〔Overnight Brief〕AI Narrows From a Narrative Trade to an Earnings-and-Cash-Flow Test: CoreWeave’s Target Drops From $150 to $100 and Nvidia Leaves Two Top-Pick Lists — Ratings Stay Constructive

0. Weekly Arc

• **[ONGOING] Weekly arc:** AI remained central across 2027 strategy meetings involving CITIC Securities, Huatai Securities, Caitong Securities, China Securities Co., Zheshang Securities, and Kaiyuan Securities, with the long-term trend described as intact but the investment phase shifting toward “selective” stock picking as low valuations, marginal liquidity improvement, and earnings repair were expected to support a new A-share repair rally ^[1][2].
• **[NEW] Evidence quality:** Five packet entries are headline-only or title-level snippets rather than full report summaries ^{[3][4][5][6][7]}.
• **[NEW] Research breadth:** Since the start of November, brokers had surveyed 328 listed companies, concentrated in electronics, machinery, and pharmaceuticals and biotechnology, while changing 18 ratings, already close to October’s full-month total ^[8].
• **[ESCALATED] Election overlay:** Morgan Stanley saw no directional easing in US-China policy, while a separate post-election analysis said roughly three-quarters of 2027 added compute had already started, making unstarted 2028 projects the main timing risk ^[9][10].
• **[NEW] Macro counterweight:** US October headline CPI rose 0.4% m/m versus a 0.3% estimate and 0.6% prior reading, while core CPI stayed at 0.3% m/m for a third month; the result lifted rate-hike expectations and pressured equity futures, Treasury yields, and precious metals ^[11].

1. Policy and Project Clock

• **[NEW] Nomura scenario:** A Democratic “blue wave” could materially benefit China’s domestic AI-infrastructure investment while producing only limited pressure on some AI export channels; in the first nine months of 2026, dollar-denominated Chinese integrated-circuit exports rose 108.1% y/y and contributed 5.6 percentage points to 21.3% overall export growth, while most AI components and other politically sensitive goods were already barred from direct US sales under existing export controls ^[12].
• **[ONGOING] Morgan Stanley read:** The bank saw no sign of policy easing and expected strategic competition and geopolitical uncertainty to persist, treating greater congressional constraints and perhaps better predictability—not friendlier policy—as the possible positives ^[9].
• **[NEW] Project timing:** Around three-quarters of 2027 new compute capacity had already started, so the election impact falls mainly on unstarted 2028 projects and depends on how long state approval pauses last rather than on whether limits are imposed ^[10].
• **[ESCALATED] Capacity risk:** Morgan Stanley estimated that 20% of new US data-center capacity planned for 2028 could slip to 2029, or about 5 GW equal to 10% of global 2028 additions, potentially creating a supply-chain gap and inventory risk for legacy memory ^[13].
• **[NEW] Financing mechanism:** CICC Managing Director and chief overseas and Hong Kong strategy analyst Liu Gang said AI can support real GDP, corporate earnings, productivity, and knowledge spillovers, but it can also crowd out electricity and financing, raise borrowing costs in traditional demand, and constrain the Federal Reserve’s room to cut rates for housing ^[14].
• **[NEW] Dispersion implication:** The common thread is tighter project timing and higher financing sensitivity, favoring companies that can convert spending into cash flow while increasing the penalty for delay ^[9][10][14].

2. AI and Corporate Signals

• **[NEW] Consumer electronics:** Several representative companies saw third-quarter gross margins fall and profits come under pressure, with memory-price increases, FX volatility, Apple’s adjusted product-release cycle, and AI cited among the factors; most still posted revenue growth as they built second and third growth curves in AI glasses, robots, gimbal cameras, automobiles, servers, and optical communications ^[15].
• **[NEW] Physical AI:** Analysts identified the scarcity of high-quality real-world interaction data capable of continuous scaling as a physical-AI bottleneck and favored the data supply chain ^[16].
• **[NEW] CoreWeave:** Evercore ISI cut its target price from $150 to $100 while retaining an Outperform rating ^[17].
• **[ESCALATED] Nvidia:** Morgan Stanley removed Nvidia from its semiconductor top-pick list after Bank of America removed Nvidia, Micron Technology, and Lam Research from its “US 1” list, while both banks retained Buy or Overweight ratings; Morgan Stanley said uncertainty around CEO Jensen Huang had weakened the risk-reward ^[18].
• **[NEW] Price reaction:** Nvidia fell 4.3% and 5.47% over two consecutive sessions, a cumulative loss near 10% and about $470 billion of market value, before rebounding more than 5% in premarket trading ^[18].
• **[ESCALATED] Legacy memory:** Morgan Stanley warned that the old-memory pricing tailwind was ending as cloud-capex uncertainty and weak consumer demand pushed SLC NAND, MLC NAND, NOR, and DDR4 toward balance around 2028; it cut ratings, price targets, and 2027-2028 earnings estimates and used 1.0x P/B as a downside anchor ^[13].
• **[NEW] Storage split:** Goldman Sachs analyst James Schneider said memory stocks often bottom near tangible book value and rated Samsung and SK Hynix Buy, although their shares were below book value, a different condition from being below tangible book value; SK Hynix’s US ADR rose about 8% on Wednesday but traded at a more than 40% premium to its Korean listing, while Micron stayed Neutral with a $1,250 target and about 5.4x tangible book ^[19].
• **[NEW] Optical switching:** University of Illinois Urbana-Champaign and Notre Dame researchers reported a 5-micron nitride quantum-well microLED optical switch with 2.75 GHz modulation bandwidth and 25 mW power; with about 90% of data-center communications within 10 meters, it was positioned as a lower-power alternative to lasers ^[20].

3. Allocation and Corporate Calls

• **[NEW] A-share setup:** Brokerages expected more structural opportunities in 2027, with a stronger second half potentially ahead and a possible rebound-retest-recovery path; however, China Securities Co. senior strategy analyst Xia Fanjie framed the year as falling earnings plus a falling discount rate and an N-shaped path, while a separate six-broker strategy synthesis cited earnings repair as a rally driver ^[1][2].
• **[NEW] Hong Kong allocation:** Multiple brokerages saw 2027 Hong Kong fundamentals and valuation potentially repairing together, with liquidity shocks easing but still a pressure; the allocation sequence was dividend stocks first to absorb volatility, then growth exposure as liquidity expectations improve, including biotechnology, AI applications, internet, autonomous driving, and embodied intelligence ^[21].
• **[NEW] Diversification:** BlackRock China head Fan Hua argued that Chinese equities, bonds, and overseas assets have low correlations and different economic and growth cycles; he said more than $200 billion of BlackRock’s $15.5 trillion in assets was invested in China, leaving substantial allocation room ^[11].
• **[ESCALATED] QDII dispersion:** By November 10, the average year-to-date QDII return had fallen from 4.46% to -4.6%, the number of products down more than 20% had widened from 47 to 69, and 139 of 339 funds with data were positive, or 41%; some interviewees said allocation may shift toward defense and diversification ^[22].
• **[NEW] China-Germany operations:** A German Chinese Chamber and KPMG Germany survey of more than 110 Chinese companies found that 40% expected revenue growth over the next 12 months, 43% planned to add staff, 30% planned to expand investment, and 27% planned production investment over five years; 34% conducted R&D in Germany, 28% collaborated with German companies on R&D, and 26% collaborated on manufacturing, with all three shares higher than in the prior survey ^[23].
• **[NEW] Airlines:** Morgan Stanley cut Air China, China Eastern, and China Southern A shares to Underweight, with target prices of CNY 4.30, CNY 2.20, and CNY 3.23 respectively, while the supplied report item provided no rationale ^[24].
• **[NEW] Defensive consumer:** Deutsche Bank upgraded Unilever from Hold to Buy and raised its target from 5,300 pence to 5,350 pence, arguing that long-term transformation and liquidity could offset AI-related risks ^[25].

4. Falsifiers and Decision Points

• **[ONGOING] AI earnings test:** The selective-AI thesis fails if revenue and free cash flow do not catch up with the financing required to build the ecosystem, because AI can crowd out electricity and capital while highly leveraged new cloud providers are most exposed to project delays ^[10][14].
• **[NEW] Project test:** Any extension of state approval pauses would worsen the supply-chain risk already implied by the expected 20% shift of US 2028 data-center capacity into 2029, particularly for highly leveraged cloud operators ^[13][10].
• **[NEW] Liquidity test:** The valuation-repair case faces a higher hurdle because the 0.4% headline CPI surprise and third consecutive 0.3% core CPI print lifted Federal Reserve hike expectations and pressured equity futures, Treasury yields, and precious metals ^[11].
• **[NEW] Breadth test:** QDII evidence favors dispersion rather than one global beta, because some Hong Kong high-dividend and healthcare products moved from losses to gains while the broader average turned negative ^[22].
• **[NEW] Evidence threshold:** The packet does not provide the underlying Goldman Sachs Fed forecast, Baird downgrade rationale, Teradyne opportunity assumptions, or full Nvidia target-cut analysis, so none is used to set the allocation conclusion ^{[3][4][5][6][7]}.