〔Day Digest〕5.148% 10Y Forces a Risk Repricing — Fed Cut Path Splits June 2027 vs 'No Sustained Cycle,' AI Capex Peak Diverges Li Bei's Mid-2027 vs CLSA's 2030 Shortage, Gold Eyes $5,000 vs Sub-$4,300

0. Weekly Arc

The US 10Y at 5.148% — highest since 2007 — is now the dominant force on the tape, dragging risk assets and gold below $4,300/oz ^[1][2][3]. The Fed-path debate is no longer a single line: June 2027-cut, H2 2027-cut, and "no sustained hike" views sit side by side ^[4][5][6][7]. On AI capex, two signature calls collide: Banxia Investment founder Li Bei expects a second leg of AI declines when quarterly capex peaks around mid-2027 ^[8], while CLSA argues compute supply-demand balance will not be reached until 2030 ^[9]. Net: a rates-led risk repricing layered over a divided AI capex call. Source-quality flag: the 10Y/5.148% print and the Tickmill risk-asset view are single-broker items ^[1]; the Meta target raise is a single-house line ^[10]; JPMorgan's "true warning line" yield band is a headline without disclosed range ^[11].

1. Rates and the Fed Path

• **[ESCALATED] Treasury yields vs. risk assets:** US 10Y intraday 5.148%, the highest since 2007, with LSEG data showing strong growth plus inflation concerns have the market pricing multiple hikes from major central banks over the next year ^[1]. Tickmill Group analyst Patrick Munnelly: "if yields keep climbing and refresh cycle highs, even AI and trade tailwinds cannot support risk assets" ^[1].
• **[NEW] Fed cut-path divergence:**
• **Citi:** no cuts until June 2027, with the Fed holding in both October and December 2026 ^[4][5].
• **Morningstar:** two cuts in H2 2027, fed funds at 2.50%-2.75% by end-2028 ^[6].
• **Goldman Sachs Asset Management:** "does not expect the Fed to enter a sustained hiking cycle," citing fading tariff and energy price pressures, little overheating, and anchored inflation expectations ^[7].
• **Money markets (LSEG):** +37bp priced across the two remaining 2026 Fed meetings ^[7].
• **[NEW] JPMorgan watch level:** flags a US Treasury yield band as the "true warning line" for US equities; the precise range is not disclosed in the headline and should be sourced before quoting ^[11].
• **[NEW] Dongfang Jincheng (Golden Credit Rating) on gold:** the September dot-plot hawkish signal still weighs, but rate-hike exhaustion, slowing yield upside, and structural central-bank/ETF demand leave gold in a weak-oscillation regime ^[2].
• Contradiction is content: money markets still price 37bp of hikes ^[7], Citi/Morningstar push the next cut to mid-2027 or later ^[4][5][6], and GS AM argues there will be no sustained cycle ^[7] — the band is wider than any single point.

2. AI Capex: Two Camps

• **[NEW] Li Bei, founder of Banxia Investment:** the current AI decline is a valuation event, not a Davis double-kill — 2027 consensus earnings were revised up on capex mark-ups, and 2027/2028 capex is likely to peak; the second leg of AI declines will come "when capex truly peaks, around mid-2027," with Q2 next year the most likely pivot and Q3/Q4 the fallback ^[8].
• **[NEW] CLSA / CITIC Securities (HK) Wealth — "AI: The biggest questions at IF26":** rejects the 2028-peak consensus; their model shows compute shortage persisting to 2030, with semiconductor stocks still cheap after cyclical adjustment and SaaS still rich; Agentic AI and consumer-friendly interfaces (e.g., Meta's Muse) skew risk to the upside ^[9].
• **Falsifiable test:** the trajectory of hyperscaler capex into H2 2026 — if quarterly prints stay above Q2 2026 levels, Li Bei's mid-2027 timeline slips right; if compute-supply additions accelerate, CLSA's 2030 balance view is at risk ^[8][9].
• Source-quality flag: both are signature views from a single investor and a single-broker report; the AI capex-to-revenue ratio in ^[12] is 59.2% in 2023 vs a higher level in H1 2026 and is the cleanest cross-check on the camp debate ^[8][9][12].

3. Single-Stock Calls

• **[NEW] JPMorgan:** raises Meta target from $820 to $920 ^[10].
• **[NEW] Citi on Xiaomi 18 Pro / Pro Max:** Pro at 5,999 yuan and Pro Max at 6,999 yuan — "lower than comparable peers with comparable or superior specs"; maintains Buy with HK$34 target; next catalysts are shipment data and Q3 2026 results ^[13].
• **[NEW] Dongfang Securities on Shanxi Fenjiu:** maintains Buy, 149.53 yuan target at 19x 2026E P/E; forecasts 2026-2028 net profit of 9.6bn / 10.6bn / 12.2bn yuan ^[14].
• **[NEW] Zhongyou Securities on Weiergo (first coverage):** Buy; Q2 2026 net profit 374.9m yuan (+63.82% y/y vs Q2 2025) on the back of copper-cl laminate cost pass-through; AI-server power PCB leader with thick-copper HDI positioning; 2026-2028 revenue forecast 2.7bn / 5.3bn / 8.4bn yuan, net profit 320m / 960m / 1,490m yuan ^[15].
• **[NEW] CITIC Securities morning call:** 2026 may be the turning year for liquid cooling industry order-to-earnings conversion ^[16].
• **[NEW] CLSA / CITIC Securities (HK) on Maybank and CIMB (IF26):** both target 13-14% ROE by 2030; Maybank to keep ~70% payout ratio, CIMB to flex capital by exiting weak units ^[17].
• **[NEW] CLSA on China Mid-Autumn + National Day travel:** combined 10-day public holiday (13 days with leave); early flight bookings +44-50% y/y, hotel pre-bookings +40% in hot cities, but the longer window inflates apparent growth and the booking rate is slowing — Flight Master projects 31.02m air passengers over Sept 25-Oct 7, daily +3.1% y/y ^[18].
• **[NEW] Financial Street Securities on Beijing Stock Exchange H1 2026:** 339 listed firms; mid-manufacturing leads with 138 firms and 312.37bn yuan market cap (37.12%); smart-tech 2026H1 net profit +177.47% y/y; mid-manufacturing and consumer-services net profit turned negative ^[19].

4. Cross-Asset Tail Risks and Strategy Roundup

• **[NEW] Morgan Stanley (Martijn Rats et al.) on a US diesel export ban:** not the base case, but if imposed, refiners would cut runs by ~2.0 mb/d, with gasoline supply falling ~0.65 mb/d even after maximizing gasoline yield — counter-intuitively lifting retail gasoline prices; "as long as a US diesel ban is still in debate, the oil market should brace for further volatility" ^[20].
• **[NEW] TD Securities (Ryan McKay) on gold:** gold broke $4,300/oz but downside is limited; expects a retest above $5,000/oz in 2027 on resurgent investor and central-bank demand ^[3].
• **[NEW] BNP Paribas Q3 2026 Eco Perspectives:** economic and financial forecasts refreshed as of September 2026 ^[21].
• **[NEW] CEPR / VoxEU on Europe:** the investment debate should shift from "how much capital can be mobilised and how quickly" to "who receives capital" — combine simpler administration with stronger analytical capacity to spot anomalies early ^[22].
• **[NEW] Strategic minerals (research report #2):** 21 provincial 15th Five-Year plans reframe critical minerals as cycle/strategic/constraint assets; Inner Mongolia targets 100+ prospecting targets; Guizhou's "461 project" covers coal, phosphorus, aluminum, manganese ^[23].
• **[NEW] AI and taxation (research series #2):** AI exposure, labor share, and tax-dependence metrics put Germany, the US, and Japan at the top of tax-system risk; the tax-response cycle has compressed from 80+ years to ~10 years across the four industrial revolutions ^[24].
• **[NEW] AI revolution series #1:** five hyperscalers' capex-to-observable-AI/cloud-revenue ratio rose from 59.2% in 2023 to a higher level in H1 2026; low vacancy and high pre-lease argue against systemic overcapacity — the cleanest cross-check on the Li Bei vs CLSA capex debate ^[12].
• **[NEW] Sina Finance daily report picks:** flagging a new-strategic-industries investment boundary report, AI-hardware scarcity with energy-price retracement, and global AI safety governance acceleration ^[25].