Research Digest — last 24h

Verdict

This is a rotation and right-tail multiple compression within AI, not a fundamental break of the AI trade; de-rated semis/hardware now have limited fundamental downside absent a new negative leg (P001, P014).

Analysis

GS's Privorotsky frames the 15 Sep move as a de-grossing event: hardware fell 500-800bps depending on vertical, with anything tied to the right tail of AI capex hit hardest, while software went vertical and Meta/Alphabet rallied (P001). The causal mechanism: where earnings and multiples relied most on continued acceleration of incremental infrastructure growth, those assumptions are judged too aggressive, so the market compresses the right tail first. Contrary evidence is Asian price action — SoftBank +7.5% and TSMC barely touched — which is hard to square with AI demand breaking (P001); a day later Asia memory, AI bottlenecks and Japan optical were higher again, and GS argues that after the derating it is difficult to see huge fundamental downside without another leg (P014). Positioning evidence is mixed: COT showed elevated non-dealer Nasdaq length into the move and GS's AI vs AI-at-risk basket plunged -10.9%, but the largest cohort endured similar price action before, leaving bulls at risk yet not demonstrably forced (P003). For investors the consequence is barbell risk within tech: momentum and AI-winner baskets have swung violently (P001, P020), so headline index exposure masks violent intra-tech dispersion. Conditional interpretation: a durable capex slowdown at hyperscalers would be the fundamental confirmation that turns this from rotation to downcycle; the cited evidence so far shows intent-to-slow commentary, not delivered capex cuts (P001).

Tension

P003 sees elevated non-dealer length and lingering Nasdaq bull risk; P001/P014 read the same tape as a contained right-tail repricing with Asia holding firm.

What changed

AI bellwethers' joint call for restraint on 14 Sep triggered a -10.9% drop in the AI vs AI-at-risk basket, reframing the debate from demand to capex discipline (P003, P001).

Desk views

GS Privorotsky: de-grossing event repricing the highest incremental infrastructure growth assumptions; hardest hit where earnings/multiples relied on continued acceleration; Asia strength suggests AI demand itself is intact.

Evidence IDs · P001

GS Quinn: violent shift tempts one to call Nasdaq bulls at risk, but non-dealer length had already been trimmed (-$7.1bn Sept 1-8 after +$51.3bn buying) and broader positioning is mixed.

Evidence IDs · P003

GS Privorotsky next day: Asia trading better (memory +3%, AI bottlenecks +2.5%); after the derating, hard to see huge fundamental downside without another leg — the issues are politics, rates and macro.

Evidence IDs · P014

Investor implication

For equity investors, discriminate within AI rather than exiting the theme: right-tail infrastructure growth names carry multiple-compression risk while software, bottlenecks and demand-side beneficiaries behave differently (P001). Next observable test: hyperscaler commentary and any delivered capex revisions, plus whether Asia hardware strength persists (P001, P014).

Trade expression

Barbell within tech rather than outright exit; favor demand-side/software over right-tail infrastructure growth names until capex guidance resolves (P001).

Invalidation

A delivered, quantified hyperscaler capex cut or renewed collapse in Asia AI hardware would shift this from right-tail repricing to a broader AI downcycle.

Verdict

The supply disruption is real and duration is highly uncertain; GS sees Brent settling ~$85 by end-2026 even with tensions possibly extending into 2027, while the near-term tape prices scarcity and stagflation risk (P013, P015, P018).

Analysis

The mechanism is a supply shock: Saudi Arabia closed its 7mb/d East-West pipeline after Thursday's attacks and an Iran-Gulf meeting on Hormuz shipping was postponed, pushing Brent to nearly $110 (P002, P018). GS estimates an August deficit of ~3mb/d vs a ~1mb/d balance estimate, and repair assessments for Yanbu exports range from 'very soon' to eight weeks — with affected volumes and outage duration highly uncertain (P018). Market consequence: higher energy prices fed stagflation fears, pushing Brent to $108.75 close (+2.90%) and the 10-year to briefly 5.04%, with the S&P 500 falling to a 6-week low (P013). Counterweights: GS commodity strategists see Hormuz flows at ~40% and Gulf flows at ~2/3 of pre-war levels and forecast Brent settling ~$85 by end-2026 even if tensions extend into 2027 (P015); Nomura, by contrast, argues this round is more 'terminal' because China is rebuilding rather than releasing reserves, and flags energy equities, commodity carry and long vol as the hiding places (P030). For investors, the key uncertainty is outage duration, not direction of the shock itself; realized prices far above GS's year-end forecast imply substantial risk premium that could unwind on repair news.

Tension

GS fundamentals point to Brent settling ~$85 by end-2026 (P015, P018); Nomura sees the shortage as increasingly terminal with China rebuilding inventories (P030); the front tape trades ~$108-110 (P013).

What changed

Pipeline closure escalation: the April attack cut flows only 0.7mb/d for four days, but the recent attack may be more severe and could threaten the remaining 2mb/d of Yanbu exports (P018).

Desk views

GS Commodities (Grigsby et al.): Brent near $110 on the Saudi East-West pipeline closure and postponed Hormuz meeting; deficit ~3mb/d in August vs ~1mb/d balance estimate; outage duration highly uncertain.

Evidence IDs · P018

GS China (Musings): Hormuz flows ~40% and Gulf flows ~2/3 of pre-war levels support the commodity team's view that Brent should settle ~$85/bbl by end-2026 even with tensions possibly extending into 2027.

Evidence IDs · P015

Nomura McElligott: this round differs because China is rebuilding crude inventories, likely in coordination with Iran; energy equities, commodity carry, rates trend and long vol are the favored uncorrelated hiding places.

Evidence IDs · P030

Investor implication

Treat the oil premium as a duration-uncertain shock: repair timelines (days to eight weeks) are the swing factor for both crude and the rates complex (P018). A resolution toward GS's $85 year-end path (P015) would relieve the stagflation pressure currently driving both bonds and equities lower (P013); a longer outage validates the Nomura regime of energy-equity and carry hedges (P030).

Trade expression

Energy equities, commodity carry and long vol as hedges per Nomura's regime framing (P030); watch for sharp downside in the risk premium on repair news (P018).

Invalidation

Rapid pipeline repair and restored Hormuz shipping toward GS's ~$85 end-2026 path (P015, P018).

Verdict

The hike itself is a non-event at ~93-94% priced (P014, P013); forward risk sits in whether the median dot shows one or two hikes in 2026 and whether Warsh frames tightening as inflation-driven or oil-driven — the answer governs rates vol, USD and equity multiples (P014, P017, P019, P021).

Analysis

Markets price 23bp for today, 52bp by year-end and 89bp by June; ING argues openness to further hikes keeps USD supported while a surprise hold would hit the dollar via both front-end repricing and a back-end selloff — a risk the Fed is aware of, keeping hold probability low (P021). GS Economics added the 25bp hike after August CPI and raised its terminal rate to 3.25-3.5%, but sees no strong economic case for hiking, attributing the overshoot to one-time factors including tariffs (P017). Privorotsky's key question: does the median dot show one or two 2026 hikes — GS's GIR expects a 10-8 majority for one — and does Warsh frame this as a response to persistent underlying inflation or an attempt to stop the energy pass-through (P014)? The vol math favors a hawkish-but-expected outcome: GS finds mildly hawkish Fed surprises deliver better vol-selling returns than holds when a hike was at risk, and a one-to-two hike median with data-dependence should let vol moderate by removing the tail of a policy error (P019). Equity read-across splits: Citi finds global equities wobble at the start of hiking cycles but climb 6/12m later (P012); UBS expects the hike but calls it a shift, not a sea change, retaining S&P targets of 8,100 year-end and 8,400 mid-2027 on earnings growth (P033). Next test is immediate: the SEP dots and press conference framing.

Tension

GS Economics sees no strong economic case for hiking and attributes the overshoot to one-time factors (P017); the market and most desks price a hawkish recalibration with ~3.5 hikes through next summer in futures (P014, P021).

What changed

August CPI pushed market pricing of a hike to ~90%, forcing the FOMC's hand and prompting GS to add the September hike and raise terminal rate to 3.25-3.5% from 3-3.25% (P017).

Desk views

GS Economics (Mericle): added the 25bp September hike after CPI; raised terminal to 3.25-3.5%; but sees no strong economic case, with the overshoot from one-time factors and core PCE improving to ~2.5% over three months.

Evidence IDs · P017

GS Rates Vol: mildly hawkish Fed surprises have historically delivered better vol-selling returns than holds when a hike was at risk; a one-to-two hike dot median with data dependence should moderate vol.

Evidence IDs · P019

ING: consensus 25bp hike to 4.0% can still lift USD; a dovish hike may not convey the discipline bond investors demand given the tightening already priced into swaps.

Evidence IDs · P021

UBS: expects the hike but sees earnings growth on a solid footing as the main driver of further equity gains; retains S&P 500 targets of 8,100 year-end and 8,400 mid-2027.

Evidence IDs · P033

Investor implication

Position for the reaction function, not the hike: a one-hike median with data-dependent framing should allow vol to moderate (P019) and is the constructive case for equities per UBS and Citi (P033, P012); a two-hike median or inflation-first framing risks validating the ~3.5 hikes futures price through next summer and re-tightening financial conditions (P014). USD direction is the cleanest expression — a hold is the low-probability, high-impact outcome (P021).

Trade expression

Vol-selling into a mildly hawkish, expected hike historically paid better than into holds, per GS Rates Vol (P019); USD longs into the meeting per ING (P021).

Invalidation

A two-hike median dot or explicit inflation-first framing that validates the full futures path of ~3.5 hikes through next summer (P014).

Verdict

Higher yields are a live risk but not yet a throttle: Berenberg sees a significant equity drawdown as unlikely until rates or rate volatility begin throttling capex, and treats pullbacks as add opportunities (P006); JPM agrees the damage so far is multiple compression, not earnings (P023); the flow floor under US credit remains foreign demand (P016).

Analysis

The mechanism: expectation pivot from cuts to hikes has driven long-dated Treasuries down ~10% and compressed forward PE ~3x, with 30-year yields at ~5.40% — driven by policy repricing, sticky inflation (oil and trade shocks), rising fiscal deficits, and possible crowding-out as foreign investors favor long-term corporate debt over Treasuries (P023). Berenberg's threshold framework: a significant equity drawdown is unlikely until higher rates or heightened rate volatility begin throttling capital expenditure across the financial and real economy — and there is little sign of that today, with solid macro momentum and healthy credit demand (P006). This explains why equities trade near all-time highs despite the bond rout. Citi adds the conditional: equities can better digest higher yields when growth stays resilient and inflation falls — suggesting short-term caution amid stagflationary risk but no abandonment of the earnings-driven upside call to mid-27 (P012). Credit is the transmission watch: foreign investors own ~29% of US corporate bonds, and 2026 YTD net purchases of $251bn put the year on pace to rival 2025's $392bn record — a flow floor that supports the 'capex not throttled' thesis but which JPM flags as a possible Treasuries-crowding-out channel (P016, P023).

Tension

Berenberg treats the bond rout as an add-equity opportunity while capex holds (P006); Citi counsels short-term caution amid stagflationary risks from geopolitics (P012); JPM documents both the resilience (earnings +16% equity return) and the pressure (PE -3x, USTs -10%) (P023).

What changed

Long yields broke higher on oil: the 10-year broke above its 2023 intraday peak to briefly touch a post-2007 high of 5.04% (P013); US 30yr has risen to ~5.40% (P023).

Desk views

Berenberg: a significant equity drawdown is unlikely until higher rates or rate volatility throttle capex in the financial and real economy; little sign of that today; pullbacks are add opportunities.

Evidence IDs · P006

JPM Global Equity Strategy: pivot to hikes cut long-dated Treasuries -10% and forward PE ~3x, but equities are still +16% on 28%/21% earnings growth for 2026/27.

Evidence IDs · P023

GS (Lynam): foreign appetite for US credit is holding — $251bn YTD net purchases through June, on pace to rival 2025's $392bn record; Europeans are 52% of net purchases.

Evidence IDs · P016

Citi: equities wobble around the start of hikes but climb 6/12m later; they digest higher yields when growth stays resilient — short-term caution, but the earnings-driven upside call to mid-27 stands.

Evidence IDs · P012

Investor implication

Watch capex and credit demand, not yield levels in isolation: as long as credit demand holds and foreign flows into US credit persist (P006, P016), higher yields compress multiples rather than earnings. The drawdown trigger would be rate-driven capex throttling — visible first in credit spreads and investment plans (P006). Conditional interpretation: the JPM-flagged shift of foreign buyers toward long-term corporate debt over Treasuries is itself a yield-pressure channel worth monitoring (P023).

Trade expression

Keep benchmark equity weighting against underweight duration per Berenberg's stance, using pullbacks to add while the capex throttle is absent (P006).

Invalidation

Evidence of capex throttling by financing costs, or a break in foreign demand for US credit (P006, P016).

Verdict

China's growth is relatively unscathed by the oil shock thanks to energy diversification, but domestic demand is deteriorating — leaving policy easing and the September 24 Trump-Xi summit as the swing factors, not the oil tape (P015, P007).

Analysis

Two CACIB/GS threads combine. On the shock: Chinese equities exhibit lower return beta to oil than global peers, supported by strategic energy diversification, sizable reserves, fiscal offset and robust refining capacity — a relative-value argument within EM equities while the rest of the world absorbs stagflation pressure (P015). On the domestic economy: August data showed further weakening of demand — retail sales growth eased again, FAI further declined — while high-tech production supported IP growth (P007). CACIB expects policymakers to push faster implementation of planned easing via fiscal and quasi-fiscal spending, with growth possibly stabilizing in coming months 'but without major rebound'; crucially, if the summit confirms a manageable US-China relationship, policymakers may feel no urgency to ease aggressively (P007). The causal mechanism for investors: the oil shock is externally disincentivizing for Chinese equities only via global risk appetite, not via direct terms-of-trade — so China is a hedge against the oil regime, while the real repricing risk is political (tariffs, Busan Truce extension, AI guardrails at the summit, P007). Next observable test: the summit itself and whether major demand stimulus appears.

Tension

Production strength vs demand weakness: high-tech output holds IP up while retail sales and FAI fade (P007) — a two-speed economy complicating both the easing outlook and equity earnings paths.

What changed

Oil back above $100 after months around $80 — yet Chinese equities' low oil beta means the terms-of-trade hit is far smaller than for global peers (P015).

Desk views

GS China Musings: China remained relatively unscathed in the oil shock due to energy diversification, reserves, fiscal offset and refining capacity; Chinese equities show lower return beta to oil than global peers.

Evidence IDs · P015

CACIB: August demand indicators further worsened despite production strength; expect faster policy easing implementation, but if the Trump-Xi summit confirms a manageable relationship, policymakers may see no urgency to aggressively ease.

Evidence IDs · P007

Investor implication

Chinese equities function as a partial hedge within EM against the oil regime (P015), but the sharper repricing risk is event-driven: the September 24 Trump-Xi meeting on tariffs, the Busan Truce and AI guardrails (P007, P015). A confirmed state visit with manageable outcomes removes the case for aggressive easing — capping the stimulus upside — while a breakdown would hit both the relationship and the oil-stressed global tape.

Invalidation

A summit breakdown on tariffs or AI guardrails, or confirmation that major demand stimulus will not materialize despite weakening data (P007).