. Over the past two days of rebound, large net selling via southbound flows and rising short-selling volume in Hong Kong stocks both indicate growing market divergence ("How Does the Market Bottom?"). By contrast, gold faces less resistance, with various factors still providing support. Tech hardware: Short-term headwinds for overseas AI infrastructure; continue to be bullish on domestic computing power development during the 15th Five-Year Plan period Recently, driven by the fact of slowing ARR growth at leading AI labs, overseas market investors have begun to persistently worry that current high returns on AI infrastructure investment are unsustainable, and they also hold a pessimistic view on North American CSP capex growth expectations for 2028. After the US midterm elections concluded on November 5, US stocks last night further reflected market concerns about increased US government AI regulation and a slowdown in AI infrastructure investment, with the Philadelphia Semiconductor Index falling more than 10% that day and the Nasdaq falling more than 5%. Looking back at A-shares, since July most Chinese AI hardware industry chain names (including overseas-chain and domestic-chain names) have corrected more than 50%. CICC believes that near-term North American data center construction will change the slope of hardware demand but will not change the long-term prosperity trend for now; domestic computing power demand may be even stronger. For overseas-chain names, CICC believes that valuation premiums for relevant companies have been significantly digested, the medium- to long-term AI prosperity remains, and the market's oversold decline reflects concerns about the pace of overseas AI data center construction. According to Wind data, CICC calculates that the SW Communications Equipment Index's current PE (TTM) is 28.92x, below the 10th percentile of the past three years. CICC believes the recent sharp drop in overseas-chain AI stock prices has fully squeezed out the valuation bubble; what remains to be clarified is the potential change in the pace of overseas AI data center construction and its impact on core suppliers' earnings realization. But after the short-term funding-side shock amplified volatility, the market's concerns about construction pace and earnings realization have been significantly released. In terms of actual numbers, the combined capex of the top four US cloud providers grew 84.2% year-on-year in 3Q26, while Microsoft Azure/AWS revenue grew strongly year-on-year over the same period, showing demand continues to convert into revenue. Looking ahead, CICC believes 2027 will still be a year of active investment by cloud providers.