Week ahead · Macro · post-FOMC · the capex verdict · August 3, 2026
The Fed held. The line didn't.
Grade sheet first, as always. Last week's four wires: the Alphabet reception wire fired — a raised capex guide ($195–205B) was sold for the worst next-day print of the megacap week, confirming the discount-the-spend regime at the top of the chain (Microsoft then complicated the regime in the best possible way; Section 4). The tariff wire resolved — the 10% surcharge expired at 12:01 a.m. Friday and a permanent two-tier Section 301 regime replaced it the same minute; duration stayed below-benchmark, which the 30-year's 5.2% close made look conservative rather than clever. The oil wire fired — WTI settled $92.19 on July 23, through the ~$90 line, with Brent through $100 — and then crude round-tripped the whole move in its worst three-day stretch since 2020. And the Micron wire never got the chance to fire against a live position: the runner was already gone.1
The one everyone will ask about is the exit, because of what this ledger went through to get here. Two weeks ago the $900 rule fired and we froze — an F on process, graded in public, repaired with a promise we made structural: a stop-loss order resting at the broker at $900, no human in the loop. In Monday and Tuesday's breakdown — China's CXMT had just debuted at a $488 billion valuation and the global chip complex was shedding a trillion dollars in three sessions — Micron broke $900 and the order sold all 15 shares at approximately $900. When Tuesday's $820.53 close fired our published $848.95 line, the book was already flat. Wednesday's $739.00 close made the machine look brilliant; Thursday's 18% rip made it look early; the ledger doesn't care either way. Twelve weeks, ~+53% blended, all realized — the full accounting, grade sheet and all, is in the closing note published last night, and Section 2 carries the summary.2
Around those two stories, the macro tape had its most consequential week since the June FOMC. Chair Kevin Warsh held with three knives out — Hammack, Kashkari, and Logan all voting to hike — and told the room there is “no soft inflation target… not on this Committee's watch.” The long end called the bluff: a violent bear-steepener took the 30-year through 5.2% at the close, its first such close in nineteen years, while the short end trimmed September hike bets. One desk called the whole move “consistent with a central bank inflation credibility shock.”3 PCE landed Thursday morning — headline down to 3.7%, core to 3.3% — cooling, and nowhere near cool enough to settle the argument.
This week the argument moves to the labor market. Jobs land Friday August 7, the July FOMC minutes — with three dissent rationales inside — on the 19th, and the war-oil CPI on August 12. Sections 6 and 7 set the lines. Everything below is scenario work with named trip-wires, not point forecasts.
1. The scoreboard: four wires, four answers
Set ex ante in the July 20 outlook, graded now — the same way we'd grade anyone else.
Wire #1: Alphabet's print meets the new bar — fired, and then refined
The wire said a capex raise that gets sold confirms the discount-the-spend regime. Alphabet raised 2026 capex guidance to $195–205B on a genuinely strong quarter — revenue +24%, Cloud +82% — and printed its first negative-free-cash-flow quarter since its 2004 IPO, with buybacks at zero for a second straight quarter. The stock was sold 7.1% the next day and dragged the tape with it. Wire confirmed — and then Microsoft rewrote the fine print a week later, which is Section 4's story. Response held: no cash deployed into the weakness before the rest of the gate cleared.4
Wire #2: The tariff handoff — resolved, as a wash with teeth
Section 122's 10% global surcharge expired at 12:01 a.m. Friday July 24, dead on its 150-day statutory cap — and the Section 301 forced-labor action took effect the same minute. Two tiers across 60 economies covering 99.4% of U.S. imports: 10% on seventeen partners, 12.5% on the rest — including China — with USMCA-compliant goods duty-free and about 2,100 tariff codes carved out. The scheduled disinflation event we flagged became a permanent regime swap with no gap: a wash on the 10% tier, a modest tightening on the 12.5% tier, and no sunset clause this time. Response held: duration stayed below-benchmark — and with the 30-year closing the week at 5.2%, the position needs no defending.7
Wire #3: Oil escalation — fired, then round-tripped
The wire: WTI closes above ~$90, or a strike touches Gulf production infrastructure beyond Kuwait's utilities. The first clause fired Thursday July 23: WTI settled $92.19 (+6.2%, fifth straight up day) and Brent settled $100.69, its first close above $100 since May — on Houthi attacks against Saudi tankers with Hormuz transits still in the single digits. The second clause did not: production and export infrastructure went unhit as far as any confirmable source shows (Houthi claims against the east–west Saudi corridor remain claims). Then diplomacy flickered, and crude gave the entire move back — the worst three-day decline since 2020 — before Wednesday's escalation snapped it right back: WTI +6.6% to $84.46 and Brent back through $90 on an Iranian missile attack against US forces and the retaliatory wave that followed. Net on the week of wires: WTI $82.49 → $84.46. A war premium that violent, that round-trips to nearly flat, is not a trade — it's a warning about being in the blast radius with size. Response held, boring on purpose: heavy cash, no war-trade heroics.6
Wire #4: The Micron runner — overtaken by its own repair
The wire had no discretion left in it by design: any close below $848.95 and the runner is sold at the next open. It fired Tuesday — $820.53, with an intraday high of $848.36 — against a book that was already empty, because the broker stop at $900, the literal implementation of the July 21 repair, had executed on the way down. The published line was the backstop; the machine in front of it went first, and higher. Section 2 logs it; the closing note grades it.2
2. The exit, executed: twelve weeks, closed ledger
The mechanics first, because after the violation note the mechanics are the whole point. The repair we published on July 21 was “a structure that doesn't need us to” — and the structure was literal: a stop-loss order resting at the brokerage at $900, the original June line, the level our published discipline had already failed once. When the CXMT shock took Micron back through $900 in the July 27–28 breakdown — Monday's close grazed $900.20, Tuesday broke it for good — the order did the only thing it knows how to do: all 15 shares, out at approximately $900, no decision, no debate, no human in the loop. The published $848.95 close-line — the one we staked this site's risk credibility on — fired at Tuesday's $820.53 close against a position that no longer existed. What we failed to do manually on July 16 at $853.20, the machine did without us at a better price.2
The ledger, closed after twelve weeks and fully realized. Coverage began May 6 at a $668 basis with a $1,100 target the Street called aggressive. The stock ran to a $1,213.56 record; we booked roughly 60% through the pre-planned scale-out at a ~$1,100 average — +65%, banked in June, the single best decision of the whole call. The ~40% runner — held one week too long, in graded violation of the first stop — exited at ~$900 for +35%. Blended: ~+53% — matching, to the digit, the “worst case” floor we published on July 8. Never selling a share sits at roughly +31%. The rule-honored counterfactual — the $848.95 close-line executing at Wednesday's open — lands near +50%: the intraday stop traded precision for certainty and, this time, beat its own backstop. The system, run through one confessed failure and one mechanical repair, did what it said it would. Full grade sheet — four rights, two wrongs, one repaired F — in the closing note.2
The postscript is the part worth framing. Wednesday, with the book flat, Micron closed $739.00 — the exit looked like genius. Thursday, Samsung printed a record quarter and warned the memory shortage worsens into 2027 and continues into 2028, Microsoft's print lit the whole complex, and Micron ripped 18% to $874.66 — the exit looked early. Six hours of genius, six hours of early. We wrote it during the BofA bounce and it holds now that we're out: rules get graded on the distribution of outcomes they produce, not on any single print — and for the first time in this saga, a violent bounce taunts nobody. It is somebody else's whipsaw now. Re-entry is a separate underwriting with its own published triggers — Section 7 — and it does not happen on green candles.2
Chart 1 — The line that didn't need us
Eight sessions: the walk down from $990, the stop executing at ~$900 into the breakdown, the published line firing against an empty book — and the 17% rip that taunts no one.
MU daily closes, July 21–30, 2026. The pre-placed broker stop at $900 — the repair promised in the July 21 stop note — sold all 15 shares at ~$900 into the July 27–28 breakdown; the published $848.95 close-line fired at Tuesday's $820.53 close against an already-flat book. July 30: +18% to $874.66, after the exit. Full accounting: the closing note. Sources: verified exchange closes; prior published notes.
3. The Warsh Fed: a hold with three knives out
The decision itself was the least interesting sentence of the afternoon: target range unchanged at 3½–3¾%, fifth straight meeting. The vote was the story: 9–3, with Hammack, Kashkari, and Logan all dissenting for a quarter-point hike — per AP, the first time three policymakers dissented in the same direction since September 2016. The statement barely moved a word from June. The chair moved plenty: “There is no soft inflation target, there is no soft implicit target — not on this Committee's watch. There is only a target, and it is 2 percent.” He called the stance “a period of watchful thinking, not watchful waiting,” called the dissents “a good family fight” he had asked for, and — asked about markets pricing a September hike — declined the anchor: “we're not going to be constrained by market prices.” On whether rates go up if inflation stays elevated: “interest rates could well be part of that solution.”3
The bond market's response was the most instructive price action of the week — a violent bear-steepener that read the hold as the mistake. The 2-year fell 4 basis points to 4.24% as September bets got trimmed; the 10-year rose to near 4.69%; and the 30-year jumped 10.5 basis points to close at 5.201% — its first close above 5.2% since 2007, with an intraday high of 5.244%. Translation: the short end priced “no hike delivered,” the long end priced a Fed that's “falling behind.” BofA's desk called the move “consistent with a central bank inflation credibility shock.” Equities agreed: the Dow fell 1,153 points (−2.19%), its worst day since April 2025, the S&P dropped 1.5% to 7,316, and the Nasdaq closed more than 10% below its high — correction territory. September hike odds, curiously, fell on the day — from roughly three-quarters priced to a coin-flip-plus — because a Fed that didn't hike with three dissents pushing may not hike at all this year. Warsh's own framing did the work the funds rate didn't: yields have risen across the curve since June by one of the largest intermeeting moves in two decades, and he blessed it — “market participants are learning to play the ball, not the referee.”3
Thursday's PCE gave both camps a card. Headline eased to 3.7% from 4.1% — the June energy give-back, exactly the rear-view-mirror dynamic we flagged on CPI two weeks ago — and core ticked down to 3.3%. Cooling, genuinely. Also: still far above target on both gauges, with July's war-oil basket unprinted. Our June lesson stands unrevised — don't fade an incentive-aligned risk. A new chair rebuilding credibility has every incentive to err hawkish, and the whole curve now agrees with the read we adopted, painfully, on June 19.3
4. The capex verdict: proof gets paid, faith gets sold
Three of the five gate prints are in, and the market has drawn a line we should take seriously — through the middle of the trade, not around it. Alphabet raised 2026 capex to $195–205B on a strong quarter and printed the first negative free-cash-flow quarter in its public history, buybacks zero: sold 7.1% next day. Meta nudged its capex floor up to $130–145B, missed on EPS via one-time charges, and offered no 2027 spending clarity: sold 8.6% to $535.25, extending a losing streak CNBC tallies as its longest on record. Microsoft spent just as hard — $35.8B in the quarter — but arrived with Azure accelerating to +43% (crossing $100 billion in annual revenue) and a decade-high beat: paid 16.9% to $456.69, its best day since October 2008, adding roughly $490 billion in market value — the largest single-session market-cap gain any stock has ever recorded. And buried in Microsoft's own numbers sat our memory thesis, itemized: roughly $25 billion of its capex framing is attributed to higher component costs — memory chief among them, as reported.4
Read the three together and the discount-the-spend regime we've tracked since June just got its second axis. The market is no longer punishing capex as a category — it is separating capex with revenue proof attached from capex on faith. Azure at +43% is proof; a negative-FCF quarter with buybacks suspended is faith; a capex floor raised into an earnings miss is faith with a legal bill. That refinement matters for every supplier position we own, because it means the gate doesn't close the question — it hands the question to the next print with proof attached. Amazon and Apple close the gate tonight; their capex and AWS/iPhone results will be handled in a short follow-up note rather than in this week-ahead — the big-4 was tracking north of $700 billion of 2026 capex before their prints.
Memory lived a full cycle inside the same five sessions — and got a new axis of its own. CXMT, China's DRAM champion, debuted Monday in Shanghai and rose 466% in a day to a ~$488 billion valuation — instantly the most valuable listed company on the Chinese mainland — and the global chip complex shed more than a trillion dollars in three sessions as the market repriced Chinese memory from 2028 threat to present-tense fact. SK hynix missed its blockbuster consensus on HBM4 timing; SanDisk collapsed 37% from its July 23 close to Wednesday's trough. Then Thursday, Samsung printed a record quarter, warned the shortage it sees worsens in 2027 and continues into 2028 — and the US complex finally bought a record print instead of selling it: Micron +18%, SanDisk +23%, the SOX +7.7%, though Seoul notably faded its own good news. Demand intact, supply arriving early, both now priced in public — that tension is the sector question our remaining memory exposure lives inside, and the closing note frames it properly.5
5. The book: smaller, marked honestly
With Micron closed, the book is the basket, the toolmaker, the sleeve, and cash — and the honest marks are ugly in the middle. The DRAM basket, held from $59.73, was $58.85 when we graded it a week ago; it printed $44.85 on Wednesday and recovered to $52.34 by Thursday's close — underwater roughly 12%, after trading 25% below basis at the low. Saying it plainly: the basket was built to own the un-capped winners of a shortage, and CXMT just put a second question under that thesis — whether the shortage's duration survives China's ramp. The Q3 contract prints (TrendForce's +13–18% DRAM band) and a proper China-supply underwrite are the adjudicators, and we owe the position the work rather than the assumption. ASML closed Wednesday at $1,550.69, 11.5% below our $1,752.87 fill — and in doing so traded into the published $1,550–1,650 add-zone for the first time since the initiation, before Thursday's bounce lifted it right back to the zone's ceiling — $1,651.44 at the close. The “paid either way” thesis had its first genuinely bad week; the add decision is live and gets its own note rather than a paragraph here. The quantum sleeve did its small, boring job: QTUM ended Thursday at $139.70, flat on the week while the pure-plays it diversifies against sit 62–77% below their highs; Benchmark reinstating Buys across them Monday (“no winner takes all”) reads to us as the sell-side arriving at our basket logic ten months late. Nothing added; QC-ADDS specs (~September 20) remain the next dated catalyst.58
And cash is the heaviest it has been since May — deliberately, and now bigger by the exit's proceeds. The deployment gate we've held since June was never “wait for the prints” as ritual — it was wait for the verdict. The verdict so far is split by the proof axis, Amazon and Apple close the gate tonight, and September now carries a live Fed. A split verdict plus a 5.2% long bond is not a table-pounding setup; it is a setup for staging into proof — basket-first where the thesis survives its underwrite, nothing on faith. The where-the-next-dollar-goes series this exit funds begins after the gate closes; the specific lines are in Section 7.
6. The week ahead: the labor market takes the stand
The Fed argument now runs through the data, and the data comes fast. Friday August 7: the July jobs report — next Friday — the first payrolls print the three dissenters will read as public advocates for a September hike, and the single number most capable of breaking the September coin-flip one way or the other. Before it: Treasury's quarterly refunding announcement Wednesday August 5 — supply meeting a 5.2% thirty-year — plus ISM manufacturing Monday and services Wednesday. Behind it: the July FOMC minutes on ~August 19, which for once are appointment reading — three dissent rationales, in writing, from the blackout's other side. Then August 12: July CPI, the first basket that carries the war — Hormuz was functionally closed the entire month — against the tariff handoff's push-pull. And Warsh speaks at Jackson Hole, August 27–29, a speech he told Wednesday's presser is “a blank piece of paper right now.”8
Earnings hand the referee's whistle from hyperscalers to the second line: Palantir Monday, AMD Tuesday — the accelerator complex's own proof-or-faith moment — Lilly and Disney Wednesday, Super Micro the following Tuesday, and Micron's own FQ4 far out on September 29, a print we'll watch from the research bench with no position and a written trigger list. The war remains the wildcard it was: strikes resumed Wednesday, transits sit near zero, and every headline is a two-sided repricing risk in crude.8
7. The book & trip-wires
The book into the week, one table; four wires below it.
| Position | Basis | Mark (7/30) | Status / thesis tag |
|---|---|---|---|
| Micron (MU) | $668 (May) | CLOSED — runner out ~$900 (Jul 27–28) | COVERAGE CLOSED — ~+53% blended, all realized (the closing note); re-entry by published triggers only |
| DRAM basket | $59.73 (week of July 7) | $52.34 (−12%) | HOLD, thesis under review — the CXMT supply question gets a real underwrite; Q3 contract prints (TrendForce +13–18% band) adjudicate |
| ASML | $1,752.87 (fill; initiation 7/15) | $1,651.44 | ACCUMULATE — the published add-zone ($1,550–1,650) traded Jul 28–29; the add decision gets its own note |
| Quantum sleeve (QTUM) | Venture-sized by design | $139.70 | HOLD — flat on the week vs. pure-plays 62–77% off highs; no adds; QC-ADDS specs ~Sept 20 |
| Cash | — | Heaviest since May (exit proceeds in) | Deployment = staged, proof-first, after the gate's last prints (tonight) and Friday's jobs |
A payrolls print hot enough to push September above ~80%, or a 30-year close above 5.30% — either one, and the defensive duration stance deepens. A sub-100K payrolls print is the other side: the first real crack in the hawkish read.
The curve did the Fed's tightening between meetings; the question is whether the data lets it stick. Response: hot branch — duration stays below-benchmark and equity adds wait for the CPI print; soft branch — we re-examine duration for the first time since June 19, in writing, with the same knife we used on ourselves then.
If the gate closes with proof-backed prints paid and no capex cuts anywhere in the five, staged deployment begins — basket-first where theses survive their underwrites. Any hyperscaler guiding capex down kills the supplier-floor thesis and the ladder with it.
Response: staged and boring — nothing on faith, at least half the cash held through the August 12 CPI regardless, and the where-the-next-dollar-goes series opens with the underwrites, not the buys.
Brent closes above $100 again, or a confirmed strike hits Gulf production or export infrastructure — fields, terminals, processing — as opposed to claims and near-misses.
The first $100 close round-tripped in three sessions; a second one, or verified production damage, converts the war premium from headline noise into a supply-math problem July's CPI will inherit. Response: unchanged and boring on purpose — heavy cash, below-benchmark duration, no war trades. Position size remains the book's only opinion about missiles.
We do not re-enter on green candles — Thursday's 18% was the free sample. Re-entry requires the published triggers: HBM4 re-qualification at scale, FQ4 realized ASPs beating the ceiling model, a genuine 200-day washout — and now an explicit CXMT/China-supply underwrite.
Two weeks ago the re-entry wire was premature — written for a position we still secretly held. It is honest now: the book is flat Micron, the DRAM basket carries the sector exposure, and the name sits on the research bench with FQ4 (September 29) the next scheduled look. Response: any re-initiation gets its own published note with fresh numbers, fresh kill-shots, and a fresh stop — one we have now demonstrated we will automate rather than trust ourselves to obey.
The market's new axis is proof, not spend. Alphabet and Meta spent on faith and were sold; Microsoft spent identically and was paid a record sum, because Azure's +43% made the spending self-evidently productive. If that axis holds through Amazon tonight and AMD Tuesday, the supplier complex we own through the basket and ASML stops trading as one blob and starts trading on whose customers can show receipts. That is a better market for research — and a worse one for momentum — which suits this desk fine.
(1) A Fed credibility spiral — if the long end keeps repricing (30yr > 5.3%) regardless of data, the discount rate does to every duration asset what it did to the Dow on Wednesday. (2) A war resolution that collapses oil — the round trip showed how fast the premium evaporates; a real Hormuz reopening revives the disinflation trade and hawkish positioning underperforms a relief rally. (3) Memory supply news outrunning demand news — CXMT's $488B debut repriced the basket faster than Samsung's shortage warning could defend it; if the China underwrite or the Q3 contract prints break the thesis, we will say so in those words and act on it. A wire is a response plan, not immunity.
Sources & footnotes
- The four wires as published: weekly-outlook-2026-07-20 (“The gate: Alphabet's print meets the new bar” / “The tariff comes off Friday” / “Oil escalation — carried, and armed” / “The Micron runner — no discretion left”), each graded against its wording as published. ↩
- The Micron exit — canonical account in the closing note (July 30): the pre-placed broker stop at $900 (the repair from the July 21 stop note) sold all 15 shares at ~$900 into the July 27–28 breakdown; the published $848.95 close-line fired at Tuesday's $820.53 close against an already-flat book; the July 21 capex-print deadline was satisfied in-window with the exit done. Daily closes: 7/21 $970.82, 7/22 $959.48, 7/23 $990.21 (intraday high $1,011.77), 7/24 $920.95, 7/27 $900.20, 7/28 $820.53 (intraday high $848.36), 7/29 $739.00, 7/30 $874.66 (+18%). Ledger, all realized: ~60% at ~$1,100 average (+65%, June scale-out) + ~40% at ~$900 (+35%) = ~+53% blended off the $668 May 6 basis — matching the July 8 published worst-case floor; never-sold counterfactual ~+31% at Thursday's close; the $848.95-close rule honored mechanically lands near +50%. Sources: verified exchange closes; the closing note; prior published notes. ↩
- FOMC July 28–29: statement and 9–3 vote with Hammack/Kashkari/Logan dissents (Federal Reserve, monetary20260729a); “first triple dissent since September 2016” per AP. Warsh quotes from the Fed's preliminary press-conference transcript (July 29): “no soft inflation target… not on this Committee's watch”; “watchful thinking, not watchful waiting”; “a good family fight”; “we're not going to be constrained by market prices”; “interest rates could well be part of that solution”; intermeeting yield rise “around the top decile” of two decades; “play the ball, not the referee.” No SEP in July; next projections September 15–16. Decision-day markets: Dow −1,153.18 (−2.19%) to 51,594.14 (worst since April 2025), S&P 500 −1.52% to 7,316.15, Nasdaq −1.74% to 24,442.94 (>10% off its high); 2yr −4bp to 4.236%, 10yr to ~4.69%, 30yr +10.5bp to 5.201% close (intraday 5.244%, highest since July 2007) per CNBC closes. BofA “credibility shock” quote as reported by Yahoo Finance. September hike odds: ~76–82% pre-meeting to a ~56–60% cluster after (CME FedWatch / Bloomberg swaps / Polymarket). June PCE (BEA, released July 30): headline −0.1% m/m, +3.7% y/y (from 4.1%); core +0.1% m/m, +3.3% y/y. Attribution note: Wednesday's moves are contaminated by same-day war escalation — flagged, not untangled. ↩
- The capex gate: Alphabet Q2 (July 22 AMC): revenue $119.8B (+24%), Cloud $24.8B (+82%), quarterly capex $44.9B, FY26 guide raised to $195–205B; FCF −$5.86B (first negative quarter since the 2004 IPO); buybacks $0 (second straight quarter); GOOGL −7.13% on 7/23 to $317.69 (Class A). Meta Q2 (July 29 AMC): revenue $60.8B (+28%), EPS $6.18 miss on ~$3.6B one-time charges, 2026 capex guide $130–145B (floor +$5B), no 2027 capex guidance (CNBC); META −8.60% to $535.25 on 7/30; losing-streak-record framing per CNBC. Microsoft FQ4 (July 29 AMC): revenue $90.0B (+17.8%, beat), adj EPS $4.74, Azure +43% (crossed $100B annual), FQ4 capex $35.8B, ~$25B of the FY26 capex framing attributed to higher component costs incl. memory (as reported); MSFT +16.94% to $456.69 on 7/30, best day since October 2008, ~$490B market-cap gain — the largest single-session gain on record, surpassing Nvidia's April 2025 mark. Amazon/Apple: report tonight, July 30 AMC; graded in the next note. Sources: company releases; CNBC/Bloomberg/Benzinga coverage; verified closes. ↩
- Memory & semis week: CXMT listed Monday July 27 (Shanghai STAR): priced ¥8.66, first-day close ¥49 (+466%), ~¥3.3T (~$488B) valuation — China's most valuable onshore-listed company (via CNBC/Fortune/Bloomberg); global chip complex shed >$1 trillion in three sessions (CNBC). SK hynix Q2 (released July 28 ET): revenue and operating profit missed blockbuster consensus on HBM4 shipment timing (mass shipments began in Q2, H2 ramp); SKHY ADR made post-listing lows before Thursday's +15% rebound. SanDisk: −37% from its July 23 close to Wednesday's trough; +23% Thursday. Samsung Q2 (July 30 KST): record revenue ₩171.5T and operating profit ₩89.49T (up ~19x y/y); memory shortage “expected to worsen” in 2027 and “continue in 2028”; Seoul faded the print (Samsung ~−1%) while the US complex rallied. SOX: closing peak 14,247.0 (June 30) → 10,447.5 July 29 (−26.7%) → 11,250.7 July 30 (−21.0%; still a bear market). DRAM basket closes: $58.85 (7/21) → $44.85 (7/29) → $52.34 (7/30). ASML: $1,550.69 (7/29) vs the $1,752.87 fill; add-zone $1,550–1,650 traded 7/28–29; $1,651.44 (7/30). 7/30 marks per the bpleon price feed. TrendForce 3Q26 forecast (July 3): conventional DRAM +13–18% q/q, NAND +10–15%. Sources: exchange data; company releases; TrendForce; CNBC/Reuters coverage; the closing note. ↩
- Oil & the war: WTI settles — $86.83 (7/22, derived), $92.19 (7/23; through the wire's ~$90 line; highest settle since early June), $89.78 (7/24), $82.61 (7/27, −7.5%), $79.26 (7/28; −14.0% in three sessions, worst since 2020), $84.46 (7/29, +6.6%). Brent: $100.69 (7/23; first close above $100 since May), $84.09 (7/28), $90.74 (7/29, +7.9%). Hormuz: transits in single digits (Kpler counted 10 on 7/23 vs ~88/day norm); escorted convoys only. War arc: US completed a 13th consecutive strike wave 7/24, suspended strikes for diplomacy over the weekend, talks collapsed 7/29 with an Iranian ballistic-missile attack on US forces (including a base in Jordan) and a major US retaliatory wave. Production infrastructure: no confirmed strike on Gulf fields or export terminals; Houthi claims on the Saudi east–west corridor unconfirmed; the reported Ras Laffan halt traces only to Iranian state-affiliated media and is not treated as fact here. Sources: exchange settlements; Reuters/CNBC; Kpler as reported. ↩
- Tariffs & Treasury: Section 122's 10% surcharge lapsed 12:01 a.m. ET July 24 at its 150-day statutory cap (the CIT had already struck it; collection ran to the sunset under a Federal Circuit stay; refunds litigation continues). USTR Section 301 forced-labor final action (released July 23, effective 12:01 a.m. July 24): 10% on 17 partners / 12.5% on the rest of 60 economies (~99.4% of imports), USMCA-compliant goods exempt, ~2,120 annex codes with 863 exempt as-entered; no sunset. Auctions: the July 22 20-year ($13B) stopped at 5.163% with a tail and heavy dealer takedown (graded D+ by InvestingLive); the July 23 10-year TIPS stopped at a 2.438% real yield — highest for the tenor since October 2008 (TipsWatch). Friday July 24 closes: S&P 7,411.98 (−0.61% wk), Dow 51,947.25, Nasdaq 24,975.82 (−2.13% wk). Sources: USTR/Federal Register trackers; Global Trade Alert; auction specialists as named. ↩
- Calendar & quantum: Amazon + Apple tonight (July 30 AMC; Apple's call is Tim Cook's last as CEO — John Ternus takes over September 1). July jobs: Friday August 7 (BLS). Treasury quarterly refunding announcement: Wednesday August 5. July CPI: Wednesday August 12 — the first full war-oil basket. FOMC minutes: ~August 19 (three-week convention). Jackson Hole: August 27–29. Next FOMC: September 15–16 (with SEP). Next-week earnings: PLTR Aug 3, AMD Aug 4, LLY + DIS Aug 5, SMCI Aug 11; MU FQ4 September 29. Section 338 Canada 50% tariff effective ~August 19 (law-firm trackers). Quantum: QTUM $132.24 (7/29) → $139.70 (7/30, +5.6%), ~flat vs the prior Friday and −22% from its 52-week high; pure-plays IONQ/RGTI/QBTS/QUBT 62–77% below 52-week highs; Benchmark reinstated Buys on IONQ/QBTS/RGTI Monday (“no winner takes all”); IONQ's Thursday pop aided by SkyWater-acquisition approval; QC-ADDS technical specs due ~September 20 under the June 22 quantum EO. Sources: BLS/BEA/Treasury/Fed calendars; company IR; exchange closes. ↩
Methodology & definitions: “Bear market” and “correction” are the standard price-action heuristics (20%+ and 10%+ off a closing high), not official designations. The “DRAM basket” refers to the memory ETF spanning the DRAM/HBM complex (Samsung, SK hynix, Micron) and NAND/storage (SanDisk, Western Digital, Seagate). Position marks are the most recent settled closes against stated basis. Realized figures reflect actual fills as disclosed in the linked notes; rule-honored counterfactuals assume execution at the close or open the rule specifies; open positions are marked, not realized. Trip-wire responses are pre-committed in the prior week's note and graded against the wording as published. Single-source figures are attributed inline to their source.
Nothing on this page is investment advice. We work in scenarios and trip-wires, not price targets — everything above is for thought and process, not for trading. Forward-looking statements are scenarios, not promises. See disclaimer.