〔Overnight Brief〕Goldman: Earnings, Not Yields, Anchor the Bull as 30Y Hits 5.3% Near 20-Year High; WTI Past $100 on Hormuz; Fed September Hike Odds Top 80%

0. Overnight Arc

A Goldman research note reframes the 'high-rate scare' as a valuation story rather than a bull-killer: the 30-year US Treasury yield at 5.3% (near a 20-year high) and 10-year near 5% have already compressed S&P 500 forward P/E from 22x to 19x, with the equity risk premium stable at 3% ^[1]. Chinese strategy desks, however, read the cross-asset setup as unambiguously hawkish — Fed September hike probability above 80% on hot US August data, WTI clearing $100/bbl on US-Iran tensions and Hormuz disruption ^[2]. The test of Goldman's 'earnings, not yields' thesis is whether A-share and US earnings can hold while yields and oil keep climbing ^[1][2].

1. Goldman vs the High-Rate Crowd

• **[NEW] Goldman Sachs (Sept 11 report):** high rates are a headwind, not a bull-market terminator; the anchor is earnings growth and unusually strong corporate balance sheets ^[1]. Historical reference: the S&P 500's average return in the 12 months following the start of a Fed rate-hike cycle is +9%, against a typical early-phase drawdown of ~2% ^[1]. Goldman economists expect a 25bp hike 'next week' ^[1].
• **[NEW] Positioning call:** avoid long-end-rate-sensitive homebuilders, embrace financials and high-growth names; AI capex, M&A and spin-offs are cited as the earnings levers ^[1].

2. Rates, Oil and the Fed Hike Setup

• **[NEW] Rates:** 30-year US Treasury yield at 5.3% (near 20-year high), 10-year near 5%; S&P 500 forward P/E has absorbed the move, falling from 22x to 19x ^[1].
• **[NEW] Oil and shipping:** WTI above $100/bbl as US-Iran tensions escalate from late August, with Hormuz Strait passage obstructed ^[2]. The oil move is the proximate driver of the renewed Fed-hike pressure ^[2].
• **[NEW] Fed path:** market-implied probability of a 25bp September hike has risen above 80% on stronger-than-expected US August jobs and inflation data; rate-hike pricing has already shown up in US bonds and equities ^[2].

3. A-Shares: Preventive Hike, Third-Stage Bull

• **[NEW] vReport macro+strategy desk:** the current Fed cycle is classified as 'preventive' rather than 'anti-inflation' — eight US hike cycles since 1982 are split using a US inflation-pressure index, with anti-inflation episodes including 1983-84, 1987-89, 1999-2000 and 2022-... ^[2]. Under a preventive hike, A-shares are not mechanically bearish; the binding variables are domestic macro/micro repair and non-extreme sentiment ^[2].
• **[NEW] Stage and sectors:** A-shares are placed in the 'third stage' of a bull market; sector cues highlight an AI hardware second wave 'brewing,' with AI applications, resource dividends, real estate and consumption also flagged ^[2].

4. Industry Cycle: Expansion Outrunning Recovery

• **[NEW] 26Q2 industry landscape tracking:** sectors are broadly progressing, with TMT and midstream manufacturing in expansion and upstream materials in recovery; consumer is moving from downcycle toward clearing ^[3].
• **[NEW] Historical rule with a 2026 kink:** long-term backtests from 2010 favor entering in recovery, with expansion second-best; 'bottom-grinding' offers limited allocation value, and contrarian clearing-period bets warrant caution ^[3]. But 26Q2 shows expansion beating recovery — a deviation attributed to TMT/tech leadership, last seen in 2021 Q1-Q3 ^[3]. For expansion-stage TMT names, the watch-item is the second derivative of revenue growth across consecutive periods ^[3].

5. Tail Read: The Decoupling War Game

• **[NEW] South China Morning Post relay:** two researchers based in Guangdong province used a quantitative trade model to 'war game' a full US-China economic decoupling; the counter-intuitive finding is that most other countries would gain ^[4]. The excerpt is truncated at 'a number of other countries, including...', so the named beneficiaries and the mechanism (trade reallocation, tariff incidence, supply-chain rerouting) are not in the packet and remain unverified ^[4].