Closing Briefing — Friday, September 4, 2026

Index levels from the Investing.com major-indices board read after the 16:00 ET close; Bloomberg's board marks the S&P 500 at 7,718.60, a 0.24-point difference discussed in Data Notes. Treasury yields are the official Daily Treasury Par Yield Curve for 4 September.

1. The payroll beat, and the fact that it changed the meeting without changing the destination. +162,000 against +55,000 took the September hike from 49.4% to 58.6% on CME's own columns, a 9.2-point move worth roughly 0.75 basis points of ZQU6 price — the contract went 96.3125 to 96.3050. That leverage, roughly ten points of probability per basis point of price for a mid-month meeting, is why a small price move looks like a large narrative move. The tell is where the rest of the strip went: the September contract cheapened 1.0 bp while December 2027 cheapened 4.5 bp, with the move rising monotonically along the curve. That is the mirror image of Thursday, when the back richened 4 to 6 bp against a pinned front. January 2027's modal range flipped back up to 4.00-4.25% and cumulative-above at December 2027 rose to 93.6% from 91.2%, so the market added to the destination rather than merely pulling hikes forward. Tradable through the front-versus-back ZQ spread in Section 12.
2. The memory bid that arrived without a headline. SanDisk +11.90% to $1,740.00, KLA +7.32%, Seagate +6.34%, Micron +6.17% to $1,017.30, Western Digital +5.86%, Teradyne +5.49%, Lam Research +5.12%, AMD +4.69%, Intel +4.51%, Applied Materials +4.31%. The vendor coverage is explicit that SanDisk released nothing: the move is sector flow against a NAND and DRAM pricing cycle whose last hard evidence — an 84.6% gross margin on $8.96bn of SanDisk revenue, and Western Digital storage prices up in the high teens year on year — still points up. SOX +3.38% against a Nasdaq 100 +0.21% is a 3.17-point relative move on a session the broad index fell.
3. Nvidia's non-participation. NVDA +0.83% on a day its own index rose 3.38%, and Broadcom +0.21%. The two largest members of the semiconductor complex contributed almost nothing to its best session of the window. That is the third consecutive session in which the complex and its megacaps have moved on different logic — Wednesday sold Broadcom's guidance, Thursday refused a 1.16% index rally, Friday ran without either name. A complex that keeps re-ranking its own leadership is a dispersion trade, not a beta trade.
4. Fair Isaac, and a regulatory re-rating done in one sentence. FICO fell 16.68% — Bloomberg marks the intraday low at roughly -21% — after FHFA Director Bill Pulte instructed Fannie Mae and Freddie Mac to accept VantageScore 4.0 from every lender, effective immediately, ending a pilot that had been limited to 50 originators since 1 May and that had already reached over 9% of GSE-securitised mortgages. "FICO has enjoyed a monopoly. No more." The read-across was indiscriminate: Equifax -6.37%, TransUnion -5.9%, Experian -4.6% in London. Pulte also said the agency is "seriously considering bi-merge," which would cut the bureaus from three pulls to two. This is a cash-flow event, not a sentiment event, and the second leg has not been priced.
5. Lululemon, and the second cut in two quarters. LULU -17.39% to $100.60, its worst session since September 2025 and the worst S&P 500 performer. Second-quarter revenue $2.4bn, down 4%, missed a $2.46bn estimate even as adjusted EPS of $2.92 beat $1.82 — a beat manufactured by 560 basis points of tariff refunds inside a gross margin up 200 bp to 60.5%, against an operating margin down 190 bp to 18.8%. Comparable sales -9%, with the Americas -12%. Full-year revenue guidance cut to $10.35-10.5bn from flat-to-down-1%, and EPS to $9.48-9.73 from $10.95-11.15. Stifel: "We are disappointed to see another sharp downwards revision." The third-quarter EPS guide of $0.93-0.98 against a $2.41 consensus is the number that re-rates the multiple.
6. The curve's shape, which contradicts the equity story. Yields rose where the Fed is priced and fell where duration is priced: the 3-year +4 bp, 2-year +3 bp, 5-year and 7-year +2 bp, 10-year +1 bp, 20-year unchanged, 30-year -1 bp. 2s30s flattened 4 bp to +87 bp and 20s30s inverted to -1 bp after seven consecutive sessions at exactly zero. A three-fold payroll beat that adds nine points of hike probability and simultaneously richens the 30-year is a market saying the hikes will work. Section 6 has the arithmetic.
7. The Asian currency block reversed its own anomaly. USD/KRW -0.84% to 1345.21, USD/TWD -0.29% to 31.6370, USD/CNY -0.15% to 6.70727 — three Asian currencies stronger on a session DXY rose 0.25%. For four sessions this report documented the opposite: a won that would not strengthen when the dollar fell. The explanation that fits both halves is flow timing, not direction — the region trades the American session's conclusion in its own next session, and Friday's Asian tape had already rallied hard (Kospi +1.64%, Taiwan +1.51%, Hang Seng +1.74%) before New York opened.
8. Gold's failed session, in the cleanest form available. Comex December gold opened $4,520.30 against a $4,539.90 settle, printed a high of $4,537.34 and never traded through the prior close, then fell to $4,412.55 and finished -1.38% at $4,477.20. A metal that cannot trade above yesterday on any tick of the session is not consolidating; it is being distributed. The mechanism is direct: a hawkish payroll raises the nominal leg of the real rate.
9. Breadth versus the index, again. 166 advancers against 325 decliners among 491 S&P 500 names captured — a 1-to-1.96 ratio — while the index fell only 0.38%. Seven of eleven Finviz groups fell. The index was held up by a semiconductor complex that is roughly a tenth of it, which is the same concentration risk the report flagged in the other direction on Thursday.
10. Oil's week, which the day disguises. WTI settled $91.22, down eight cents, and Brent $95.85, up 33 cents — but the week ran +9.38% and +8.80%, the strongest since mid-July, on renewed U.S.-Iran fighting and an Iranian expansion of its Hormuz vessel blacklist. Reuters reports U.S. diesel at record highs, with Ukrainian strikes on Russian refineries compounding Middle East risk, while Iraq lifted August exports to 2.34m barrels a day from 1.35m. ANZ raised its short-term Brent forecast to $95, noting that "elevated inventories helped absorb the initial supply crisis" — the buffer argument, now thinner.

Source: Finviz group screener, Performance table view (g=sector&v=140&o=name), read after the close. 1-Day is the Change % column, 1-Week Perf Week, YTD Perf YTD. Finviz classification, not GICS.

YTD reconciliation, and it is the cleanest of the reporting window. Compounding each group's 3 September YTD by Friday's one-day move reproduces the published YTD to within 0.03 percentage points for ten of eleven groups. Worked examples: technology 1.2493 × 1.0077 = 1.25892, or +25.89% against a published +25.89%, deviation 0.000 pp; energy 1.4002 × 0.9924 = 1.38956 → +38.96% against +38.96%, deviation 0.000; consumer cyclical 0.9641 × 0.9894 = 0.95388 → -4.61% against -4.60%, deviation 0.01 pp. The single wider case is industrials at 0.06 pp — 1.1095 × 1.0038 implies +11.37% against a published +11.43% — and real estate, which carried the widest deviation for two consecutive sessions, now reconciles to 0.01 pp. No group requires a carry-forward flag.

The rotation is the exact inverse of Thursday's and it is a rate rotation. Thursday put industrials, financials, consumer cyclicals and communication services in front on a front-end rally; Friday's payroll reversed the front end and only three of eleven groups closed green. Consumer cyclical is the worst at -1.06%, which is both the rate leg and the Lululemon leg; healthcare -0.94% and communication services -0.92% follow. Energy fell 0.76% on a session crude was flat, keeping the pattern this report has flagged all week in which the equity refuses to follow the barrel, and it still leads the year at +38.96%.

The composition traps to name, and one of them is the whole session. Technology at +0.77% massively understates the semiconductor complex: SOX rose 3.38% while the group's software and services leg was sold — ServiceNow -2.97%, Accenture -3.31%, Salesforce -1.97%, Synopsys -5.40%, Cadence -3.99%, Autodesk -8.26%, PTC -6.04%, Intuit -3.37% and Adobe -6.73%. On Thursday the same group's +1.25% overstated semis, because software led and SOX added 0.11%. Two consecutive sessions in which the Finviz technology print points the wrong way about chips is a reason to read SOX directly rather than the group. Financial at -0.63% conceals the credit-bureau wreck: Fair Isaac -16.68%, Equifax -6.37% and TransUnion -5.9% sit inside a group whose banks were nearly unchanged — Wells Fargo +0.87%, Goldman +0.07%, Bank of America -0.06%, Morgan Stanley +0.26%. Consumer defensive at -0.90% is broad rather than food-specific this time, with the nine packaged-food names of Section 12 item 7 down an average of 1.47%, so the sub-industry underperformed its own group by only 0.57 pp against Thursday's 3.77.

Higher — the memory and equipment complex, and almost nothing else.

• SanDisk (SNDK) +11.90% to $1,740.00 — best performer in both the S&P 500 and the Nasdaq 100. No company announcement; vendor coverage attributes the move to sector flow on NAND pricing. The last hard datapoint remains Q4 revenue of $8.96bn at an 84.6% gross margin.
• KLA (KLAC) +7.32%, Seagate (STX) +6.34%, Micron (MU) +6.17% to $1,017.30 on 32.30m shares, Western Digital (WDC) +5.86%, Teradyne (TER) +5.49%, Lam Research (LRCX) +5.12%, Applied Materials (AMAT) +4.31%, Skyworks (SWKS) +3.58%, Qorvo +2.45%, Microchip +1.45%.
• AMD +4.69% to $477.57, Intel (INTC) +4.51%, Super Micro (SMCI) +4.54%, Corning (GLW) +5.68%, Texas Instruments +1.78%, Analog Devices +1.62%, NXP +1.26%, ON Semiconductor +0.99%, Monolithic Power +0.84%.
• Nvidia (NVDA) +0.83% on 116.07m shares and Broadcom (AVGO) +0.21% — the two largest chip names contributed almost none of a 3.38% index move.
• NRG +6.42%, Constellation Energy (CEG) +4.88%, Vistra (VST) +3.52% — the power complex bid alongside AI hardware for a third session this week.
• Oracle (ORCL) +3.08% into a results date that has just moved (Section 5). Eaton +3.46%, Cummins +2.80%, Caterpillar +1.72%, JB Hunt +2.81%, Delta Air Lines +1.78%, United Airlines +2.5%, Aptiv +3.04%, BorgWarner +2.75%, Builders FirstSource +2.51%, Best Buy +3.18%, Estee Lauder +2.97%, Franklin Resources +3.52%, Home Depot +0.94%, Lowe's +1.26%, Dell +1.50%, Arista +1.22%.

Lower — regulation, guidance and duration.

• Lululemon (LULU) -17.39% to $100.60 — full detail in Section 2 item 5. Worst S&P 500 performer and worst session since September 2025.
• Fair Isaac (FICO) -16.68%, Equifax (EFX) -6.37%, TransUnion (TRU) -5.9% — the FHFA VantageScore directive. Experian fell 4.6% in London; TransUnion is the read-across name with the least mortgage-mix disclosure.
• Autodesk (ADSK) -8.26%, Adobe (ADBE) -6.73%, PTC -6.04%, Synopsys (SNPS) -5.40%, Cadence (CDNS) -3.99%, Tyler Technologies -4.02%, Paycom -3.68%, Intuit -3.37%, Accenture -3.31%, ServiceNow -2.97%, Salesforce -1.97%. Adobe's decline follows the announcement that Anil Chakravarthy becomes chief executive on 1 December — and the company now reports on 10 September, one day later than the calendar carried on Thursday.
• Tesla (TSLA) -5.96% to $353.94 on 59.33m shares, the weakest of the megacaps, after a Cybercab launch Bloomberg describes as having underwhelmed investors.
• Netflix (NFLX) -5.35%, DoorDash -4.63%, Gartner -4.62%, Palantir (PLTR) -4.50%, Hewlett Packard Enterprise -4.48%, Axon -4.15%, Roper -3.57%, FactSet -3.51%, T-Mobile -3.46%, Fox B -3.39% and Fox A -3.38%, Cognizant -3.60%.
• Apple (AAPL) -2.51% to $319.97 and Microsoft (MSFT) -2.03% — the two largest index weights both fell more than the index, which is how a 0.38% decline sits alongside a 3.38% semiconductor rally.
• Albemarle -4.45%, Robinhood -2.09%, Moderna -2.23%, Vertex -2.12%, Bristol-Myers -1.87%, Regeneron -1.87%, Newmont -1.79%, Diamondback -1.83%, Chevron -1.29%, ConocoPhillips -1.08%, Occidental -0.94%. Freeport-McMoRan +0.23% on a flat copper session.
• Samsara (IOT) +3.7% is the day's counter-example on the software side, after raising full-year revenue and adjusted EPS guidance. Not an S&P 500 constituent.

Analyst actions, 4 September.

• Upgrades. Morgan Stanley raised Shell (SHEL) to Overweight from Equal Weight, price target $101.30 from $81.60 — a 24% target increase and the largest of the day. Bernstein took Ormat (ORA) to Market Perform from Underperform, target $112, and W.P. Carey (WPC) to Equal Weight from Underweight, target $80. Piper Sandler moved Simmons First National (SFNC) to Overweight, target $27.50 from $26. Baird raised Sonida Senior Living (SNDA) to Outperform, target $43 from $37.
• Downgrades. Craig-Hallum cut Ambarella (AMBA) to Hold from Buy with the target down to $70 from $95, a 26% reduction on a day the semiconductor index rose 3.38% — the sharpest disagreement with the tape on the board. Morgan Stanley cut TotalEnergies (TTE) to Equal Weight, target $90, on the same morning it upgraded Shell, which is a pair recommendation rather than a sector call. UBS cut Hannon Armstrong (HASI) to Neutral, target $44 from $51. Canaccord cut UiPath (PATH) to Hold, target $17. StoneX cut Cal-Maine (CALM) to Hold with no published target.
• Initiations. StoneX started PulteGroup (PHM) at Buy, target $148, and Lennar (LEN) at Hold with no target — a split call inside homebuilding on the day the 3-year cheapened 4 bp. Stephens started Paychex (PAYX) at Equal Weight, target $132, and Asure Software (ASUR) at Overweight, target $11. Canaccord started Belite Bio (BLTE) at Buy, target $303.
• Percentage upside is not asserted for these calls because same-session closing prices for the non-S&P-500 names were not independently captured; the target changes are stated as published. See Data Notes.

Sourcing, disclosed. The Earnings Whispers day pages remain behind a cookie-and-usage-agreement consent banner, which this unattended session did not accept. The rosters below are captured from the Nasdaq earnings calendar API for each date and screened name by name against an S&P 500 constituent list. Nasdaq publishes a before-open or after-close bucket rather than a clock time, so no clock times are asserted this session; confirm every time against company investor relations before trading a date.

None. The week is finished; its last S&P 500 reporters were Thursday's, and the reactions are in Section 2 and Section 4. The next reporter is Wednesday of next week.

Mon 9/7. U.S. equity markets are closed for Labor Day. The capture returns no S&P 500 reporter for the date.

Tue 9/8. No S&P 500 reporter on either bucket. This is a change from the prior calendar and it is the most consequential line in this section — see the diff below.

Wed 9/9. AMC: Cooper Companies (COO).

Thu 9/10. AMC: Oracle (ORCL), Adobe (ADBE), Copart (CPRT).

Fri 9/11. BMO: Kroger (KR).

Changes vs. the prior calendar (9/3 report):

• Oracle (ORCL) has moved from Tuesday 8 September to Thursday 10 September, and it has acquired a bucket. The prior four consecutive captures placed Oracle on 9/8 with no before-open or after-close designation published. This capture places it on 9/10 in the after-close bucket. A date change accompanied by the appearance of a previously missing bucket is the signature of a publisher correcting a placeholder rather than of a company re-scheduling, but the direction of the correction cannot be inferred from one capture. Treat 10 September after the close as the working date and confirm with Oracle investor relations; this report will hold the 9/8 slot as vacated only after a second agreeing capture, per the two-capture rule.
• The consequence is a three-name Thursday. Oracle, Adobe and Copart now report into the same after-close window on 10 September, where the prior calendar spread them across two sessions. Three after-close prints on one evening compresses the reaction into a single overnight session.
• Adobe (ADBE) and Copart (CPRT) are unchanged on 9/10 in the after-close bucket for a third consecutive capture. Cooper Companies (COO) is unchanged on 9/9 after the close and Kroger (KR) unchanged on 9/11 before the open.
• No additions and no removals beyond the Oracle re-dating.
• Dual listings deduped; no dual-listed constituent appears on the covered dates this capture.
• Non-members on the same dates, listed so nobody mistakes their absence for an omission: Grifols (GRFS), Apartment Investment (AIV) and Dynagas LNG (DLNG) on 9/7; Casey's (CASY), ServiceTitan (TTAN), GameStop (GME), Braze (BRZE), ABM, United Natural Foods (UNFI), Innovage (INNV) and Mission Produce (AVO) on 9/8; Chewy (CHWY), SailPoint (SAIL), Core & Main (CNM), AeroVironment (AVAV), Korn Ferry (KFY), Signet (SIG), American Eagle (AEO), Academy Sports (ASO) and Caleres (CAL) on 9/9; Descartes (DSGX), Macy's (M), RH, National Beverage (FIZZ), Hub Group (HUBG), Zumiez (ZUMZ), Designer Brands (DBI), Lovesac (LOVE) and 1-800-Flowers (FLWS) on 9/10; Hooker Furnishings (HOFT), Rent the Runway (RENT) and MoneyHero (MNY) on 9/11. Borderline membership cases are listed in Data Notes and conservatively excluded.
• What the forward calendar hands the desk. A blank Monday and a blank Tuesday, then a single medical-device name on Wednesday, then everything at once. Thursday 10 September after the close now carries the month's most important artificial-intelligence capital-expenditure disclosure, a software franchise five days into a new chief executive, and an auction operator eight days past a JPMorgan upgrade that lifted its target 25% — three prints into one overnight window, with Kroger the following morning as the only consumer read of the week. The reaction function to carry is the one Section 4 has documented twice this week in opposite directions: guidance was rejected on Wednesday evening and rewarded on Thursday morning, with the same names moving both ways inside eighteen hours, and then a third guidance cut was punished 17.39% on Friday. Size the reaction, not the number. Note also that the compression removes the Tuesday hedge the desk had penciled in: whatever position expresses a view on the AI capital-expenditure cycle now has to carry an extra 48 hours of unrelated tape before it gets its answer.

Source: U.S. Department of the Treasury Daily Treasury Par Yield Curve Rates, Text View for September 2026, read after publication on the evening of 4 September. Changes are versus the 3 September official row (1-day) and the 28 August official row (1-week).

The read: a belly-led bear flattener, and the diagnostic is that the market priced tighter policy without pricing more term premium. The payroll beat cheapened exactly the tenors where the September and October meetings live — the 3-year 4 bp and the 2-year 3 bp — while the 10-year rose only 1 bp, the 20-year not at all and the 30-year richened 1 bp. That is a 4-to-minus-1 spread between the cheapest and the only richer point on the curve, and it is the opposite configuration to the one that dominated the second half of August, when the long end led every selloff. Three-fold beats on payrolls do not usually produce a lower 30-year yield. This one did, because the mechanism the market chose was "the Fed will act, and acting works": 2s30s flattened 4 bp to +87 bp and 2s10s flattened 2 bp to +41 bp, both against a week in which they were 2 bp and 1 bp the other way. On a one-week view the whole curve is still cheaper — the 5-year and 7-year 6 bp, the 10-year 5 bp, the 30-year only 2 bp — so Friday extended the week's flattening rather than reversing it.

The 20s30s inversion is the single new fact on this page. For seven consecutive sessions the spread printed exactly 0 bp with both tenors at 5.25%. On Friday the 30-year richened a basis point and the 20-year did not, taking the spread to -1 bp — the first non-zero print of the run and the first inversion. One basis point is not a dislocation, but the run of zeros was itself the evidence this report used for a buyback floor at the very long end, and the floor has now moved. It matters for the position in Section 12 item 3, whose invalidation sits at -3 bp: the trade is 1 bp worse and 2 bp from being wrong.

Vendor cross-checks reconcile, which is worth stating because the official row published late. WSJ's bond board at the 5:00 p.m. ET close marks the 10-year at 4.789% (+1.6 bp), the 30-year at 5.247% (-0.4 bp), the 2-year at 4.372% (+2.1 bp), the 3-year at 4.453% (+3.7 bp), the 5-year at 4.551% (+2.7 bp) and the 7-year at 4.659% (+2.0 bp). Every one of those agrees with the official par row to within a basis point and every direction agrees exactly, including the 30-year's decline. Bloomberg's board marks the 10-year at 4.78. There is no level dispute this session; the differences are the ninety minutes between Treasury's 3:30 p.m. bid-side snapshot and the 5:00 p.m. quotes.

The bill curve is doing something the coupon curve is not, and it is a financing signal. The 1-month fell 4 bp to 3.79% — the largest decline anywhere on the fourteen-point curve, on the day the market added nine points of hike probability — while the 3-month rose 2 bp to 3.91%. That takes the 3-month-minus-1-month spread to 12 bp from 6 bp, a doubling in one session, and it is the correct sign for a meeting being priced: a 3-month bill spans the September and October meetings, a 1-month bill spans neither in full. But the level move in the 1-month is too large for that alone, and it points to bill supply and settlement rather than policy. Off the published table, the 1.5-month rose 1 bp to 3.83%, the 2-month fell 1 bp to 3.90% — so the 1.5-month-to-2-month gap this report has tracked narrowed to 7 bp from 9 bp — the 4-month rose 1 bp to 4.00% and the 6-month rose 3 bp to 3.98%, closing the four-month-above-six-month anomaly to 2 bp from 4 bp. All four are off-table under the trimmed tenor set and are cited here because they carry the story; the funding read sits with the money-market data in Section 9 block b.

Source: Federal Reserve Bank of New York Economic Indicators Calendar for September 2026 (all times Eastern), cross-checked against the Wall Street Journal and Investing.com economic calendars. Consensus figures are carried where independently verified; where none is verified, the sensitivity note describes what the market is positioned for instead of asserting an expectation.

Current week (Aug 31 - Sep 4) — still to come

Nothing remains. The week's calendar is exhausted; Friday's 08:30 release was its last item of any sensitivity, and the 10:00 and 12:45 items that followed it are Low.

Next week (Sep 7 - Sep 11)

The week after, for planning: Empire State on 15 September; retail sales, imports and exports and business inventories on 16 September, the morning the FOMC begins; claims, housing starts and the Philadelphia Fed survey on 17 September; industrial production on 18 September.

Current target range: 3.50%-3.75%. A three-fold payroll beat took the September meeting from a coin flip back to odds-on and, unlike Thursday, it moved the destination more than the schedule.

CME FedWatch headline — 16 September 2026 meeting. Data as of 4 Sep 2026, 05:01:29 p.m. CT (6:01 p.m. ET), read from the FedWatch probability table. Contract ZQU6, mid price 96.3050, prior volume 52,145, prior open interest 242,269.

Provenance of every column, stated — and Thursday's live read corrected by 0.8 of a point. The footer timestamp reads 05:01:29 CT with no meridian; the page's own "Last Updated" line reads 04 Sep 2026 05:13:01 PM CT, twelve minutes later, so it resolves as p.m. and is an indicative read taken about an hour after the 4:00 p.m. CT ZQ session close rather than a settlement snapshot. The 1 DAY column carries the legend date 3 September and prints 49.4%, against the 50.2% this report published from CME's live column on Thursday evening — a -0.8 pp correction, recorded here rather than left standing, and the second-tightest reproduction of the reporting window after Wednesday's 0.9. 1 WEEK (28 August) at 57.0% and 1 MONTH (4 August) at 58.4% carry genuine reference dates and are used in the calculations below. The Investing.com matrix underneath is timestamped 4 Sep 2026, 05:45 p.m. EDT and is the primary source for parts (a), (b) and (c).

The CME-versus-Investing.com gap, quantified, and it stayed narrow. CME puts the September hike at 58.6% at 6:01 p.m. ET; Investing.com at 58.4% at 5:45 p.m. ET — a 0.2 percentage-point difference across sixteen minutes, identical to Thursday's gap and far inside the 1.3-to-2.0 point range of earlier in the window. The arithmetic explains it: Investing.com publishes the September future at 96.303 against CME's 96.3050, two-tenths of a basis point of price. Because the 16 September meeting sits mid-month, only about 47% of the contract's averaging period is affected, so one basis point of ZQ price is worth roughly ten percentage points of hike probability. That leverage is the most important number in this section: Friday's entire 9.2-point repricing is about 0.75 bp of contract price, from 96.3125 to 96.3050.

One-day, one-week and multi-day momentum, and a symmetry worth naming. The September hike rose 9.2 pp on CME's own columns (49.4% to 58.6%) and 8.0 pp on Investing.com's own columns (50.4% to 58.4%). Investing.com's "previous day" column reads 50.4%, exactly what this report published from it on Thursday — the first zero-discrepancy reproduction of the reporting window, after 0.3, 2.3 and 2.0 point gaps in the three prior sessions. The multi-day read is the discipline. Thursday's Waller collapse took the meeting from 63.2% to 49.4%, a 13.8-point fall; Friday's payroll took it from 49.4% to 58.6%, a 9.2-point rise. Net of both, the meeting is 1.6 points more hawkish than the 1 WEEK column of 57.0% and 0.2 points above the 1 MONTH column of 58.4% — so a fortnight containing a hawkish Jackson Hole, an oil shock, a dovish governor and a three-fold payroll beat has moved the September meeting by two-tenths of a point. Further out, every 2026 horizon rose: October's cumulative-above to 69.9% from 64.1%, December's to 85.6% from 82.7%, and the peak of the 2027 strip to 96.0% at September 2027 from 94.8% on Thursday. The probability of a cut at any 2026 meeting remains 0.0%.

(a) Current-year meeting distributions

Investing.com Fed Rate Monitor, updated 4 Sep 2026 05:45 p.m. EDT. Format: current [prior day] [prior week]. Modal range in bold.

Sums are 100.0% at all three meetings. Three observations. First, September has resolved out of the coin flip in one session and it has resolved to within a point of where it sat a week ago — 58.4% against 55.9% on the vendor's own prior-week column. Second, the December contest has narrowed again and the mode is holding by a thread: one hike leads at 41.4% against 35.8% for two hikes, a 5.6-point gap where Thursday's was 9.9 and Wednesday's three-tenths. That is a second consecutive session in which the gap moved, and it moved back toward two hikes — the buckets went 32.7 to 35.8 for +50 while +25 went 42.8 to 41.4. Third, the tail was bought back, exactly reversing Thursday's sale: October's +50 bucket rose to 16.1% from 13.9% and December's +75 bucket to 8.4% from 7.2%, returning 3.4 points of tail mass to the 2026 distribution against the 7.3 points that left it on Thursday. The two-day, twelve-point unwind of the oil-shock tail this report described on Thursday has retraced roughly half of itself in a single session.

(b) Next-year meeting path

Modal range, its probability, and the cumulative probability above and below the current 3.50%-3.75% range, with the contract price that draws it.

The 2027 strip cheapened more than the front, and that inverts Thursday's message. The eight contracts print 96.015, 95.915, 95.865, 95.795, 95.775, 95.760, 95.755, 95.775 against Thursday's 96.035, 95.935, 95.885, 95.820, 95.800, 95.795, 95.790, 95.820 — 2.0 to 4.5 basis points cheaper, with the cheapening rising monotonically along the curve. Set that against the front: September 2026 cheapened 1.0 bp and December 2026 2.5 bp. Thursday was a near-uniform 4-to-6 bp richening of the back against a pinned front, which this report read as the market moving hikes later without removing them. Friday is the mirror image and it is more than a reversal of timing: January 2027's mode has flipped back up to 4.00-4.25% at 37.1% from 3.75-4.00% at 36.6%, and cumulative-above at December 2027 has risen to 93.6% from 91.2% while the first non-trivial cut probability fell to 0.5% at 3.25-3.50% from 0.9%. A back-led cheapening with a rising mode is the market adding to the terminal rate, not merely pulling hikes forward. On a two-day view the destination is now higher than it was before Waller spoke.

(c) Year-end probability ladders

Year-end 2026 — the 9 December meeting.

Cumulative above the current range: 85.6%. Cumulative below: 0.0%. Sum: 100.0%.

Year-end 2027 — the 8 December meeting.

Cumulative above the current range: 93.6%. Cumulative below: 0.5%. Sum: 99.9%.

Rounding, transparently. Every figure is reproduced at the vendor's own one-decimal precision. Column sums of 99.9% or 100.1% are rounding artefacts of that precision, not missing probability mass; no cell has been rescaled, and cells shown as 0.0% are ranges the vendor's own card either omits entirely or publishes as zero, which under CME methodology means a probability below the rounding floor. The 2027 year-end ladder sums to 99.9% for exactly this reason, as do the March, June, July, September and October 2027 rows; the 2026 year-end ladder and the three 2026 meeting rows sum to exactly 100.0%.

(a) IG and HY credit spreads

ICE BofA option-adjusted spreads via FRED. FRED publishes with a one-business-day lag: the levels below carry the 3 September 2026 as-of date, not the 4 September close. Same-day direction is cross-checked against the cash-market proxies underneath and against Bloomberg and WSJ credit coverage.

CDX — the six-step ladder was worked again in the local Chrome browser and is reported so the gap stays auditable. (1) Bloomberg in Chrome: /markets and /markets/rates-bonds both rendered fully; a full-text scan of the rates-and-bonds page returns zero occurrences of the string, and the Friday markets wrap carries equity and rate-path colour with no credit-index level. (2) WSJ Market Data bonds page in Chrome: rendered with its Treasury, consumer-rate and government-bond tables populated; no CDX quote. (3) Cbonds carries a dedicated CDX.NA.IG 5Y index page which loaded, and it masks the level behind a subscription wall — the page prints *** bps with a previous value of *** dated 02/09/2026 and names IHS Markit as the calculating organisation. S&P Dow Jones Indices and ICE publish index-family, options, tranche and methodology documentation rather than the daily running spread. (4) FT Markets Data and Reuters credit wraps returned no dated CDX quote for 4 September. (5) TradingView and Barchart symbol searches for CDX resolve to unrelated instruments, and the CME CDS index product pages carry specifications rather than levels. (6) Cash-market proxies, labelled as proxies: HYG closed $79.16, -0.06%, and LQD $105.48, -0.02% at the 4:00 p.m. ET close, against Thursday's $79.21 and $105.50 — both reconciling to the cent against the prior edition. No CDX level is published here, because an undated third-party digest number is not a CDX level.

The tail stopped widening, and that is the first thing this section has been able to say for four updates. CCC tightened 2 bp to 1,051 on the 3 September stamp after 16, 7 and 4 bp of widening in the three prior updates, so the three-session, 27 bp run is broken. HY tightened 1 bp to 265 and IG was unchanged at 81, taking the CCC-minus-HY differential to 786 bp from 787 — a single basis point of relief off a window extreme, and 163 bp wider than where 2026 opened. Against that, the one-week picture is unchanged in character: IG and HY are each 2 bp wider on the week and CCC 20 bp, so the tail is still doing four-fifths of the widening at a tenth of the index weight. The cash market, for a third consecutive session, declined to say anything at all: HYG -0.06% and LQD -0.02% on Friday against +0.13% and +0.14% on Thursday and +0.01% and +0.12% on Wednesday — a cumulative 0.08% and 0.24% across three sessions in which the lagged index moved 3 bp and then 1 bp the other way. The divergence this report flagged on Thursday has not resolved; it has gone quiet on both sides.

(b) Money-market & funding plumbing

New York Fed reference rates, published at approximately 8:00 a.m. ET for the prior business day. The rates below carry the 3 September 2026 effective date. Rate up = red.

The plumbing is tightening at the margin and the evidence is now three-sided. SOFR printed 3.66%, one basis point above the 3.65% IORB, after a session at exactly the administered rate — and its 99th percentile at 3.74% sits 9 bp above the rate itself on $2.949tn of volume, so the tail of the distribution is paying up even where the median is not. The standing repo facility took $33m, small in absolute terms and notable because it is not zero: a facility designed to be used only when private repo is inconvenient has been used. Alongside it, reverse repo take-up rose to $702m from $525m — cash going the other way on the same day — and reserve balances fell $10.4bn to $2.9249tn, their fourth consecutive weekly decline. Two-way facility usage with reserves falling is the configuration that precedes a funding squeeze rather than the squeeze itself, and the September quarter-end is three weeks away.

The off-table bill tenors belong here, because they are financing rather than policy. The 1-month bill fell 4 bp to 3.79% on a day the market added nine points of hike probability — the largest decline on the fourteen-point curve and the wrong sign for policy. It is the right sign for bill supply: a 1-month bill richening while the 3-month cheapens 2 bp to 3.91% widens the 3-month-minus-1-month spread to 12 bp from 6 bp. Off-table, the 2-month fell 1 bp to 3.90% and the 1.5-month rose 1 bp to 3.83%, narrowing that pair to 7 bp from 9 bp, while the 4-month rose 1 bp to 4.00% against the 6-month up 3 bp to 3.98%, closing the four-month-above-six-month anomaly to 2 bp from 4 bp. Two of the three bill anomalies this report has tracked narrowed and one opened. The one that opened is at the very front, which is where the standing repo facility just took its first non-zero print of the window.

(c) Rates volatility & swap spreads

The MOVE vintage is disclosed rather than smoothed, and this session it is a hard stop. The Investing.com series carries a 03/09 date stamp and reproduces Thursday's level, change and percentage exactly — 74.68, -5.03, -6.31% — which means it did not publish a 4 September value at all. On top of that it fails its own internal consistency check for an eighth consecutive session: the card's "previous close" of 95.74 sits outside its own 74.68-79.71 day range. The level is published with its vintage; the change is withheld, and the MOVE-versus-VIX ratio above is computed on a stale numerator and flagged as indicative. What can be said without the series: Friday's rate move was a 4 bp cheapening at the 3-year against a 1 bp richening at the 30-year, which is a curve event rather than a volatility event, and the option-market read of it will not be visible until the series republishes. Swap spreads at the 2-year, 10-year and 30-year were not obtainable from a primary source this session and are not asserted; the substitute evidence is the 20s30s inversion to -1 bp in Section 6, which is the same question asked of the cash curve.

(d) Issuance, leveraged loans & private credit

The September calendar is the constraint and it has not been tested yet. The month opened against an expected $215bn of IG supply — the figure this report has carried since the start of the week — into a 4.78% ten-year and a curve that has just flattened 4 bp at 2s30s. Friday was the last session before a three-day weekend, so the primary market was effectively shut; the test is the week of 7 September, which has only three usable issuance days before the 10-11 September data block and only six before the FOMC. JPMorgan's syndicate view, published on 24 August, is that the market can absorb the stampede; that view has not yet met a week in which the front end cheapened 4 bp on a payroll. The Treasury buyback operation on 9 September, described by Treasury only as "at least double" its usual size, lands in the middle of it and is the single scheduled event most likely to move the long end. On the loan side no updated Morningstar LSTA print was obtainable this session and none is asserted. The named private-credit watch item is unchanged and unresolved: Broadcom's contingent residual-value guarantees to two AI laboratories are vendor financing that no published spread series captures, and Broadcom closed +0.21% on a day its index rose 3.38%.

Levels from the TradingEconomics currency board read after the U.S. close. Every row carries a Sep/04 date stamp and a completed-session change column: the spot FX week ends at 17:00 ET on Friday, and the capture sits after that boundary with no new session open behind it, so the vendor's own %Chg column is reproduced this session rather than reconstructed — the opposite of Thursday, when a live clock forced a 24-hour reconstruction. The basis change against the prior edition is reconciled in Data Notes. Corroborated against the Investing.com majors board, whose final ticks stamp 16:58-16:59 ET. Quote basis: USD per unit for EUR, GBP, AUD and NZD; units per USD for JPY, CHF, CAD, KRW, TWD, CNY, MXN, SGD and INR.

The take: the Asian block did on Friday the exact opposite of what it did for four sessions, and the reversal is the most informative thing on this page. For four consecutive sessions this report documented a won, a Taiwan dollar and a yuan that would not strengthen when the dollar fell — USD/KRW down only 0.18% on Thursday's 0.59% dollar break, the Taiwan dollar 0.14% and the yuan 0.02%. On Friday the dollar index rose 0.25% and all three currencies strengthened anyway: the won 0.84%, the Taiwan dollar 0.29%, the yuan 0.15%. A currency that ignores a dollar decline and then rallies into a dollar advance is not trading the dollar at all. The explanation that fits both halves is timing rather than direction: Asia trades the previous American session in its own next session, and Friday's Asian tape — Kospi +1.64%, Taiwan +1.51%, Hang Seng +1.74% — was pricing Thursday's dovish repricing hours before New York reversed it. That leaves a specific and datable gap: the region has now bought a softer American rate path that the payroll destroyed at 08:30 ET, and with U.S. markets shut on Monday there is no cash market to arbitrate the correction until Tuesday.

The yen is the cleanest expression of the two-day whipsaw and it is the least symmetric. USD/JPY rose 0.29% to 156.233 after falling 1.73% on Thursday, so the pair gave back roughly a sixth of a move it made in one session, and it remains 2.39% stronger on the week. Both legs of Thursday's argument survive: the American leg reversed — the 2-year cheapened 3 bp and the 3-year 4 bp — while the Japanese leg did not, with 10-year JGBs richening a further 3.7 bp to 2.911% and the Bank of Japan decision on 18 September still carrying Nomura's extreme-case scenario of three consecutive hikes. A carry pair whose foreign leg keeps richening while its domestic leg cheapens should not be giving back yen strength on a 0.25% dollar move, and the fact that it gave back so little — 0.29% against a 1.73% gain — is the market saying the same thing the bond leg is.

The European currencies did nothing, and that is itself a data point. EUR/USD fell 0.10% and GBP/USD 0.06% on a session in which the U.S. 3-year cheapened 4 bp and 10-year Bunds richened 1.2 bp — a rate differential move of five to six basis points that produced a tenth of a percent of currency. Against that, USD/CAD rose 0.31% and USD/CHF 0.28%, so the dollar's gain was concentrated in the commodity and haven crosses rather than in the two currencies with the largest rate gap to trade. Sterling holding 1.35048 after Thursday's 10 bp gilt rally is the continuation of the observation this report made a day ago: the fiscal-premium reading has lost its cleanest evidence, and the November budget is back to being a scheduled risk rather than a daily one.

Settlement basis, stated, and reconciled to the prior edition. All rows are quoted from the Investing.com per-contract board read after the 17:00 ET Friday close, with each contract's "Prev. Close" field reconciling to the prior edition's published settle — WTI $91.30, Brent $95.52, gold $4,539.90, silver $67.70 and RBOB $3.13 all match. WTI, RBOB, heating oil and natural gas are on the October contract; gold, silver and copper on December; Brent on the November contract, front since the 23 August rollover. No front-month roll occurred this session. Weekly and year-to-date columns are TradingEconomics spot returns on the vendor's Sep/04 date stamp; the daily settles are the futures basis named in each row.

\*YTD figures marked with an asterisk are TradingEconomics spot year-to-date returns, not futures returns on the contracts quoted above. Weekly columns are TradingEconomics and Investing.com one-week changes on the same caveat. Mixing the two would be a basis error; they are presented in separate columns for exactly that reason.

Gold's session is the one to study, because the shape is more informative than the size. The contract opened at $4,520.30 against a $4,539.90 settle, printed a session high of $4,537.34 and never traded through the prior close on any tick, then fell to $4,412.55 before finishing $4,477.20, down $62.70 or 1.38%. A gap-down open that fails to fill on a 186,451-lot day is distribution, not consolidation. The mechanism is direct and it is the exact inverse of Thursday's: a three-fold payroll beat raises the nominal leg of the real rate — the 2-year cheapened 3 bp and the 3-year 4 bp — while the inflation leg was untouched, and the dollar added 0.25%. So gold moved 5.5 times the dollar in the other direction, against Thursday's 4.8 times in the same direction. Silver fell 1.31% to $66.820, less than gold in percentage terms, so the gold-silver ratio slipped to 67.00 from 67.05 and silver has now outperformed for a third consecutive session including a down one. And copper closed unchanged at $6.6700, +0.05%, refusing the precious selloff outright — which is the cleanest available confirmation that the metals move was monetary rather than industrial, because the industrial metal did not participate in either direction.

Energy broke internally again, and this time the split is products against crude. WTI settled $91.22, down eight cents, and Brent $95.85, up 33 cents — so Brent outperformed for the first time in three sessions and the Brent-WTI differential widened to $4.63 from $4.22. Underneath, RBOB rose 2.07% and heating oil fell 1.01%, the second consecutive session in which the distillate leg has lost money while gasoline gained. The week is where the story lives: WTI +9.38% and Brent +8.80%, the strongest since mid-July, on U.S.-Iran fighting that Reuters describes as the fiercest since July, an Iranian expansion of its Hormuz vessel blacklist, and Israeli Defence Minister Israel Katz's threat to "cripple" Iranian military and civilian infrastructure including energy facilities. Against the escalation sit three deflators the market is weighing: Iraq lifted August exports to 2.34m barrels a day from 1.35m in July, President Putin signalled possible Ukraine negotiations, and ANZ — while raising its short-term Brent forecast to $95 — notes that "elevated inventories helped absorb the initial supply crisis." Reuters separately reports U.S. diesel at record highs, which is the physical market saying the opposite of what the October crack just did.

The crack spreads, on a consistent October basis against October WTI at $91.22:

• Distillate crack: $4.5500 × 42 - $91.22 = $99.88, down $1.97 from $101.85.
• Gasoline crack: $3.2000 × 42 - $91.22 = $43.18, up $2.81 from $40.37.
• The differential narrowed $4.78 to $56.70 from $61.48 — a second consecutive session of narrowing, and larger than Thursday's $4.81 only in the sense that it comes on top of it, for $9.59 across two sessions.

That two-day, $9.59 collapse in the distillate-over-gasoline differential is the vindication of Thursday's decision to close the position rather than defend it, and it is also a warning about the replacement thesis nobody has put on. Retail diesel is at record highs while the October crack has fallen $9.59 in two sessions — a physical market and a paper market pointing in opposite directions inside 48 hours. Either the futures curve is telling you the record retail print is the top, or the crack is dislocating from the barrel it refines. The structural numbers still favour distillate on the spot basis — heating oil +114.30% year to date against gasoline's +86.88% — so this is a timing question rather than a thesis question, and it is worth watching rather than trading until the week of 7 September gives it a second observation. Natural gas at $2.940, up 0.89%, reclaimed Thursday's giveback and remains -20.28% on the year, the only major commodity on the board still deeply negative.

Desk-style ideas for institutional investors. Each carries an explicit expression, catalyst and invalidation. These are not personalized investment advice; verify independently and size to your own mandate before acting.

1. The rates trade — long ZQZ6 against short ZQZ7 paid on day two; hold the quarter

Mark first. Long ZQZ6 (December 2026) against short ZQZ7 (December 2027), DV01-matched one-for-one at $41.67 per basis point per contract, entered Thursday at 96.085 / 95.820 for a spread of 26.5 bp, quarter size. Friday's mark: ZQZ6 96.060, ZQZ7 95.775 — a spread of 28.5 bp. That is +2.0 bp on the day, worth +$83.33 per contract pair before costs on a quarter, and the position is +2.0 bp from entry.

Why it paid, and it is not the reason the entry note gave. The thesis was that the terminal rate is anchored and the 2026 path is not, so a back-end contract should be the stable leg. Friday inverted that: ZQZ6 cheapened 2.5 bp and ZQZ7 cheapened 4.5 bp, so the spread widened because the back moved more, not because the front held. The trade made money on the correct sign for the wrong mechanism, and that is worth writing down rather than claiming foresight — the payroll added to the destination, with January 2027's mode flipping back up to 4.00-4.25% and December 2027 cumulative-above rising to 93.6% from 91.2%.

The modal path, base case and tails. Modal path: a 25 bp hike on 16 September is now odds-on at 58.6% on CME and 58.4% on Investing.com, with ease at 0.0%; one hike is modal at October (53.8%) and at December (41.4%); the 2027 strip modes at 4.00-4.25% at all eight meetings. Base case: one hike in September, a second priced by December at 85.6% cumulative-above, and a terminal rate that has now risen rather than merely shifted in time. Tail one, and it is the whole trade: CPI on 11 September, which Waller made the deciding print and which now meets a labour market that just beat three-fold; a hot core number takes September through 70% and, more importantly for this spread, keeps the back end cheapening faster than the front. Tail two: a soft CPI that reinstates Thursday's parallel richening of the back, which is precisely the configuration that would narrow this spread. Practical implication: at roughly ten percentage points of probability per basis point of ZQ price for a mid-month meeting, the front leg has limited room left — the September contract has moved 1.3 bp down and 0.75 bp back up across two sessions — so the spread's carry now depends on the 2027 leg, and it should be sized as a terminal-rate view rather than a timing view. Catalyst: the buyback operation 9 September; PPI 9/10; CPI 9/11; the 15-16 September FOMC. Invalidation, unchanged: the spread through 22.0 bp; or the September cumulative hike back below 45% on either vendor; or the December 2026 mode reverting to two hikes. Sizing: a quarter, at $41.67 per basis point per pair.

2. Long the 20-year against the 30-year, on the November refunding — a basis point worse and two from wrong; hold the half

Expression:* long the 20-year bond against short the 30-year, DV01-neutral, half size. Mark: 20s30s at -1 bp, with the 20-year at 5.25% and the 30-year at 5.24% — 1 bp against the position on the day, and the first non-zero print after seven consecutive sessions at exactly zero. What changed, and it is the thing the trade was built on. The seven-session run of zeros was the evidence for a buyback floor holding the very long end together. On Friday the 30-year richened 1 bp on a three-fold payroll beat while the 20-year did not move, which is the floor flexing in the one direction the position cannot afford. Against it: the 9 September operation is now five days away, Treasury has described it only as "at least double" its usual size, and the $215bn September corporate calendar has not begun to price. The dislocation the trade waits for is closer, but so is the stop. Catalyst: the 9 September buyback operation and its maturity buckets; Treasury's 4 November quarterly refunding. Invalidation, unchanged: 20s30s through -3 bp, or an explicit Treasury statement ruling out changes to long-end auction sizes. Sizing: a half, unchanged. **Do not add on the loss.

3. Protection on the CCC cohort funded in IG — the first losing update after three gains; hold the half

Expression: long CCC-exposed credit protection (or short a levered-loan / CCC-heavy vehicle) against long IG cash. Mark: CCC 1,051 bp, -2 bp; HY 265 bp, -1 bp; IG 81 bp, unchanged on the 3 September FRED update, taking the CCC-minus-HY differential to 786 bp from 787 — a 1 bp loss, against gains of 13, 5 and 3 in the three prior updates for a cumulative +20 bp. The honest reading: a one-basis-point give-back is not a reversal, but it is the first time in four updates that the tail has stopped widening, and the position has no other engine. The cash proxies remain silent for a third session — HYG -0.06% and LQD -0.02% — so the pair is neither confirmed nor contradicted by anything tradeable. A position paid only by a one-day-lagged series, and now not paid by it, is a position to hold at half and to judge on the next two updates. Action: hold the half; do not add. Catalyst: the September IG calendar clearing into a $215bn month; PPI 9/10 and CPI 9/11; quarter-end funding from the 15th. Invalidation, unchanged: the differential back through 750 bp, or IG widening beyond 90 bp, which would mean the aggregate has joined the tail and the pair no longer isolates anything. Sizing: a half, unchanged.

4. Short the packaged-food cohort against long consumer defensive — paid on day one; hold the half

Mark. The nine-name basket closed Campbell's -3.37%, General Mills -2.47%, JM Smucker -1.89%, Bunge -1.17%, Lamb Weston -1.13%, McCormick -0.97%, Conagra -0.90%, Kraft Heinz -0.70%, Tyson -0.66% for an average of -1.47%, against consumer defensive at -0.90% — a 0.57-point gain on the pair on day one. The honest reading. This is a fraction of Thursday's 3.77-point dispersion and it comes on a down day for the group, which is the weaker of the two ways to win: the basket is falling with its sector rather than away from it. But the invalidation was written as "the basket recovering more than half of Thursday's underperformance within three sessions," and session one extended the underperformance instead. Walmart fell 1.18% and Costco 1.04%, so the group's Thursday distortion has partly unwound and the comparison is cleaner than it was. Catalyst: Kroger on 11 September (Section 5) as the channel read; any second payout cut in the cohort. Invalidation, unchanged: the basket recovering more than half of Thursday's underperformance within three sessions, which would mark it a one-day sympathy move; or a takeover approach for any basket constituent. Sizing: a half, dollar-neutral. Two sessions of the three-session window remain.

5. Long volatility on the semiconductor complex — the invalidation clock has started; keep the quarter, one strike from wrong

Mark. The Lululemon leg was taken off in Friday's pre-market as instructed; LULU closed -17.39%, so the leg is banked in full and is no longer part of the book. The remaining position is October volatility on the semiconductor proxy, a quarter, expressed in premium rather than delta. The honest reading, and it cuts against the position. The invalidation was written as "SOX outperforming the Nasdaq 100 on two consecutive up sessions, which would say the non-participation was positioning rather than doubt." Friday was the first of those two sessions: SOX +3.38% against the Nasdaq 100 +0.21%, on an up session for both, a 3.17-point outperformance and the largest of the window. One more and the trade is wrong on its own terms. Against that, the internals argue the dispersion is still there. The complex rallied 3.38% while Nvidia added 0.83% and Broadcom 0.21% — its two largest members contributed almost none of it — and Craig-Hallum cut Ambarella to Hold with the target down 26% to $70 on the same morning. A sector that rallies without its leaders and takes a target cut inside the move is dispersing, not re-rating. Action: keep the quarter, do not add, and mark it out on the next up session for the complex. If SOX outperforms again on an up day, close it and record the loss rather than re-reading the invalidation. Catalyst: Oracle, Adobe and Copart on 10 September after the close (Section 5); CPI 9/11. Invalidation: as written — a second consecutive up-session outperformance. Sizing: a quarter.

6. New: short the credit-bureau complex against long the S&P 500 financials — a regulatory re-rating has a second leg

Thesis. The FHFA did not criticise the bureaus on Friday; it removed their moat. VantageScore 4.0 is now accepted by Fannie Mae and Freddie Mac from every lender, effective immediately, ending a pilot that had been capped at 50 originators since 1 May and that had already reached over 9% of GSE-securitised mortgages by 31 August — an adoption rate achieved under a cap. "FICO has enjoyed a monopoly. No more." The market marked it as a single-day event: FICO -16.68% (an intraday -21% on Bloomberg's mark), Equifax -6.37%, TransUnion -5.9%, Experian -4.6% in London. What it has not marked is the second announcement in the same interview: the agency is "seriously considering bi-merge," which would cut a lender's required pulls from three bureaus to two and remove roughly a third of the mortgage-channel unit volume from whichever bureau loses the tie-break. Expression: short an equal-weighted basket of FICO, EFX and TRU against long the S&P 500 financials sector, dollar-neutral, half size. The pair isolates the regulatory event from the rate move, which matters on a day the 3-year cheapened 4 bp and the banks did nothing — Wells Fargo +0.87%, Goldman +0.07%, Bank of America -0.06%. Why now rather than after the bounce: a pricing-power removal by directive re-rates the terminal margin, not the next quarter, and the sell-side has not had a weekend to publish revised mortgage-channel assumptions. The three-day weekend is the asymmetry — the first downgrade cycle lands into a Tuesday open. Catalyst: sell-side revisions in the week of 7 September; any FHFA follow-through on bi-merge; the bureaus' own investor communications. Invalidation: the basket recovering more than half of Friday's decline within five sessions, which would say the market reads the directive as volume-neutral; or an explicit FHFA statement retreating from bi-merge. Sizing: a half, dollar-neutral. The risk to name honestly: this is a crowded short after a 17% day, borrow will be expensive, and a single conciliatory line from the agency reverses it violently.

Prior closes, marked forward. The October distillate-versus-gasoline crack pair, closed on Thursday at a $61.48 differential, would have lost a further $4.78 on Friday as the differential fell to $56.70 — so the decision to close on thesis invalidation rather than wait for the $60 stop saved that amount, and the position is recorded as having ended at the right level. The short-debasement basket against long dollar, closed on Thursday when both invalidation triggers fired, would have gained on Friday for the first time since it was closed: gold -1.38%, silver -1.31% and DXY +0.25%, so both legs would have paid on the same session, roughly 1.60 points. That is the uncomfortable mark and it is recorded rather than omitted — the trade was closed one session before the tape agreed with it, which is what a two-trigger invalidation is designed to do and what it costs. The AI-halo basket against long Nvidia, closed on 1 September, would have gained on Friday: Palo Alto +0.38%, ServiceNow -2.97%, Fortinet -0.05% and Adobe -6.73% for a four-name average of -2.34% against Nvidia +0.83%, a 3.17-point gain on the pair — its best session since being closed, and the second consecutive session in which the closed pair has swung more than three points.

The vol note. VIX closed 14.53, up 1.47%, with a session range of 13.80 to 14.58, on a day the index fell 0.38% — a 3.9-to-1 ratio of volatility gain to index decline, which is a market adding a little hedge rather than repricing risk. The four-session path is 16.34 → 15.20 → 14.32 → 14.53, so Thursday's window low held and Friday added a fifth of a point back. A 14.53 handle asks for roughly a 0.92% daily move, against a market that has just delivered 1.06%, 0.38% and a 3.38% move in one of its largest sub-indices. The rates side of the comparison is unavailable this session: MOVE did not publish a 4 September value, so its 74.68 carries a 3 September vintage and the change is withheld (Section 9 block c). That removes the cross-market divergence read entirely for one session and leaves the equity surface to be judged on its own terms — which is that it is cheap into a calendar containing a three-day weekend, a two-day gap, PPI and CPI on 10 and 11 September, and a meeting priced at 58.6%. The instruction is unchanged and now cheaper to execute: own the 10-11 September inflation block outright, finance it in the 18-19 September meeting expiry where an odds-on meeting still carries the richest premium on the curve, and do not sell the front weekly into a Tuesday reopen that has three days of unarbitraged Asian repricing behind it.

Crowded consensuses to stress-test, with the numbers.

1. "The payroll settled it." The strip says 58.6% on CME and 58.4% on Investing.com, up from 49.4% and 50.4%. Stress test: the 1 WEEK column reads 57.0% and the 1 MONTH column 58.4%, so a fortnight containing Jackson Hole, an oil shock, a dovish governor and a three-fold payroll beat has moved the September meeting 1.6 points on the week and 0.2 points on the month. Nothing has been settled; the same 0.75 basis points of ZQ price that moved it on Friday can move it back on 11 September. And the print itself is contested — ADP counted 38,000 private hires on Wednesday against an establishment survey whose beat is concentrated in 59,000 food-service jobs against a twelve-month average of 12,000, the most revisable cohort in the report.
2. "The terminal rate is anchored." It moved on Friday and the strip is explicit about it. Every 2027 contract cheapened 2.0 to 4.5 bp with the move rising monotonically along the curve, January 2027's mode flipped back up to 4.00-4.25%, cumulative-above at December 2027 rose to 93.6% from 91.2% and the first non-trivial cut probability fell to 0.5%. Stress test: this report argued on Thursday that a parallel richening of the back with a pinned front was a timing repricing. The correct reading of two sessions together is that the destination is where the volatility actually is, and it is being marked in both directions inside 48 hours.
3. "The long end is the problem." Still not — and the evidence just changed sign in the other direction. The 30-year richened 1 bp on a three-fold payroll beat, taking 20s30s to -1 bp after seven consecutive sessions at exactly zero and flattening 2s30s 4 bp to +87. Stress test: a long end that rallies when the front cheapens is a market that believes the hikes will work, and it is the one configuration in which the $215bn September corporate calendar and the 9 September buyback of unspecified size can both be absorbed. The counterweight remains Ben Emons of FedWatch Advisors, whose reading of forwards and long-term spot yields implies a hike path of roughly 120 basis points against the 60 bp the market prices through year-end.
4. "Credit is fine because credit is tight." IG at 81 bp is unchanged on the day and 2 bp wider on the week; HY at 265 bp is 18 bp through January; and CCC tightened 2 bp to 1,051 to break a three-session, 27 bp widening run. Stress test: the relief is one basis point on a series that is 20 bp wider on the week and 163 bp wider on the year, and the tradeable proxies have now been silent for three consecutive sessions — HYG and LQD have moved a cumulative 0.08% and 0.24% while the lagged index moved 3 bp then 1 bp the other way. The newer risk is not in spreads at all: SOFR is back above IORB at 3.66%, the standing repo facility took a non-zero $33m, reverse repo rose to $702m and reserves fell a fourth straight week to $2.9249tn, three weeks before quarter-end.
5. "Volatility is right to be calm." VIX at 14.53 asks for roughly a 0.92% daily move into a three-day weekend, an unarbitraged Asian repricing, PPI and CPI on 10 and 11 September and a meeting priced at 58.6%. Stress test: the usual cross-check is unavailable, because MOVE did not publish a 4 September value and its 74.68 carries a 3 September vintage — the series has now printed a "previous close" outside its own day range for eight consecutive sessions. An equity surface at a 14 handle with no functioning rates-volatility comparison is a surface being judged on half the evidence.

The two-sided geopolitical tape. Escalation is now doing what the barrel would not do a week ago. Reuters describes the fiercest U.S.-Iran clashes since July, with American attacks killing and wounding dozens of Iranian civilians; Israeli Defence Minister Israel Katz has threatened to "cripple" Iranian energy facilities; and Iran has expanded its blacklist of vessels for the Strait of Hormuz, though Iraqi tankers were approved. Crude answered with its best week since mid-July — WTI +9.38%, Brent +8.80% — and U.S. diesel at record highs, compounded by Ukrainian strikes on Russian refineries. Against that: Iraq lifted August exports to 2.34m barrels a day from 1.35m in July, President Putin signalled possible Ukraine negotiations, and ANZ notes elevated inventories absorbed the initial supply crisis while raising its short-term Brent forecast to $95. Canada's retaliation on $20bn of U.S. goods lands 8 September, the first trading day after the holiday.

Structural watch items. SOFR at 3.66%, one basis point above the 3.65% IORB, with a 99th percentile at 3.74% on $2.949tn of volume; a standing repo facility take-up of $33m alongside reverse repo at $702m; reserve balances at $2.9249tn, down $10.4bn on the week and falling for a fourth consecutive week; a 1-month bill 4 bp richer at 3.79% on a hawkish payroll, widening the 3-month-minus-1-month spread to 12 bp from 6 bp; $215bn of expected September corporate supply into a 4.78% ten-year with only three usable issuance days before the data block; the FHFA's VantageScore directive and its unpriced bi-merge follow-on; Oracle's move to 10 September putting three after-close prints into one evening; Japan's 10-year 3.7 bp richer at 2.911% with the yen 0.29% weaker at 156.233 and a Bank of Japan decision on 18 September; and Adobe's incoming chief executive Anil Chakravarthy, now taking over five days before a results date that has itself moved.

Full Data Notes & Conflicts, the Weekend / Asia read-through and the categorised source links are in the companion file US_CrossAsset_Daily_2026-09-04_DataNotes.txt.