Overnight & Global
The de-escalation dividend is not a straight line — on 6 Aug it partly reversed. (1) The deal stalled, and the war premium re-bid. The Bessent/Trump "Wednesday or Thursday" signature did NOT land — Rubio called it "progress but not finality," and what is taking shape is a narrow Iran-Oman navigation channel (coordinates agreed, a joint statement "in the final stage"), not the broad US-brokered package Washington wanted, with the Strait still physically shut. Oil, which had crashed ~5% two sessions running pricing the reopening in, re-bid +2.3% to $79.80, and gold rose +2.3% — the haven/war-premium tell. (2) A growth scare compounded it. ADP printed +44k (vs ~70k expected, the weakest of the year); the 2-year eased to 4.18% and the 10-year to 4.61%, and the gold bid firmed. The market began de-risking into Friday's payrolls (consensus ~83k, the whisper now skewed lower). (3) The AI-tail snapped — the market repriced the COST of the build. AMD gave back -7.0%: a clean beat (data-centre +107%) sold on a gross-margin miss (54% vs ~56%) from the Helios ramp — growth is real but expensive. Alphabet -4.0% on a capex hike to $195-205bn. The speculative long-duration names took the worst: the SpaceX tracker -13.6% (capex + a lockup expiry), the quantum complex -14% to -26% (D-Wave/Rigetti report today, pre-profit). But NVDA HELD +3.4% — Washington approved ~10 Chinese firms to buy the H200, so its AI spend is someone else's cost and its revenue. The tape bifurcated cleanly: sell the names where AI spend is a cost with a distant/uncertain return; hold the demand-confirmed picks-and-shovels. Own the AI monetisers + the cyclical/value/energy barbell; the tail is the AI-cost names + a deal fracture.
Session Read
THE STREET vs DELPHIC — external research vs our re-run radar (houses anonymised, as we always do). We re-scrutinised the fresh external research against our re-run radar; the full adjudication is in the External Read digest, anonymised. Where we AGREE: AI-capex is the master variable and the market IS the AI trade; the deleveraging is largely done (our Risk axis eased). The re-adjudication on 6 Aug — the market repriced the COST of the build, not the demand. The bears' thesis (hyperscaler FCF deteriorating as capex outruns cash, AI-name CDS the tell) got its evidence in the tape: AMD sold on a margin miss, Alphabet on a capex hike, the speculative names crushed — but the bulls' point held too, because NVDA rose on a concrete demand unlock (the H200 China approval). Our synthesis: this is differentiation, not a bubble bursting — own the demand-confirmed monetisers/picks-and-shovels, fade the AI-cost/ROI-distant names, and carry AI-name CDS as the leading tell. A hawkish outlier: one Fed official argued it is "time to raise rates" — in tension with the weak ADP; the curve sided with the growth scare (yields down). Where Delphic had the EDGE — oil/Hormuz: we HELD Phase 2 through the near-deal optimism and the stall alike, never over-calling a reopening the flow never delivered.
The deal stalled — the war premium re-bid, and gold went with it. The Bessent/Trump "Wednesday or Thursday" signature did NOT land. Rubio: "progress but not finality." What is materialising is a narrow Iran-Oman navigation channel (coordinates agreed, a joint statement "in the final stage") — not the broad US-brokered ceasefire/nuclear package; there are no direct US-Iran talks, and the Strait is physically still shut (~2 transits/day). Oil, which had crashed ~5% two sessions running pricing the reopening in, re-bid +2.3% to $79.80, and gold rose +2.3% — the haven/war-premium tell. Our SoH Monitor HOLDS Phase 2: the signature is the swing — it firms (oil down, risk-on) or keeps slipping / fractures (oil up). A slipped signature is a warning, not a trigger.
The AI-tail snapped — the market repriced the COST of the build; NVDA the exception. The reversal was idiosyncratic to long-duration / high-cost AI. AMD -7.0% gave back its beat on a gross-margin miss (54% vs ~56%) from the Helios AI-infrastructure ramp — "growth is real, but it's expensive." Alphabet -4.0% lifted 2026 capex to $195-205bn with more guided for 2027 — the fear that AI-compute cost outruns near-term returns. The speculative tail took the worst: the SpaceX tracker -13.6% (Q2 AI capex ballooned to $15.8bn + a post-IPO lockup expiry), the quantum complex -14% to -26% (D-Wave/Rigetti report today, pre-profit, negligible revenue). But NVDA held +3.4%: Washington cleared ~10 Chinese firms to buy the H200, so its AI spend is someone else's cost and its revenue. The signal for the monitor: the market pays the demand-confirmed picks-and-shovels and fades the names whose spend is a cost with a distant/uncertain return — differentiation, not a top.
The macro — a growth scare (ADP +44k) eased the curve into Friday's payrolls. ADP printed +44k (vs ~70k expected, the weakest of the year; leisure/hospitality -11k, trade/transport -8k) — a genuine growth wobble. The 2-year eased to 4.18% and the 10-year to 4.61%, and gold caught a bid. The regime read stays a reflationary late-cycle, but the growth axis just flashed a caution and the inflation tail is still deflating (oil off the pre-deal level even after the bounce). One Fed official argued for a HIKE (a hawkish outlier), but the curve sided with the growth scare. The next read is Friday's July payrolls (consensus ~83k, unemployment ~4.2%); with ADP at 44k the whisper is skewed below consensus — a soft print + lower oil would fade the hawkish-hike case the Street is positioned for.
Greece — the banks CONSOLIDATE (5-Aug close); Metlen H1 today, PPC reaffirmed. On the latest Athens close (5 Aug) the banks took profit after the record run — National -1.8%, Eurobank -1.4%, Piraeus -0.5%, the banking index ~-1.2% — a healthy digestion at a 17-year-high index, not a reversal; leadership rotated to GEK Terna +2.6%, Aegean +2.2%. The corporate season delivered: PPC reported adj. EBITDA €1.2bn, reaffirmed FY guidance (€2.4bn EBITDA / €0.7bn net) and confirmed a DPS of €0.80; Coca-Cola HBC beat and raised; Cenergy a record H1 + a €1.15bn ADMIE interconnection award. Metlen reports H1 today (6 Aug) — the read on whether the energy/metals complex confirms the supercycle (gallium the swing). Politics: the Western Attica wildfire relief; polling frozen in the August blackout.
The macro read. Read the 6-Aug tape as a reflationary late-cycle taking a growth-scare wobble, with the AI-cost trade de-rating: the Iran deal stalled (oil + gold re-bid), ADP missed (the curve eased), and the market repriced the cost of the AI build — punishing the margin/capex/ROI-distant names while holding the demand-confirmed one (NVDA). Own the demand-confirmed AI monetisers + the cyclical/value/energy barbell, keep the skeptic's AI-CDS tell, and watch Friday's payrolls + the Iran signature. On Hormuz, HOLD Phase 2 — the deal is unsigned and the Strait physically still shut.
Theme of the Day
The de-escalation dividend is not a straight line — on 6 Aug it partly reversed. (1) The deal stalled. The Bessent/Trump "Wednesday or Thursday" signature did NOT land — a narrow Iran-Oman navigation channel is taking shape, not the broad US package, with the Strait still physically shut. Oil re-bid +2.3% to $79.80 and gold +2.3% — the war-premium/haven tell. (2) A growth scare compounded it. ADP +44k (vs ~70k) eased the 2-year to 4.18% and the 10-year to 4.61% and firmed the gold bid, with the market de-risking into Friday's payrolls. (3) The AI-tail snapped on COST. AMD -7.0% (a beat sold on a gross-margin miss from the Helios ramp), Alphabet -4.0% (capex to $195-205bn), the SpaceX tracker -13.6% and the quantum complex -14/-26% — but NVDA held +3.4% on the H200 China approval, its AI spend being someone else's cost and its revenue. The tape bifurcated: sell the AI-cost / ROI-distant names, hold the demand-confirmed picks-and-shovels — differentiation, not a top. Own the demand-confirmed AI monetisers + the cyclical / value / energy barbell, keep the AI-CDS tell, and watch Friday's payrolls + the Iran signature.
The Call — the daily cross-check on the Regime Radar
The daily cross-checks the Regime Radar, and the 6-Aug read is a reflationary late-cycle taking a growth-scare wobble: resilient growth but a soft ADP (+44k) that eased the curve (2Y 4.18 / 10Y 4.61), an inflation tail still deflating, and an AI complex that de-rated on COST — the market punished the margin/capex/ROI-distant names (AMD, Alphabet, the speculative tail) while holding the demand-confirmed one (NVDA +3.4% on the H200). A within-complex re-sort, NOT a broad top. The rotation favours the demand-confirmed AI monetisers + cyclicals / value / energy. Own that barbell, watch Friday's payrolls and the Iran signature — the deal that drives oil is unsigned and the Strait is physically still shut (HOLD Phase 2).
