〔Day Digest〕AI Capex Faces Its Payoff Test: 493 S&P 500 Companies Show Flat Revenue per Employee as the 10-Year Yield Drops 17 bp - A-Shares Face a 3,700-Point Bottom Debate

0. Weekly Arc

The dominant fault line is whether AI spending is becoming an earnings engine or remaining a cost center, with Apollo warning on realized returns while Nomura and CITIC defend the broader opportunity ^[1][2][3]. Apollo Global Management Chief Economist Torsten Slok says revenue per employee is rising at the seven major technology companies but flat at the other 493 members of the S&P 500, with no sign yet that AI is lifting productivity ^[1].

Yet Nomura Securities expects the memory-chip shortage to persist through 2028 and says the sector is severely undervalued, while CITIC Securities argues that an AI pullback is not the boundary between bull and bear markets ^[2][3]. Source quality is uneven: the CITIC AI note and the Nomura Fed preview arrive only as headline-level briefs, so their claims carry less evidentiary weight than the fuller Apollo, Goldman Sachs, Bank of America, and foreign-flow reports ^{[1][2][4][5][6][7]}.

1. Enterprise AI Economics

• **[ONGOING] Operating evidence.** Slok says mainstream U.S. companies are reducing AI investment and treating token consumption as a cost center rather than a growth source because the spending is squeezing margins ^[1]. Meta's more aggressive AI plan and higher capital spending plan triggered a share-price decline and pushed a credit-risk metric to a record high ^[1].
• **[ONGOING] Infrastructure case.** Nomura says concerns about AI investment are overreacted to, the memory-chip shortage will run through 2028, and sector valuations are severely underestimated ^[3]. The disclosed bullish case therefore rests on a persistent hardware bottleneck and depressed valuations, not on demonstrated enterprise-wide productivity gains ^[1][3].
• **[NEW] Allocation view.** The CITIC headline says the AI pullback is not a bull-bear divide and that non-AI businesses may face a more favorable environment; the supporting detail is not supplied in the packet ^[2].
• **[NEW] Near-term test.** Investors are watching whether enterprise AI spending produces the expected return as earnings season begins ^[1].

2. Global Rates and Technology Risk

• **[ESCALATED] Bond move.** Goldman Sachs said the 10-year Treasury yield fell 17 bp on Thursday, its largest one-day decline since March 2023, and moved back below 5% ^[4]. Goldman Sachs Fixed Income, Currency and Commodities strategist Cosimo Codacci-Pisanelli attributes the move to the AI unwind: equity selling pushed capital into long-duration Treasuries and restored bonds' hedge role against stocks ^[4].
• **[ESCALATED] Fed path.** Goldman still treats a 25 bp December hike as its base case, says markets may struggle to reprice the full move, and warns that AI-related debt supply plus energy and inflation risks can rebuild rate pressure ^[4].
• **[ONGOING] Tactical stance.** Bank of America strategist Michael Hartnett recommends tactical caution through mid-November while U.S. midterms, French bond-market turmoil, and AI financing risk remain unresolved ^[5]. His tactical call is falling yields, stocks, and the dollar, with the 30-year Treasury heading toward 5% and the dollar index below 100 ^[5]. Global equity funds still took in $55.6 billion for the week and technology funds $18.8 billion, the highest technology inflow since July, but Bank of America's bull-bear indicator fell from 7.4 to 7.2 ^[5][8][9].
• **[ESCALATED] Technology correction.** Deutsche Bank's multi-asset team says technology has fallen about 7% while the rest of the S&P 500 has gained about 5%, leaving a 12-percentage-point gap and roughly 5 percentage points of room to the long-term trend-channel floor ^[10]. Positioning is not yet extreme: about $30 billion flowed into technology funds over two weeks, although the latest session saw a net outflow ^[10]. Without a tariff or geopolitical shock, Deutsche Bank views the move as sector rotation rather than a systemic U.S. equity decline, with earnings support and continued fund flows deciding the next leg ^[10].
• **[NEW] Regional read-through.** Goldman Sachs cut its 12-month Kospi target by 25% as concerns over the AI trade intensified ^[11].

3. China Bottoming and Capital Flows

• **[NEW] Index reversal.** Xiangcai Securities says the Shanghai Composite briefly fell below 3,600 on October 23 before recovering to close higher, while the Shenzhen Component and ChiNext formed bullish-engulfing patterns ^[12]. It calls the move sentiment repair rather than a full earnings-driven reversal and says the Shanghai Composite may build a stage bottom near 3,700 ^[12].
• **[NEW] Flow reality.** Guosen Securities estimates northbound funds were net sellers of RMB5.6 billion during October 19-23, versus RMB14.2 billion the prior week, while flexible foreign capital swung to RMB0.4 billion of net inflow from RMB6.8 billion of net outflow ^[6]. In Hong Kong for October 14-20, total foreign and Stock Connect flows were a net HK$3.8 billion inflow: stable-type foreign capital contributed HK$14.8 billion, flexible-type foreign capital HK$1.6 billion, and Stock Connect recorded an HK$11.8 billion outflow ^[6]. The combination is mixed: the index produced a one-day V while estimated northbound selling continued, albeit at a smaller weekly amount ^[12][6].
• **[ONGOING] Bottom framework.** Huajin Securities' history of A-share bottoming periods found strong-trend TMT in seven instances, cyclicals in two, medicine in one, and electrical equipment in one, each with excess returns versus the CSI 300 ^[13]. Its screen looks for valuation percentiles below 30%, lower fund and foreign holdings, lower margin balances, and lower turnover shares, and it favors oversold technology growth and selected cyclicals; high oil prices support petrochemicals while AI demand supports related small metals ^[13].
• **[NEW] Beijing exchange.** The BSE 50 rose 3.67% for October 19-23 and closed at 984.90, while the broader A-share week saw only seven industries advance ^[14]. On October 23, the BSE 50 gained 1.88%, turned back above its five- and ten-day moving averages, and traded RMB16.93 billion as attention increased ^[15]. The regular report says the rally still depended on a few heavyweight names, with breadth and volume insufficient ^[14].

4. Sector and Theme Positioning

• **[ONGOING] Growth rotation.** A brokerage strategy roundup indexed by Sina Finance says technology growth and high-end manufacturing remained active, with computers, machinery, and automobiles leading, while coal, property, communications, and electrical equipment pulled back ^[16]. It retains the medium-term case for AI computing and semiconductors, sees communications electronics lagging computers, and says high-end demand and copper-clad-laminate price increases support PCB makers, subject to third-quarter results, orders, and valuation ^[16]. Robotics, new-energy vehicles, auto parts, and equipment remain active but need order and earnings validation, while business-development deals support the innovation-drug outlook ^[16].
• **[NEW] Physical AI.** Guotai Haitong's theme barometer shows trading heat edging up, with embodied intelligence and other physical-AI themes leading while semiconductor-equipment, PCB, and domestic-demand themes weakened ^[17]. Theme turnover averaged RMB694 million a day at a 3.08% turnover rate, and both measures edged higher ^[17]. Catalysts include the first batch of national standards for humanoid robots and a push by the Ministry of Commerce and eight other departments into trade, logistics, and household applications ^[17].
• **[NEW] Industrial policy.** A Guotai Haitong policy index says AI-related policy supply ranked first among eight industries in 2025, with policy resources concentrated in leading provinces such as Guangdong, Beijing, and Shanghai ^[18].
• **[NEW] Biopharma.** A new market-and-capital report says China's listed biopharma industry is moving from betting on pipelines toward commercial delivery and earnings realization, while its market-cap share has declined and valuation-management tools remain limited ^[19]. The study covers 508 A-share healthcare companies and 302 Hong Kong healthcare companies over October 2021-September 2026 ^[19]. Tsinghua University Global Securities Market Research Institute secretary-general Chi Yongsheng says the sector remains at a historically low relative valuation and that investors now care more about commercialization and earnings than pipeline or clinical headlines ^[19]. The report recommends adding AI tools to R&D, conducting stage-by-stage development diagnostics, and using equity incentives more effectively ^[19].

5. Falsification Checklist

• **[ONGOING] Enterprise AI.** The operating test is whether revenue per employee and margins begin improving outside technology; if token spending remains a cost center without productivity gains, Apollo's warning remains intact ^[1].
• **[ESCALATED] Technology correction.** The rotation thesis is invalidated if a tariff or geopolitical shock turns sector selling into a systemic equity decline ^[10].
• **[NEW] China bottom.** The V-shaped-bottom case is not confirmed by price action alone and weakens if earnings do not validate the rebound while breadth and volume remain thin ^[12][14].
• **[ESCALATED] Rates.** The yield-decline thesis weakens if AI capex, debt-financing supply, and energy-led inflation revive upward pressure on yields ^[4].
• **[ONGOING] Checklist.** The next decision points are enterprise AI returns, third-quarter earnings and order delivery, technology-fund flows, December Fed repricing, and any renewed inflation or energy-price shock ^{[1][10][4][15][16]}.