The S&P 500 gained 3.57% on the week to a record 7,757.46, with the Nasdaq 100 up 5.12% and every major average finishing green. The driver rotated daily but never left one axis: a Strait of Hormuz reopening was priced, unpriced and repriced across four sessions, and then a July payroll contraction of 23,000 removed a September rate hike from the front end on Friday and delivered the record. The defining feature of the week is that both of its dominant inputs — an unsigned Middle East arrangement and a labour market shedding jobs — were read by the market as relief rather than as risk.

1. The Hormuz Round Trip: Strikes Called Off → a Deal “Today or Tomorrow” → Iran’s Restrictive Draft — Crude Ends the Week Down 8.8% With the Strait Still Shut

The core facts:Monday, President Trump held off on threatened strikes against Iran and said talks to revive cargo traffic through the Strait of Hormuz would begin; WTI fell 5.52% and Brent 5.00%. Tehran immediately said no negotiations with Washington were under way, only Oman-mediated discussions on managing the strait. Tuesday, Treasury Secretary Bessent told CNBC a deal could land “today or tomorrow” with “freedom of movement” and no Iranian toll; Brent broke $80, falling a further 6.08% to $78.68. Wednesday the EIA confirmed a roughly 2.5 million-barrel crude build against an expected draw — a second consecutive build breaking a seven-week drawdown streak — and a Houthi missile strike that set the Saudi tanker Encelia alight off Yanbu lifted Brent above $80 intraday before it closed lower anyway. Thursday, Iranian state agency Fars published a draft transit plan materially more restrictive than the market had priced; Brent rallied 4.72% to $83.20 and WTI 3.60%. Friday gave the entire spike back, WTI closing at $77.15 and Brent at $82.27, on reports the draft has still not cleared Iran’s parliament. The strait has been largely blocked since February 28, 2026. No agreement has been signed.

Why it matters:Five sessions of violent repricing occurred without a single barrel moving through the strait. What traded was not supply but the market’s estimate of when supply returns, and by Friday that estimate had hardened into an assumption: crude fell on a day whose only fresh news — a parliamentary hurdle — pointed toward delay. A market that sells the commodity on a bearish headline for the event it is pricing has stopped treating that event as conditional. The week also quietly supplied the one leg that owes nothing to diplomacy. Two consecutive inventory builds are a physical fact, and they are why WTI fell 8.80% against Brent’s 6.46% — the US-specific benchmark underperformed the global one during a global chokepoint disruption, which is the wrong way round unless domestic fundamentals are doing independent work. That gives the disinflationary impulse an anchor that survives a diplomatic failure. What it does not do is compensate the risk. Energy fell 3.66% on the week against the S&P’s 3.57% gain — a seven-point gap, and the sector that has led the index all year (+28.61% YTD) was the only one to lose more than 1% on Friday’s record day (see the sector rotation table and the Energy table in Section B; Chevron and Exxon both sit in the weekly decliners). The asymmetry from here is unattractive: a reopening on restrictive terms is worth far less than the clean reopening now embedded in the price, and a collapse in the talks reprices violently from a level carrying no cushion.

What to watch:The Iranian parliament’s vote on the transit draft — now the specific gate between a framework and an actual reopening. The clause that matters is whether transit conditions apply to all flags or only to designated corridors, since that is precisely the routing question that collapsed the signed June 17 memorandum after traffic had already resumed. Tanker transit counts, not statements, are the only confirmation worth acting on.

↑ back to summary

2. Five Fed Officials Spend Four Days Arguing for a Hike — and One Payroll Print Ends the Debate on Friday

The core facts:Monday, New York Fed President Williams said the Fed will raise rates if inflation does not ease, citing June PCE at 3.7% and putting a 2028 date on the return to target. Tuesday, Philadelphia’s Paulson kept a hike explicitly on the table, called the 9–3 July hold “not a close call,” and estimated underlying inflation at 2.4%–2.8%. Wednesday, Minneapolis’s Kashkari said “now is the time” to begin raising and named September 15–16 as a possible start, while Governor Lisa Cook — a permanent voter — said the committee may not have “the luxury of waiting.” Thursday, St. Louis’s Musalem argued for tighter policy on a labour market he described as “stabilized.” Friday, July nonfarm payrolls contracted 23,000 against a consensus near +83,000. September hike odds fell across the week from roughly 67% to 40%. Both the 2-year and 10-year yields ended the week down exactly 6.9 basis points, and the S&P 500 closed at a record 7,757.46.

Why it matters:Four consecutive sessions of escalating hawkish rhetoric — rising in seniority from a regional president to a sitting Board governor — moved the priced probability of a hike in the opposite direction every single day. That is the week’s most useful institutional fact, and it is not a story about the Fed being wrong. It is a story about the market having decided the argument will be settled by data the committee has not yet seen, and then being handed exactly that data on Friday. Musalem’s description of a “stabilized” labour market was contradicted by the BLS inside twenty-four hours. The market’s reading is legible in the composition rather than the direction. The Russell 2000 outpaced the Dow by nearly four to one, the 2-year fell twice as far as the 10-year, the dollar softened and volatility declined — four signatures of policy relief, and none of the four that a genuine growth scare produces. Utilities and Real Estate, the two weakest sectors on Thursday, both turned green on Friday. This is a rate trade, cleanly expressed. The discomfort is that the same print is also the strongest argument against the rally, and the market declined to price it. Payrolls have gone negative, average hourly earnings growth has slowed to its weakest since May 2021, and none of that supports the +50.4% blended earnings growth the index is capitalising. Reading a job loss purely as a policy input works only while the weakness stays inside hiring and out of demand — and copper, down 1.87% on the same session gold rose 2.31%, is already taking the other side (see the Commodities table in Section B).

What to watch:Wednesday’s July CPI and Thursday’s PPI are the two prints between here and the September meeting that can restore the hawkish case; a hot core reading would reopen the argument the payroll number appeared to close. Watch whether Kashkari, Cook or Musalem revise the “stabilized labour market” framing, and whether the Russell 2000 holds its leadership over the Dow — the cleanest single test of whether the rate-relief trade has legs.

↑ back to summary

3. The AI Trade Stops Paying for Revenue and Starts Pricing What It Costs — Four Beat-and-Raise Prints Sell Off in One Week

The core facts:Tuesday the semiconductor complex reversed violently, Intel and SanDisk each adding 10.84% and Technology leading all sectors at +4.25%. That evening AMD beat on revenue, earnings and margin, guided Q3 roughly $480 million above consensus, and fell 8% after hours on capital expenditure of $808 million against a $299 million model — a roughly 40% sequential cut to free cash flow. Wednesday the selling reached names with no earnings event: Lam Research −3.25%, Palantir −2.61%, Applied Materials and Texas Instruments down 2–3%. NVIDIA diverged, rising 3.43%. Wednesday evening SanDisk reported revenue up 372% at an 84.6% gross margin with a $14 billion buyback and fell; Western Digital beat, raised and fell; both closed down roughly 10% and 11% respectively on Thursday. AppLovin missed revenue by 1% and lost a fifth of its value. Thursday, Dell gave back 5.41% and Arista 2.53% while Microsoft added 2.54%. Disclosed 2026 capex targets stand near $220 billion for Amazon, $200 billion for Alphabet, $175 billion for Microsoft and $130 billion-plus for Meta. Arista, Shopify and Cloudflare all beat, raised and rallied. Technology finished the week +6.92%.

Why it matters:Four companies beat and raised inside four sessions and were sold anyway. That is not a demand verdict — AMD’s data-centre revenue more than doubled, SanDisk expanded gross margin by 5,800 basis points, Cloudflare’s AI-agent traffic crossed half its network — and it is not a sector-wide exit either, since Technology finished the week as the second-strongest sector. What the week did was install a new sorting rule, and by Friday it had been applied consistently enough to read. Capital intensity is now the discriminator. Companies that capture AI spend without buying capacity to serve it — Arista at a 49.9% operating margin, Cloudflare raising both revenue and earnings guidance, Microsoft booking Azure consumption — held or gained. Companies that must fund growth from the cash-flow statement were marked down regardless of the income statement. The useful test is that this reached names with no disclosure of their own: Lam Research and Palantir fell on Wednesday having reported nothing, which is what distinguishes a theme re-rating from a run of bad prints. Two things argue against reading it as the start of an unwind, and they matter. NVIDIA rose against the complex on its worst day, which is rotation toward unquestioned margin rather than exit. And New York Fed President Williams explicitly declined to call AI a bubble on Thursday, grounding it specifically on the observation that the borrowing funding the buildout sits on balance sheets generating the earnings to carry it — a financial-stability judgement, not a valuation one. That is the difference between a repricing that costs investors money and one that reaches credit and costs the economy growth.

What to watch:Applied Materials on Thursday, August 13 is the cleanest available test — wafer-fabrication equipment orders either corroborate the capex the hyperscalers have guided to or they do not, and it is the one print that speaks for the whole chain rather than a single link. Cisco on Wednesday, August 12 carries the networking read. Watch investment-grade spreads for the technology and utility issuers financing data-centre construction, which is the channel Williams is implicitly saying remains healthy.

↑ back to summary

4. SpaceX Posts a Record Loss of Confidence and Then Absorbs $100 Billion of Unlocked Stock — the Shares End the Week Up 22.8%

The core facts:Tuesday after the close, SpaceX delivered its first earnings report as a listed company: revenue of $7.81 billion against $6.83 billion consensus, up 92% year over year; GAAP EPS of −$0.09 against −$0.23 expected; Starlink subscribers doubling to 12 million; the AI segment turning EBITDA-positive on $14.1 billion of new contracts. Capital expenditure was $18.4 billion against roughly $13 billion modelled, of which $15.83 billion was AI-related, and average revenue per Starlink subscriber fell 22% year over year. The stock fell 13.60% Wednesday to an all-time low of $108.29. Also Wednesday, COO Gwynne Shotwell outlined a full Starlink Mobile network built on roughly 65 MHz of spectrum acquired from EchoStar, with next-generation satellites in 2027; American Tower, Crown Castle and SBA Communications each fell about 6%. Thursday the first post-IPO lock-up released up to 911.5 million shares — roughly $100 billion — lifting the free float from 4.9% to 11.8% of shares outstanding; the stock traded to a new low of $105.11 and closed up 6.14%. Friday it added 15.83% to $133.11. A further 319 million shares unlock August 12.

Why it matters:Lock-up expiries resolve downward with enough regularity that the exceptions are worth studying. This one more than doubled the tradable float in a single session, into a stock that had made an all-time low the day before, on a week when the broad market was distracted — and the buyer base took all of it and bid the shares 22.8% higher across five days. Until Thursday the float was too small for price to say anything reliable about depth of demand. It now has said something, and it is the first genuine information the market has had about this constituent since the June listing. The company is also the cleanest single illustration of story #3’s sorting rule, which is why it belongs beside it rather than inside it. Every operating metric beat, and the stock still lost 13.6% on the capital-expenditure line alone — $18.4 billion of quarterly spend against $7.81 billion of quarterly revenue means the equity is the funding instrument and dilution is structural rather than calendar-driven. The Starlink Mobile disclosure adds a second-order consequence the tower REITs priced immediately: spectrum acquisition is the input a satellite operator does not need to fill coverage gaps and does need to compete for primary connectivity. Two constraints belong on the constructive read. Thursday released the first tranche, not the last — more than 4 billion shares are expected tradable by year-end, several times what cleared this week. And the close still sits roughly 15% below the $135 IPO price and about 41% below the June 16 peak.

What to watch:The August 12 unlock of 319 million shares — a second absorption without a new low would establish Thursday as depth rather than a squeeze. Watch whether the stock reclaims $135, the level at which remaining locked holders move from underwater to profitable and the supply calculus changes, and whether management supplies a capital-expenditure ceiling, the one disclosure that would let the market underwrite the spending.

↑ back to summary

5. Alphabet Loses Its DeepMind CEO, Its Chief Scientist and Several Gemini Leaders on One Day — With $200 Billion of Capex Guided and the Flagship Model Still Unshipped

The core facts:Wednesday, Alphabet announced an overhaul of its artificial-intelligence organisation. Demis Hassabis gave up the DeepMind chief executive role to become DeepMind chair and chief scientist across Alphabet, retaining Isomorphic Labs. Google chief scientist Jeff Dean, a 27-year veteran, is leaving alongside senior fellow Sanjay Ghemawat to found Discovery Loop, an independent public benefit corporation in which Google will invest. Several Gemini model leaders are also departing. Koray Kavukcuoglu, DeepMind CTO and Google’s chief AI architect, becomes senior vice president of DeepMind reporting directly to Sundar Pichai. Alphabet Class A fell 4.03% to $362.43, the session’s third-largest mega-cap decline and the principal driver of Communication Services’ 2.36% drop — the worst of the eleven S&P sectors that day. The reshuffle lands with the flagship version of the latest Gemini model still unreleased against a planned June launch, and with 2026 capital expenditure guided near $200 billion.

Why it matters:Executive reshuffles at large technology companies are ordinarily noise and the market ordinarily ignores them. It did not ignore this one, and the reason is the conjunction rather than any single departure. A company committing roughly $200 billion of capital in 2026 simultaneously lost the executive running its research organisation, the technical centre of gravity of the last 27 years, and several leaders of the specific model franchise that spending is meant to produce — while that model runs roughly two months past its planned launch. Any one of those is survivable. Together they pose a question the market cannot answer from outside: did the delay cause the reorganisation, or will the reorganisation extend the delay? The 4% decline is the price of that ambiguity rather than a verdict on the people involved. The timing against story #3 is what elevates this above a governance footnote. On the same session the market began separating AI revenue from the capital required to produce it, the complex’s second-largest spender disclosed that the organisation converting that capital into product is being rebuilt. Investors are being asked to underwrite $200 billion against a research leadership reconstituted that morning and a flagship model they have not seen. Sector arithmetic confirms it stuck: Communication Services is now the worst three-month sector in the index at −8.17% (see the sector rotation table in Section B), and its +1.18% week was the fourth weakest despite a broadly rising tape. The genuine counterweight is that Hassabis is not leaving, has moved to a wider remit, and Dean’s exit is amicable enough that Google is investing in his new vehicle.

What to watch:The release date of the flagship Gemini model — shipping it retires the delay question and reframes the reshuffle as a reorganisation around a finished product rather than a stalled one. Watch for further Gemini team departures, which would shift this from planned succession to exodus, and watch Communication Services against the index, since the sector has now been the quarter’s laggard through two separate causes.

↑ back to summary

6. Trump Signs a Section 232 Polysilicon Proclamation Pairing a 15% Tariff With Minimum Import Prices — the Price Floor, Not the Tariff, Is the Instrument

The core facts:Flagged Wednesday as imminent, the proclamation “Adjusting Imports of Polysilicon and its Derivatives into the United States” was published dated August 6 after Thursday’s close, with the equity reaction running through Friday. It pairs a 15% ad valorem tariff with minimum import price floors across the chain — reported at $21 per kilogram for polysilicon and $0.38 per watt for modules, with separate minimums for wafers and cells — framed under Section 232 as protection for US polysilicon production against Chinese dominance of both the solar and semiconductor feedstock chains. Importers investing in US wafer and cell production can offset the cost. Estimated impact on imported solar cells is roughly $0.10 per watt, or $600–$800 per installation. Measures take effect at 12:01 a.m. Eastern on December 4, 2026. First Solar rose 7.73% in Thursday’s after-hours session and traded up 8.95% at $266 pre-market Friday, with SolarEdge, Sunrun and Enphase also higher; CEO Mark Widmar called it “one of the most strategically significant trade measures in decades.”

Why it matters:The headline 15% is the least consequential number in the action. A minimum import price is a categorically different instrument from a duty: an ad valorem tariff can be absorbed through margin, currency, transshipment or under-invoicing, and Chinese solar supply chains have spent a decade demonstrating exactly that against successive US measures. A floor cannot be absorbed, because it does not tax the price — it sets it. At $0.38 per watt for modules, competing below a defined level becomes impossible regardless of cost structure, which is why the domestic producer with the most integrated US manufacturing base reacted as it did. This is also a semiconductor action that has been reported as a solar one. Polysilicon is the feedstock for wafers, and invoking Section 232 places it alongside steel, aluminium and copper — the buildout of domestic chip capacity is now being treated as a supply-chain security problem rather than a subsidy problem, a framing that tends to survive changes of administration in a way tax credits do not. It also carries materially less litigation risk than the Section 301 programme in story #7, which rests on a contested forced-labor rationale rather than national security. The cost side is real and lands on the other big theme of the week: every imported module gets more expensive from December 4, and the largest incremental buyer of US utility-scale solar is the data-centre complex, whose capital intensity the market spent this week scrutinising. Raising the input cost of power generation for that buildout is a genuine offset to the domestic-manufacturer gain, and the four-month lead time invites an import pull-forward that flatters Q4 volumes and depresses 2027.

What to watch:Whether the minimum prices extend to finished wafers used in semiconductors or stop at solar-grade material — that single scope question decides whether this is a solar story or a chip-supply-chain story. Watch module import volumes into the fourth quarter for a pull-forward, and whether utility-scale developers begin flagging higher project costs in guidance, which is where the cost side first becomes visible.

↑ back to summary

7. Three Multi-State Suits Hit the Section 301 Tariffs in Forty-Eight Hours While Canada’s 50% Section 338 Deadline Closes to Eleven Days

The core facts:Monday, twenty-five Democratic-led states sued in the US Court of International Trade — the venue with exclusive first-instance jurisdiction — challenging 10% and 12.5% tariffs imposed July 24 on 60 trading partners under Section 301 over forced-labor allegations, arguing the rationale is a pretext to revive duties the Supreme Court already struck down. Tuesday a third multi-state action was filed against the USTR, using USTR officials’ own public statements as evidence the programme’s purpose was revenue rather than remediation; the underlying investigation was a single blanket three-month review covering more than 99% of US imports, with Section 232 goods carved out. Private actions from Burlap and Barrel and Collective Horology seek removal and refunds. On the Canadian track, an additional 50% duty on roughly $20 billion of goods across 554 tariff lines takes effect August 19 under three Section 338 proclamations signed July 20 — the statute’s first invocation since the 1940s, with no exemption for USMCA-originating goods. Trade Minister LeBlanc and chief negotiator Charette made a second Washington trip in two weeks; Thursday, Prime Minister Carney called Canada’s posture already “quite firm.” USTR Greer told a Senate committee he hopes to present interim USMCA options before year-end.

Why it matters:The market has largely stopped pricing tariff headlines, and this is the week that habit became expensive to keep. Three multi-state actions inside forty-eight hours in the one court with exclusive jurisdiction is a different proposition from a trade-association complaint: the plaintiffs have standing difficult to contest, and the third filing attacks the statutory authority itself rather than the rate. A claim built on the agency’s own statements that the forced-labor rationale was pretextual either survives a motion to dismiss or ends the programme — there is very little middle ground, and a blanket three-month investigation spanning 99% of imports is an unusual administrative record to defend. The exposure runs both directions, which is why this is uncertain rather than negative: firms that absorbed the duties face refund upside and continued margin drag on the two outcomes, firms that passed them through face the mirror image, and neither sits in consensus estimates because the timeline is unknown. The Canadian track is the half with a hard date and it is now inside two weeks. The structural feature that matters more than the headline rate is the absence of a USMCA carve-out: a 50% duty applying without regard to origin rules removes the mechanism by which North American manufacturers have organised cross-border supply chains for three decades, and firms cannot re-source around a rule that ignores sourcing. That is why the $20 billion trade-flow figure understates the disruption — the affected volume is modest, the invalidated planning assumption is not. Carney describing his position as already firm eleven days out signals the required concessions have been identified and declined. Greer’s interim-USMCA track is genuinely two-sided but the timelines do not meet: relief before year-end arrives months after August 19 has already changed behaviour.

What to watch:August 19 is the hard date. Any announcement before it that Canada has withdrawn the provincial alcohol bans, the dairy quota rule or the auto surtax is the only fast path to suspension, since those three actions are the stated trigger. Watch for a preliminary injunction in any of the three Court of International Trade actions, which would suspend collection and force immediate re-estimation across import-heavy sectors, and watch North American auto and rail names for whether the market believes the deadline holds.

↑ back to summary

8. Big Pharma Concedes It Cannot Grow Organically — a Reported $400 Billion Merger, a $5.7 Billion Pipeline Charge and a Cluster of Deals in Four Sessions

The core facts:Monday the Financial Times reported AstraZeneca and Bristol Myers Squibb have held months of talks on a merger valuing the combined company near $400 billion, which would create the world’s fourth-largest drugmaker. The reaction was starkly asymmetric: AstraZeneca fell as much as 7% and closed down close to 9% in London while Bristol Myers gained roughly 6% in US premarket; analysts were described as “perplexed” by the logic, and the combined oncology overlap would attract significant antitrust scrutiny. Both companies declined to comment. The same session, Lilly fell 2.39%, AbbVie 2.33% and Merck 1.87%, and Healthcare closed red as one of only three negative sectors on a day the S&P rose 1.48%. Three of five definitive deals announced Monday were healthcare — KKR/Integer at a $5.7 billion enterprise value, Curium/Lantheus at up to $114.50 a share, and Indivior/Supernus. Tuesday, Merck raised full-year revenue guidance to $66.3–$67.3 billion while cutting adjusted EPS guidance to absorb a one-time $5.7 billion charge, equal to $2.31 per share, on the Terns Pharmaceuticals acquisition. Argus upgraded Bristol Myers Squibb twice in three sessions.

Why it matters:The market’s verdict on the reported merger was delivered immediately and unambiguously: the buyer was punished considerably harder than the target was rewarded. That asymmetry is the signature of shareholders concluding an acquirer is paying up to solve a problem rather than to capture an opportunity — here, the patent-cliff arithmetic facing large-cap pharma through the second half of the decade. Buying scale in oncology does not manufacture new molecules; it consolidates two pipelines under one cost base and buys time. What makes this a week-level story rather than a Monday headline is that Merck then did the same thing in miniature and disclosed the price. A $5.7 billion charge converting a revenue beat into a profit-guidance cut is the identical admission with an invoice attached, and it arrived within twenty-four hours of the first. Two incumbents, two sessions, one conclusion: organic growth will not close the gap. The read-through runs wider than the four names. When a sector’s largest constituents signal that scale must be bought rather than built, every standalone pipeline in the cohort is repriced against that admission — which depresses multiples across the group while simultaneously embedding a takeover option in the mid-caps, and explains why three of Monday’s five definitive deals were healthcare assets. The counterweight is genuine and keeps this uncertain rather than bearish. This is a report of talks, not an announcement: no terms exist, no board has approved anything, and the sources themselves flagged that no agreement may result. Healthcare also finished the week up 2.24% and remains the strongest three-month sector at +12.65%, so the consolidation signal is a re-rating of how growth is financed rather than evidence the sector is deteriorating.

What to watch:Whether either board confirms or denies the talks — continued silence after a move of this magnitude is itself informative — and whether AstraZeneca recovers the 9%, the cleanest measure of whether shareholders will tolerate the deal being pursued at all. Watch business-development commentary from remaining large-cap pharma for whether Terns was the last acquisition of its size or the first of several.

↑ back to summary

9. The White House Restarts the Clock on Removing Fed Governor Lisa Cook — and the 21 Days Run Straight Into the September FOMC

The core facts:Deputy White House Chief of Staff Daniel Scavino sent Federal Reserve Governor Lisa Cook a letter dated August 5, disclosed publicly Friday, demanding she answer allegations that she made false statements on mortgage applications and giving her 21 days to submit evidence or argument to the White House. The allegations originated with William Pulte, who oversees Fannie Mae and Freddie Mac. The procedural posture traces to June, when the Supreme Court ruled 5–4 that the Federal Reserve occupies a special place in government and that a president may remove a governor only for cause — expressly leaving it to the lower courts to determine whether these particular allegations meet that standard. The ruling was widely reported at the time as blocking the removal; it did not, it constrained the method. Coverage was confirmed by the Washington Post and NPR affiliates. Cook is a permanent voter and had, two days earlier, said the committee may not have the luxury of waiting to raise rates. Markets showed no visible reaction: the 2-year yield fell 4.8 basis points, the dollar index declined 0.32% and gold rose 2.31%.

Why it matters:The market treated the June ruling as the end of this matter, and that is the misreading the letter corrects. What the Court held was not that Cook is unremovable but that she is entitled to process, and that whether mortgage allegations constitute cause belongs to the lower courts. The administration has now started exactly that process, and the arithmetic of the timetable is the whole story: 21 days from August 5 places a White House determination inside the window running up to the September 15–16 FOMC — the meeting the market spent this entire week repricing from a possible hike to a hold, per story #2. A contested removal proceeding against a sitting governor overlapping with that decision is a materially different institutional backdrop than the one now priced. The mechanism that matters for portfolios is not who occupies the seat. Federal Reserve independence is priced into the long end of the Treasury curve and into the dollar as a standing assumption rather than as a variable, and assumptions of that kind reprice discontinuously rather than gradually. There is a second irony worth holding: Cook is one of the officials who spent this week arguing for tighter policy, so the removal effort is not obviously an attempt to install a more dovish committee than the one already there. The market’s non-reaction is genuine evidence and deserves weight — participants have seen versions of this and concluded, so far correctly, that it resolves without changing the committee. But the standing basis for that view was the June ruling, and the June ruling has just turned out to permit precisely the outcome its interpreters said it foreclosed.

What to watch:The 21-day response deadline in late August, immediately preceding the September FOMC — a White House determination inside that window is the escalation that forces a market reaction. Watch the 30-year yield and the dollar rather than the front end, since institutional-credibility risk prices in term premium and the currency, not in the near-term policy path.

↑ back to summary

1. July Payrolls Contract 23,000 Against a +83,000 Consensus — the First Negative Print in Months (BLS, Friday Aug 7)

What they’re saying:Nonfarm payrolls fell 23,000 in July against a Dow Jones consensus of +83,000, a 106,000 swing. Government payrolls dropped 53,000, with additional softness in retail, leisure and hospitality and slower healthcare hiring. The unemployment rate ticked down to 4.1% from 4.2% — but for the wrong reason, as the participation rate slipped to 61.4% and fewer people were working or looking for work. Average hourly earnings growth slowed to 3.2% year over year, the weakest pace since May 2021.

The context:The market read this as a policy input rather than a growth input, and the composition of the reaction shows the reading was deliberate rather than reflexive. The 2-year fell 8 basis points intraday to 4.16% against the 10-year’s 6, the Russell 2000 outpaced the Dow nearly four to one, the dollar softened and the VIX declined — four signatures of rate relief, none of which a genuine growth scare produces. September hike odds fell from above 50% to roughly 40%, and the S&P closed at a record. But the print landed one day after St. Louis Fed President Musalem argued for tighter policy on a labour market he called “stabilized,” and Richmond’s Barkin, speaking hours after the release, called 4.1% unemployment “neither loose nor tight” while flagging “cause for caution” if deterioration continues. Copper fell 1.87% the same session gold rose 2.31% (see the Commodities table in Section B) — the industrial complex pricing the same number as demand rather than as policy.

What to watch:The August employment report on September 4, which lands after the September 15–16 FOMC rather than before it — meaning this print, not a revision to it, is the labour data the committee takes into the meeting.

2. The Growth-Versus-Hiring Gap Opens to the Widest of the Cycle — ISM Manufacturing at a Four-Year High and GDPNow Near 6% While Every Labour Instrument Softens (ISM / Atlanta Fed / Census, Mon–Fri)

What they’re saying:Monday’s ISM Manufacturing PMI printed 55.6 against a 54.0 consensus and 53.3 prior — the strongest since May 2022 and a seventh straight expansionary month — with the employment index jumping to 52.8 from 49.7 and Chicago PMI corroborating at 57.6. The Atlanta Fed’s GDPNow nowcast for Q3 leapt to 6.2% from 5.0%, lifting personal-consumption nowcasts to 4.6% and private investment to 17.9%, before easing to 5.8% by Thursday. Against that: ADP private payrolls added just 44,000 in July against consensus near 68,000–70,000; JOLTS openings fell to 7.359 million from a revised 7.537 million; the ISM services employment index reversed 3.8 points to 47.4; June factory orders fell 0.3% with core orders down 0.4% against a +0.4% expectation; and the trade deficit narrowed to $73.3 billion on falling exports and imports alike.

The context:GDPNow tracks hard spending and investment data; payrolls track hiring. The two have never been this far apart in this cycle, and the week is what opened the gap — a nowcast near 6% and a negative payroll print inside five sessions. The reconciliation that fits all of it is that firms are converting demand into output rather than into headcount, which is a productivity story for margins and a confidence story for employment, and the data cannot yet distinguish which. Note the internal contradiction on the goods side too: a manufacturing survey at a four-year high in the same week factory orders missed badly and the trade deficit improved on shrinking volumes on both sides. Surveys measure sentiment about direction; orders measure dollars. The market resolved the ambiguity by ignoring the growth leg entirely — yields fell on the labour data and never responded to the nowcast.

What to watch:The next GDPNow update and whether it is revised down to absorb the weaker labour backdrop — a nowcast that holds near 6% through a negative payroll month would make the divergence structural rather than a timing artefact.

3. Low-Fire, Low-Hire Hardens: Claims Below 200K for a Third Week, Layoffs at a Two-Year Low, Unit Labour Costs Undershoot by Nearly a Point (BLS / Labor Dept / Challenger, Thursday Aug 6)

What they’re saying:Initial claims for the week ended August 1 came in at 199,000 against a 202,000 consensus, a third straight week below 200,000 — the longest such streak since 1969 — with the four-week average falling to 198,750, its lowest since September 2022. Challenger reported 33,429 announced job cuts in July, down 27% from June and the lowest monthly total in two years, with year-to-date cuts of 477,033 running 41% below the same period in 2025; AI was the leading cited cause for a fifth consecutive month at 10,970 cuts. Q2 nonfarm productivity rose 1.4% while unit labour costs rose just 1.3% against a 2.1% consensus, with year-over-year productivity up 2.2%.

The context:Taken together these three releases describe the specific labour market the Fed has spent two years asking for: nobody is being fired, labour costs are not financing inflation, and output per hour is rising. It is also the exact configuration that makes the softness elsewhere in the week readable as a hiring freeze rather than a downturn — a labour market shedding 23,000 payrolls while claims sit at a 57-year low is not one where separations are driving the decline. The market gave the disinflationary content almost nothing on the day: yields rose 5.8 and 6.4 basis points because Brent added 4.72% the same morning. Two inflation channels moved in opposite directions and the one with the larger daily variance won the tape. The caveat that keeps this from being unambiguous is the Challenger composition — AI leading the cited causes for a fifth straight month says the displacement running underneath the falling headline is structural and concentrated in white-collar and technology roles.

What to watch:Thursday’s initial claims for the week of August 8 — a fourth consecutive sub-200,000 print alongside a negative payroll month would confirm that hiring, not firing, is the entire story. Q3 preliminary productivity in early November tests whether the cost restraint is durable.

4. Treasury Lifts Q3 Borrowing by $68 Billion to $739 Billion and Sets a $125 Billion Refunding for Next Week (US Treasury, Wednesday Aug 5)

What they’re saying:The quarterly refunding statement raised the July–September net privately-held marketable borrowing estimate to $739 billion, $68 billion above the May forecast, citing lower projected net cash flows. Treasury separately detailed a $125 billion August refunding covering $96.3 billion of maturing notes and bonds plus roughly $28.7 billion of new financing, with 3-year, 10-year and 30-year auctions scheduled for August 11–13. The revision sits against a roughly $1.9 trillion FY2026 deficit projected by the CBO.

The context:A $68 billion upward revision inside a single quarter produced no visible concession in yields on the day — the 10-year actually fell 1.3 basis points, and the tape attributed that to the ADP miss. That is the normal pattern and the reason to carry this forward rather than dismiss it: supply is the slowest-acting of the major yield inputs, and it moves the curve at auction rather than at announcement. The August 11–13 sequence is therefore the event. Three auctions across the curve inside three days arrive into a front end that spent this week repricing the September meeting by nearly thirty percentage points, and the 30-year leg on August 13 is where any indigestion appears first. The offsetting bid is real: if the labour softening in stories 1 and 3 persists, duration demand rises for growth reasons and comfortably absorbs the incremental supply. The risk case is the opposite pairing — a hot CPI on Wednesday that revives the hike trade, immediately followed by the quarter’s largest supply event.

What to watch:The August 13 30-year auction tail and bid-to-cover, the cleanest single read on whether $739 billion of quarterly supply is being absorbed comfortably. A 10-year back-up above 4.70% before the August 11 auction would signal the market is demanding concession.

5. Consumer Credit Beats by a Third as Revolving Balances Accelerate — in the Same Week Wage Growth Hit a Five-Year Low (Federal Reserve G.19 / NY Fed, Friday Aug 7)

What they’re saying:Total consumer credit rose $14.17 billion in June against a $10.5 billion consensus, reversing May’s $1.08 billion decline. The Fed’s G.19 showed overall credit expanding at a 3.3% annual rate, with revolving credit-card balances growing at a 6.0% annual pace against 2.3% for nonrevolving. Separately, the New York Fed’s Survey of Consumer Expectations showed one-year-ahead inflation expectations easing to 3.6% in July from June’s 3.7%, which had been the highest reading since September 2023.

The context:Revolving credit growing at nearly three times the nonrevolving rate reads either as confident households or as households substituting credit for income, and the same week supplied the reason to worry about the second: average hourly earnings growth slowed to 3.2%, its weakest since May 2021. Credit accelerating while wage growth decelerates is the combination that precedes delinquency cycles, and it is worth flagging precisely because the headline print looks like strength. The easing in inflation expectations cuts the other way and matters for the Fed — it interrupts two straight months of rising short-term expectations and removes an anchoring complication from the labour-market debate. But the survey was fielded before Friday’s payroll print, so it carries no information about how households read a contracting labour market.

What to watch:Tuesday’s Q2 Total Household Debt release from the New York Fed, which carries the delinquency detail this credit build makes material. Watch whether revolving growth persists alongside slowing wage gains in the July G.19.

1. Eli Lilly (LLY): +4.86% on the day, close $1,191.94 by Friday | The Season’s Largest Guidance Raise, Funded Entirely by Volume

The Numbers:Released BMO Wednesday, August 5. Revenue $22.97 billion, up 48% year over year and 11.07% above the $20.69 billion consensus. Non-GAAP EPS $8.38 against $6.17 expected, a 35.86% beat and up 33%; GAAP EPS $7.94 against $7.33, up 26%. Mounjaro and Zepbound delivered a combined $14.9 billion — worldwide Mounjaro revenue up 91% to $9.9 billion with US revenue up 45% to $4.8 billion, and US Zepbound revenue up 44% to $4.9 billion. Full-year 2026 revenue guidance was raised to $85–$87 billion from $82–$85 billion; adjusted EPS guidance was narrowed to $35.50–$36.50 from $35.50–$37.00. Market capitalisation roughly $1.10 trillion.

The Problem/Win:A $2.28 billion revenue beat and a $3 billion lift to the bottom of the full-year range is the largest single upward revision of the reporting season, and the composition is what makes it credible rather than the size. Management explicitly flagged that Zepbound’s 44% US gain came despite lower realised prices, which means the franchise is absorbing price erosion and compounding anyway — a market-share result, not a pricing one. The qualification sits in the same release and is easy to miss: the EPS range was narrowed by trimming the top end from $37.00 to $36.50 even as revenue guidance rose $3 billion. That gap quantifies the margin cost of the volume, and it is the number to track rather than the headline.

The Ripple:As a trillion-dollar company trading above $1,100 a share, Lilly supplied the single largest contribution to Wednesday’s Dow record of 54,349.00 on a session when the S&P, Nasdaq 100 and Russell all fell — a price-weighted index effect worth understanding before reading that record as breadth. Healthcare gained 1.19% against a falling index. The competitive read-through arrived the day before from two directions at once: Pfizer’s berobenatide Phase 2b data suggesting weight loss comparable to tirzepatide, and Amgen’s MariTide progressing through Phase 3 — two credible monthly-dosing challengers to Lilly’s weekly incumbency surfacing within hours of each other. Lilly’s strength also came as Novo Nordisk fell on its own obesity-franchise concerns.

What It Means:The incretin franchise is still accelerating three years into launch while conceding price, which is the profile of a business taking share rather than harvesting it. Lilly is now large enough that its surprises move the index, and the narrowed EPS range is the market’s reminder that this growth is not free.

What to watch:Realised price per Zepbound prescription in Q3 — continued erosion at 44% volume growth is tolerable, but the trade-off breaks if volume decelerates first. Watch the widening gap between the raised revenue range and the narrowed EPS range, and the timing of the MariTide Phase 3 readout against Pfizer’s berobenatide programme.

2. Advanced Micro Devices (AMD): −7.04% on the day | Every Line Beat Except the Two That Set the Multiple

The Numbers:Released AMC Tuesday, August 4. Record revenue $11.5 billion, up 50% year over year and 13% sequentially, against $11.28–$11.31 billion consensus. Non-GAAP diluted EPS $1.66 against $1.62 expected, up 246% year over year. Data Center revenue $6.7 billion, up 107% and now 58% of the company, on fifth-generation EPYC processors and Instinct MI350 GPUs. Client and Gaming $3.8 billion, with Client up 23% and Gaming down 31%; Embedded $977 million, up 19%. Q3 guided to approximately $13.0 billion ±$300 million against $12.5 billion consensus, with non-GAAP gross margin guided near 56%. Capital expenditure was $808 million against roughly $298 million modelled, and free cash flow fell to $1.56–$1.6 billion from $2.57 billion in Q1. CEO Lisa Su said Data Center revenue should more than double again in 2027.

The Problem/Win:Two lines did all the damage and neither is on the income statement. Capital expenditure at $808 million against a $298 million model is not a modelling variance — it is a 171% overshoot that cut free cash flow roughly 40% sequentially. The second is subtler and arguably worse: Q3 gross margin guided to approximately 56% is effectively flat against Q2, which tells investors that a 107% increase in data-centre revenue is arriving with no margin expansion at all. Growth that requires proportionally more capital and delivers no incremental margin is worth a lower multiple than growth that does not, regardless of its rate. The Gaming segment’s 31% decline is the other detail the headline obscures — the 50% consolidated growth rests entirely on one segment now representing 58% of revenue.

The Ripple:The print reached companies with no reporting event of their own, which is what separates a re-rating from a bad quarter: Lam Research fell 3.25%, Palantir 2.61%, and Applied Materials and Texas Instruments 2–3%, all on Wednesday, none on a company-specific disclosure. NVIDIA diverged upward at +3.43%. Arista — reporting within the same hour with 37.7% revenue growth and operating margin expanding to 49.9% — rose, and the pairing supplied the week’s cleanest natural experiment on where in the AI stack margin can be defended. Sandisk and Western Digital then delivered the same lesson from storage two days later, both beating and both falling double digits.

What It Means:Demand is settled and no longer the variable. What this print repriced is the conversion rate — what share of AI revenue a merchant supplier keeps after paying for the capacity to deliver it. On this quarter’s evidence the answer is less than the market had assumed, and the 2027 doubling Su guided to implies the capex line does not normalise soon.

What to watch:Whether Q3 gross margin prints above the 56% guide — expansion would retire the thesis that AMD is buying its growth, and is the single number that resolves this. Watch second-half capital-expenditure guidance for whether the $808 million run rate repeats, which would make the free-cash-flow compression structural, and Applied Materials on August 13 for whether the equipment orders corroborate the spending.

3. Palantir Technologies (PLTR): +29.45% on the day, +39.78% on the week | The Commercial Book Is Now Within $45 Million of the Government Book

The Numbers:Released AMC Monday, August 3. Revenue $1.935 billion against roughly $1.81 billion consensus, up 93% year over year and 19% sequentially. Adjusted EPS $0.41 against $0.34 expected; GAAP net income $1.062 billion, or $0.41 per share, against roughly $329 million and $0.13 a year earlier. Segment detail: US commercial $764 million, up 149% year over year and 28% sequentially; US government $809 million, up 90%; international government $181 million, up 42%; international commercial $182 million, up 26%. Adjusted gross margin 86%; adjusted operating income $1.194 billion for a 62% adjusted operating margin; GAAP net margin 55%. FY26 revenue guidance raised to $8.15–$8.158 billion from $7.65–$7.66 billion, US commercial guidance to above $3.424 billion implying at least 134% growth, adjusted operating income to $4.889–$4.897 billion and adjusted free cash flow to $4.5–$4.7 billion.

The Problem/Win:The segment arithmetic is the disclosure that matters and it is easy to lose behind the headline growth rate. Total commercial revenue of $945 million now sits within $45 million of total government revenue of $990 million — roughly one quarter from a balanced mix, and Palantir got there by growing the commercial side rather than shrinking the government side, which itself grew 90%. That retires the single most persistent structural objection to the name. The margin profile is the second half of it: a 62% adjusted operating margin on 93% revenue growth inverts the usual trade, since most software businesses buy that growth rate with margin rather than expanding it. Raising full-year revenue guidance by roughly $500 million at the half-year implies the second half is tracking materially above the April plan.

The Ripple:Palantir was the largest single contributor to the Nasdaq 100’s 3.32% Tuesday advance and the week’s biggest weekly mover at +39.78%. It drew same-session upgrades to Buy from Citi, which raised its target to $235 from $210, and from Deutsche Bank, then added 10.32% on Friday on a positive Bank of America view. It also supplied the fundamental justification for the enterprise-software leg of the week — the applied-AI layer monetising what the infrastructure layer is funding — alongside Shopify at +17.02% and Cloudflare at +13%. Notably it did not escape Wednesday’s capex re-rating, falling 2.61% on no company-specific news.

What It Means:This is the clearest audited evidence yet that AI spending is converting into application-layer revenue at scale, and management guided as though the curve steepens rather than flattens. The unresolved question is entirely valuation and comparison base: 149% commercial growth cannot be annualised indefinitely, and a 39.78% week prices a great deal of the next several quarters.

What to watch:Whether US commercial revenue holds above 100% year-over-year growth next quarter as the comparison base steepens — that single number is what the raised $3.424 billion full-year guidance depends on. Watch US commercial customer count and net dollar retention in the filing for evidence the 149% is broadening rather than concentrating in a handful of large accounts.

Q2 2026 earnings season is 88% complete, and Friday’s calendar was empty of mega-caps — the largest reporter was Vistra Corp at $47.4 billion. Next week reopens with the index’s single largest company and then turns to the AI infrastructure chain.

Berkshire Hathaway (BRK.B) — BMO, Monday, August 10 — the week’s only mega-cap on Monday’s calendar at a $1.01 trillion market capitalisation, with consensus of $5.04 per share on $96.52 billion of revenue; broader estimates look for a roughly 12.9% year-over-year earnings decline on 4.3% revenue growth. Key focus: what Greg Abel does with a record $397.4 billion cash and short-term investment position in his second full quarter as chief executive — buybacks versus acquisitions is the question the market is actually asking — plus insurance underwriting income and float in a softening market. Results are filed over the weekend, so Monday’s session is the first opportunity to price them.

Cisco Systems (CSCO) — AMC, Wednesday, August 12 — expectations are for mid-teens revenue growth on AI networking orders, with gross margin guided to contract roughly two percentage points year over year. Key focus: whether AI order growth is large enough to offset the margin compression it costs, and the read-through to Cloudflare’s agent-traffic thesis.

Applied Materials (AMAT) — AMC, Thursday, August 13 — consensus of $3.39 per share, up 36.7% year over year, on revenue of $9.01 billion, up 23.4%. Key focus: the cleanest available read on whether semiconductor capital-equipment orders match the spending the hyperscalers have guided to, which is the central unresolved question in the AI capital-intensity debate.

Deere & Co (DE) — BMO, later in the week — calendar sources conflict between Thursday, August 13 and Friday, August 14, so the day should be confirmed before positioning. Key focus: agricultural capital-goods demand and dealer inventory, the cleanest read available on farm-sector credit conditions.

With 88% of the index reported and blended growth running at +50.4%, next week is the season’s last cluster of consequence; the remaining calendar thins materially thereafter.

1. Does Wednesday’s CPI restore the hawkish case a single payroll print appeared to bury? Five Fed officials spent four sessions arguing for a September hike and the market cut the odds every day anyway. Core at or below Paulson’s 2.4%–2.8% estimate vindicates the market; a hot services core hands the argument back to a committee that never conceded it.

2. Can a Hormuz reopening survive the Iranian parliament — and what is Energy worth if it cannot? Crude fell 8.8% on the week without a single barrel moving through the strait, and Friday it fell again on news pointing to delay. The sector is down 3.66% on the week while still leading the index at +28.61% year to date. The parliamentary vote is now the gate, and the June memorandum that already collapsed did so on precisely this routing question.

3. Does the AI capital-intensity verdict survive Applied Materials on Thursday? Four companies beat and raised this week and were sold on the cash-flow statement. Wafer-fabrication equipment orders either corroborate the roughly $725 billion of guided 2026 hyperscaler capex or they do not — and unlike any single vendor’s print, this one speaks for the whole chain. Cisco on Wednesday carries the networking read.

4. Does $125 billion of supply find buyers if CPI runs hot first? The 3-, 10- and 30-year auctions on August 11–13 straddle Wednesday’s inflation print. A benign CPI leaves duration well bid on the growth read; a hot one puts the largest supply event of the quarter into a front end trying to reprice a hike. The August 13 30-year tail is the number that matters.

5. Do two Fed governance stories converge? Chair Warsh has floated cutting FOMC meetings from eight to six as part of a broader retreat from forward guidance, while the White House has started a 21-day clock on removing Governor Lisa Cook that expires just before the September meeting. Fewer scheduled decisions concentrate volatility into each remaining one; a contested seat concentrates it further. Watch the 30-year and the dollar, not the front end.

WHY THIS CHARTFour of the week’s five charts examined one market each — US versus world valuations, manufacturing’s hiring turn, AI export concentration, the collapse in frontier token pricing. Friday’s examines the assumption underneath the week itself: the index closed at a record because a negative payroll print priced a September hike out of the front end, and this chart is the argument that the front end is the only place that relief ever lands.

ORIGINAL CHART ANALYSIS — FROM FRIDAY’S MIBThe Chair said the right words, and the long bond sold off. On 29 July, Kevin Warsh told markets there is no soft inflation target, only 2%. The 30-year yield rose 12bp during those remarks. The 2-year fell 4bp. The front end took him at his word. The back end did not. The disbelief has a price. The 30-year TIPS yield now pays 2.96% real, the highest real return the long bond has offered since the crisis. That is compensation for doubt rather than for policy. Investors want three points above inflation for thirty years. The deficit runs at 5.8% of GDP, and the supply itself rebuilds the premium. That real yield is also the rate every distant equity cash flow is discounted at. For three decades that yield fell, 875 basis points in all, and a falling discount rate did much of the work. The engine now runs in reverse and the index is at a record anyway. Earnings growth is outrunning the hurdle, not escaping it. A policy rate mean-reverts. A term premium need not. Easing would leave the back end untouched, or lift it — so the 30-year mortgage stops waiting on the funds rate. Watch the dissenters: if the Fed actually hikes, the long bond should rally. A weekly close back under 5.2% would make four years of failure the story again. The front end still answers to the Fed. The back end has stopped asking. What it means: a Fed cut will not rescue housing or rate-sensitive equities, because their discount rate is set at the back end. Take policy relief in short duration, and reprice long-dated assets at 3% real.