Macro & Geopolitical Intelligence

Read: the 10Y fell 14bp over the week (5.24% → 5.10%) while the Nasdaq lost 2.15% on Friday and PPI printed 6.3%. Treasuries rallying through hot inflation data is not a dovish repricing — Reuters attributes it to investors fleeing AI-related credit into duration. That is a credit bid wearing a duration mask, and it is the same stagflation-lite tell that appeared on 15 Oct when a hot CPI raised October hike odds while the long end rallied. The Warsh put remains OFF: with ~89.5% of a December move priced and GDPNow at 3.6%, there is no growth excuse for cuts — every oil-fed CPI and PCE print is a live tightening event. With the blackout in force, the next Fed information is the 28 Oct statement itself.

Read — a rare AU–US divergence. RBA November hike odds have been marked down ~9pp to 16–17% and the cumulative-to-February strip from 59.4% to 49.1%, on two independent sources (ASX 95.365 and centralbank.watch 17.1%). That is the market deciding Q3 CPI on 29 Oct will not force the RBA's hand, even with Brent at $110. The ASX close is the trap: 8,806.70 (+1.00%) was struck hours before SOX broke 3.83%. Monday's 10:00 AEST open should gap down, and the banks/miners rotation buffer that saved the index on prior KOSPI circuit-breaker days is the only thing that would stop it. Watch the futures-gap versus cash-close divergence, not the open print.

Read — this is a valuation risk right now, not a Strait risk. TSMC delivered a record quarter (+50.9% YoY revenue, +69% net income, 66.4% gross margin) and raised Q4 guidance to +58–61%, and the ADR still fell 2.28% — its third consecutive "beat-but-sold" session. When record results and raised guidance cannot lift a stock, the bind is duration and credit, not geopolitics. Trigger indicators for the next 90 days: (1) any PLA exercise announcement naming the 2026 post-Xi truce window, (2) TSMC's January capex guide as the AI-demand tell, (3) Taiwan's special ammunition-stockpile bill (30 → 120 days) as a Washington-alignment marker. Risk level: MODERATE — military posture is the quietest it has been all war, while the equity risk premium does the damage.

Read — do not read the spread as healing. Urals at $114.30, a $4.44 premium to Brent, is the giveaway: a normally-discounted Russian grade cannot trade above Brent unless sanctions and war-risk routing have broken the arbitrage. When the constraint is structural rather than grade-specific, the front spread compresses precisely because nothing is arbitrageable. $100+ Brent is now a two-session fact, and it feeds headline CPI/PCE with a ~2-month lag — which is why 9 Dec, not 28 Oct, is the live Fed meeting.

New supply-chain channel: data centres are now military targets. Ukraine struck a VK data centre in St Petersburg overnight, following strikes on multiple Yandex facilities, after Russia hit Ukrainian data centres — Zelensky's "respond in kind". This is the first war in which commercial AI/compute infrastructure appears on both sides' target lists. It does not move a share price today; it is a structural input to where hyperscalers will agree to build.

Read — "threat up, rhetoric down" remains the most oil-bullish combination available, and this weekend added a de-escalatory line without removing a single munition. Trump told a Kentucky rally the war will end "very quickly" and that fuel will fall below pre-war levels; Araghchi floated a "political process"; CENTCOM insists the strait is open. Simultaneously Riyadh is under aerial-threat alerts, an RPG reached a tanker near Jubail, and the MOAB remains explicitly on the table after 3 Nov. That is pressure being maximised for a pre-midterm exit he can declare, not for de-escalation. Do not fade crude on "ending soon" quotes.

The one genuinely new risk this weekend is geographic. Every prior attack was a transit-lane event. An RPG 78nm from Jubail is a loading-terminal event, and it is the first credible mechanism by which the 68% that still flows could actually stop. Trigger indicators, next 30 days: (1) any strike on Ras Tanura or Ju'aymah loadings, (2) Houthi anti-ship ballistic missiles emplaced on the Perim/Yemen coast, (3) minesweeping confirmation from a non-belligerent navy. Day counts vary by tracker: skill formula 233, straits.live 231 since closure, Gulf News 228.

Read — gold's refusal to confirm is the signal. On a weekend with aerial-threat alerts over Riyadh, a tanker hit near Jubail and Brent above $110, spot gold is flat to slightly lower versus Friday. That is not a systemic-risk bid — it is the market pricing a supply-specific oil shock, not a credit or geopolitical event. It matches Friday's pattern, where gold's bid was attributed to credit quality rather than war risk. Two notes on the tape: the KOSPI's −11.2% week (6,384 → 5,668) is the largest weekly destruction since the July crash and closed below the July low — a bear regime, not a dip. And Friday's Asia closes pre-date the US rout by ~14 hours, so every Asia level above is a pre-rout print, labelled as such.

Reconciliation gate: PASS. Total A$120,373 against Friday's A$120,358 (+0.01%), with every position's P&L reproducing to 0.1pp — MU +32.1%, MSFT +33.0%, META +20.6%, TSM +6.2%, RDDT −0.2%, AVGO −20.0%, QQQU −45.5%, CBRS −56.1%, TSXU −91.5%. FX direction verified: AUD-per-USD 1.4398 (0.6945), AUD-per-HKD 0.18346. No cost-basis or direction error. There has been no session since Friday, so the book is unchanged by design.

What actually matters this week is concentration, not level. Six of nine US positions are pure AI-supply-side (MU, AVGO, TSM, CBRS, TSXU, QQQU) at A$42.2k, plus META/MSFT as AI spenders at A$41.2k. Friday demonstrated that the suppliers and the spenders now fall together because the binding constraint is financing cost, not end-demand — so there is no internal hedge. The only offsets in the book are cash and gold, and gold is currently refusing to bid.

Three live positions of risk into Wednesday. (1) META reports Wed 21 Oct after the close — A$18.4k, +20.6%, and the largest single-position P&L contributor after MSFT; a capex-guided miss hits both the spender and the supplier sleeve at once. (2) TSXU at −91.5% is now a A$730 residual — it has no recovery maths left, only a tax-loss and optionality value; treat it as a lottery ticket, not a position. (3) MU at $878.84 has given back the $900 line; at +32.1% it is still the book's best performer, and the CXMT-driven dip will look like a buying opportunity. It is not a cycle yet — a 4F² roadmap is not qualified server silicon. Wait for CXMT DDR5 RDIMM qualification with an OEM.