〔Day Digest〕Denominator Shock: 10Y Tops 5.3%, Hyperscaler FCF Cracks -$100B, KOSPI -7.57% — A-Share 'Bottoming' Thesis Faces the $1.7T 2027 Showdown

0. Weekly Arc

The dominant story on the research-report board is convergence: at least six Chinese brokerages — Everbright, Guosen, Huajin, Shenwan Hongyuan, Guojin and the Sinolink AI investment-research group — are calling a near-term bottom for A-shares, anchored to ETF net subscriptions and a -40% drawdown template ^{[1][2][3][4][5][6]}. The contradictions are external: US 10Y above 5.3% and 30Y at 5.73% have triggered a KOSPI -7.57% single-day move and a Philadelphia Semiconductor six-day losing streak, transmitting a "denominator" valuation shock into high-duration growth ^[7]. Two clocks must be outrun — BofA's hyperscaler FCF math (<-$100bn combined 12-month) and Michael Hartnett's Bull & Bear indicator at 7.4, which has ended a five-month sell signal but is not yet a buy ^[8][9][10].

1. The A-Share 'Bottoming' Consensus

• **[NEW] Everbright Securities:** market "may gradually stabilize" as ETF flows turn to net subscriptions, framed as medium- and long-term capital stepping in, but the same note flags Iran-US uncertainty, sticky US yields and a lack of fresh AI narrative ^[1].
• **[ESCALATED] Guosen Securities:** tech chip-clearing and sentiment release are "more than halfway through"; leveraged capital, public equity funds, short-term foreign and insurance have cleared 10-20%, while private, long-term foreign and "fixed-income plus" have cleared 30-40% ^[2]. Three watch items: industry catalysts, chip structure, overseas liquidity ^[2].
• **[NEW] Huajin Securities:** A-shares may short-term bottom; the 11 historical bottoming episodes since 2010 averaged 49 trading days, with domestic policy as the core driver and at least one of manufacturing PMI, industrial profit, property sales, retail or exports turning up ^[3].
• **[NEW] Shenwan Hongyuan Strategy:** short-term adjustment end signal has appeared; AI industry lacks universal computing-power inflation and overseas CSP credit/liquidity pressure is transmitting down the capex-inflation chain, while October capital supply/demand is unfavorable ^[4].
• **[NEW] Guojin Strategy:** the worst environment has passed; AI investment pushed up rates and global capital costs, but as AI narrows to US tech, US tech's own cracks should support a real-rate decline and a gold window ^[11].
• **[NEW] AI Investment Research Group model:** bonds remain bullish, equity adjustment risk not yet cleared; gold and silver models stay bearish (gold 30 days, silver 15 days) ^[6].
• **[ONGOING] Divergence:** Guosen's north-bound tracker estimates a 142bn yuan net outflow last week with flexible foreign capital alone at -68bn yuan, partially offsetting the broad-ETF inflow narrative ^[12].

2. The External Denominator Channel

• **[ESCALATED] US rates:** 10Y above 5.3%, 30Y at 5.73% after US September CPI at 3.7% headline and 2.6% core both beat expectations; Philadelphia Semiconductor's six-day losing streak transmitted to KOSPI's -7.57% single day ^[7]. SW Level-1 sectors closed 2 up / 29 down, with food & beverage +1.89% and bank +1.52% the only gainers and communications -10.87% the worst ^[7].
• **[ESCALATED] Geopolitics:** the Middle East stalemate keeps refined-product prices elevated; insufficient refining capacity, high transport costs, low inventories and the European heating season sustain an inflationary impulse ^[13].
• **[NEW] Mechanism read:** a SW strategy tracker frames the week as a "denominator" valuation kill rather than earnings deterioration; high-duration growth is being systematically repriced while dividend/defensive shares hold ^[7].

3. AI Credit Math and Hartnett's Pivot

• **[NEW] BofA (hyperscaler free cash flow):** combined 12-month FCF has fallen below -$100bn, with capex consuming 95% of operating cash flow; by contrast, Nvidia, Micron and Broadcom 12-month FCF has surged to near $500bn ^[8]. Hyperscaler WACC has jumped from 4% in 2023 to 10%, and since December 2024 the market has revised 2027 capex up 81% while AI-related revenue consensus is up only 24% ^[8].
• **[NEW] BofA (Michael Hartnett):** Bull & Bear indicator fell 0.7 points to 7.4, ending the 5-month sell signal triggered May 26, but tactical stance stays "defensive-leaning" ^[9][10]. Two preconditions for a buy — a Fed October 28 hike to restore credibility and AI funding relief — are currently unmet, with October hike odds collapsing from >50% to 14% ^[9]. CoreWeave 5-year CDS implies ~63% default; Nasdaq 100 fell nearly 6% on the week, the worst since "Liberation Day" ^[9].
• **[ESCALATED] BofA flow tape:** tech funds absorbed $10.7bn last week with US large caps the primary target; global equities took in $64.5bn while cash saw $53.1bn in outflows ^[14][15].

4. The 2027 Showdown

• **[NEW] Goldman Sachs vs Paul Tudor Jones (relay):** 2027 brings roughly $1.7T of US stock unlock plus ~$600bn new issuance — the largest net public-float increase since 2000 — set against ~$1.7T in projected buybacks and >$2T in total corporate demand ^[16]. Goldman: supply pressures valuation but does not end the bull; Tudor Jones: 2000-2002 parallel, may short for a year ^[16].
• **[NEW] Citi (relay):** S&P 500 fair P/E band 17.1-22.1x versus 22.4x current; AI profit growth and shifting investor structure are weakening the rate-to-equity linkage, but the marginal buyer is the unresolved question ^[17].
• **[NEW] 2027 parallels:** a sector teaser flags a 2027 "investment-revaluation first year" with cash-flow improvement, earnings recovery and AI demand as the trinity ^[18]; Goldman separately projects higher 2027 average mortgage rates will depress US housing sales ^[19].
• **[NEW] BofA (dollar):** the dollar rally has stalled, with the source of selling to be confirmed ^[20].

5. What Would Falsify the Bottom Call

• **Chip prices:** Hartnett flags a chip price drop as the marker for AI credit tightening and capex-cycle peak; a sustained Philadelphia Semiconductor bounce would be the first confirmation of a buy ^[9].
• **Hartnett's two preconditions:** a Fed October 28 hike (now 14% priced) and AI funding relief — without both, the 7.4 print is a pause, not a pivot ^[9].
• **Bond-equity rotation:** the AI Investment Research Group model still flags equity adjustment risk not cleared despite a 30-day gold bearish streak and a confirmed bond bullish model; a flip on either would be the structural tell ^[6].
• **Sourcing flag:** the dollar-stall note ^[20] and the 2027 sector-revaluation teaser ^[18] are single-source headlines with no detail — treat as thin. The Tudor Jones short thesis is relayed via a magazine summary, not a primary source ^[16].