Closing Briefing — Wednesday, September 2, 2026

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 one group has drifted. Compounding each group's 1 September YTD by Wednesday's one-day move reproduces the published YTD to within 0.10 percentage points for ten of eleven groups. The exception is technology, at 0.345 pp: 1.2253 × 1.0042 implies +23.04% against a published +23.39%. Real estate is the next widest at 0.097 pp (implied +7.91% against +7.81%), reopening a drift this report tracked and closed in late August. Clean examples for contrast: energy 1.4073 × 1.0017 = 1.4097 → +40.97% against +40.96%, deviation 0.009 pp; financials 1.0692 × 1.0114 = 1.0814 → +8.14% against +8.14%, deviation 0.001 pp. The technology gap is the one to watch, because it is the second-largest group by weight and the drift appeared on a day the group moved only 0.42%.

The rotation reversed Tuesday's defensive tilt without becoming a growth tape. Ten of eleven groups closed green and the leaders are cyclicals — materials +1.83%, communication services +1.16%, financials +1.14% — while the three defensives that led on Tuesday sat at the back of the pack: consumer defensive +0.23%, utilities +0.16% and, at the very bottom, real estate -0.51%, the only red group. CNBC's GICS tally agrees on both ends, marking materials up 1.6% as the leader and real estate down 0.6% as the laggard intraday, and counting nine of eleven sectors higher against Finviz's ten-up-one-down-one-flat. Note what did not lead: energy at +0.17%, on a day crude settled higher and inventories drew 4.45m barrels.

The composition traps to name. Technology at +0.42% understates nothing this time — SOX rose 0.45%, almost exactly in line — but the group's internals are extreme: Dell +15.81% and Nvidia +3.21% against Microsoft -0.84% and Palo Alto Networks -9.28%, so the bucket's small move is a genuine cancellation rather than a hidden story. Communication services at +1.16% is carried by Charter +8.74%, Paramount Skydance +3.98%, Fox Corp A +2.89% and Netflix +2.38%, not by Alphabet A at +0.63% or Meta at +2.47% alone. Industrials at exactly 0.00% is the flattest group print of the reporting window and sits on a -3.55% week, the worst on the board, with CNBC noting the group is down 7% quarter-to-date against energy's +22%. On the year the extremes are unchanged in shape and wider in level: energy +40.96% and technology +23.39% against consumer cyclical -4.92%, communication services -1.43% and utilities -1.04%.

Higher

Lower

Analyst and corporate actions

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 / 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.

Thu 9/3. BMO: Campbell's (CPB), Toro (TTC). AMC: Lululemon Athletica (LULU).

Fri 9/4. No S&P 500 reporter on either bucket.

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

Tue 9/8. Timing bucket not published: Oracle (ORCL) — the reviewed calendar carries no before-open or after-close designation; confirm with company investor relations.

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

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

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

Changes vs. the prior calendar (9/1 report):

Source: U.S. Department of the Treasury Daily Treasury Par Yield Curve Rates, Text View for September 2026, read after publication. Changes are versus the 1 September official row (1-day) and the 26 August official row (1-week).

Off-table bills, extracted and reported here because one of them carries the entire day's information. 1.5 Mo 3.87% (-1 bp on the day, +9 bp on the week), 2 Mo 3.89% (0 bp, +9 bp), 6 Mo 4.00% (0 bp, +6 bp) — and the outlier, 4 Mo at 4.02%, up 5 basis points on the day and 14 on the week, the single largest move anywhere on the curve on a session when nothing else moved more than two. The 4-month bill now yields 2 bp more than the 6-month and 10 bp more than the 3-month, an inversion inside the bill strip that is a supply and settlement artefact rather than a policy signal — the tenor spans the 16 September meeting from the far side without spanning the 28 October one. It belongs with the funding data in Section 9 block b, not with the coupon curve.

The read: a curve that did not move at all, after a session that reached a three-year high intraday. Every published spread is unchanged on the day, and seven of the ten published tenors printed exactly Tuesday's yield. The two that moved lower are the front — 1-month -2 bp and 1-year -2 bp — with the 3-year and 5-year a single basis point richer and the entire 7-to-30-year sector unchanged. That is not a bull steepener or a bull flattener; it is a policy-front bid with a frozen coupon curve, and the diagnostic is the gap between the intraday and the close. CNBC recorded a 10-year high of 4.814%, a level not seen since November 2023, in the European morning, and Bloomberg closed the tenor 2 bp lower at 4.78% — so the market bought roughly three and a half basis points of ten-year duration between the London morning and the New York close, and the official par construct ended flat.

What that repricing was made of, and the mechanism is the front rather than the term premium. The 1-year fell 2 bp to 4.16% and the September hike probability fell 4.9 points on CME's own columns on the ADP miss, so the front-end bid is policy timing. The long end simply refused to follow: the 30-year is unchanged at 5.27% for a second consecutive session, and on the week it is the least-moved point on the curve at +9 bp against the 2-year's +20. Bloomberg's own headline the same day — "Bessent's Bond Gains Wiped Out as Treasury Yields Jump Again" — is about a level that has now stopped rising, three weeks after the 19 August buyback announcement and one week before the 9 September operation that is supposed to at least double the programme's size.

The week is still an unambiguous bear flattener and Wednesday did not dent it. On five sessions the 2-year is 20 bp cheaper at 4.39%, the 5-year 17 bp, the 3-year 16 bp and the 7-year 15 bp, against the 10-year's 13 bp, the 20-year's 10 bp and the 30-year's 9 bp. 2s30s has flattened 11 bp on the week and 2s10s 7 bp; 3M10Y has widened 6 bp because the bill is anchored. The vendor gap is one basis point and routine: Bloomberg's live board closes the 10-year at 4.78%, -2 bp, against the official par 4.79% and 0 bp, a 1 bp difference in level and 2 bp in change — a 4:59 p.m. live mark against a 3:30 p.m. bid-side construct, on a day the intraday range was unusually wide.

Source: Federal Reserve Bank of New York Economic Indicators Calendar for September 2026 (all times Eastern). Consensus figures are carried where independently verified against the Wall Street Journal's economic calendar, Investing.com's calendar or a named survey; 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

Next week (Sep 7 - Sep 11)

Current target range: 3.50%-3.75%. A soft private payroll took nearly five points off the front meeting, and for once both vendors and the whole strip agree on the direction.

CME FedWatch headline — 16 September 2026 meeting. Data as of 2 Sep 2026, 04:57:35 p.m. CT (5:57 p.m. ET), read from the FedWatch probability table.

Provenance of every column, stated — and Tuesday's live read corrected by 0.3 of a point. CME's NOW column carries a 04:57:35 CT timestamp against a wall clock after 5:00 p.m. ET and a 4:00 p.m. CT ZQ session close, so it resolves as p.m. and sits within the hour after the close; treat it as indicative to roughly one to three percentage points. The 1 DAY column carries the legend date 1 September and prints 67.2%, against the 66.9% this report published from CME's live column on Tuesday evening — a +0.3 pp correction, the second-tightest reproduction of the reporting window after Tuesday's exact match, and the correction is recorded here rather than left standing. 1 WEEK (26 August) and 1 MONTH (31 July) carry genuine reference dates and are used in the calculations below. The Investing.com matrix underneath is timestamped 2 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. CME puts the September hike at 62.3% at 5:57 p.m. ET; Investing.com at 60.4% at 5:45 p.m. ET — a 1.9 percentage-point difference across twelve minutes, against 1.3 pp on Tuesday and 2.0 pp on Monday. The gap is large in probability and trivial in price. Investing.com publishes the September future at 96.300, up 1.0 bp from Tuesday's 96.290; 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. The entire 1.9 pp vendor gap is therefore about nineteen-hundredths of a basis point of contract price, and the day's 4.9-point CME move is about half a basis point. That leverage remains the most important number in this section.

One-day, one-week and multi-day momentum, and a vendor trap to disclose. The September hike fell 4.9 pp on CME's own columns (67.2% to 62.3%) and 5.2 pp against the figure this report published from Investing.com on Tuesday (65.6% to 60.4%). Investing.com's own "previous day" column reads 67.9%, which would make its internal move -7.5 pp; that column is a fixed daily snapshot rather than the prior evening's live value, it disagrees with what this report published twenty-four hours earlier by 2.3 pp, and the discrepancy is carried in Data Notes. On a one-week view it is 62.3% against 36.6% on CME and 60.4% against 36.0% on Investing.com, so 25.7 pp and 24.4 pp in five sessions — still a large repricing, but 1.6 and 6.2 points narrower than the week that ended Tuesday, which is the first deceleration of the run. The one-month column remains the discipline: on 31 July CME priced 67.0%, so Wednesday's 62.3% is 4.7 points below where the market sat a month ago — the round trip completed on Tuesday has now gone through to the other side. Further out, every 2026 horizon retreated: October's cumulative-above fell to 71.4% from 77.4%, December's to 88.5% from 91.0%, and the peak of the strip is 96.7% at July 2027 against 97.1% on Tuesday. The probability of a cut at any 2026 meeting remains 0.0%.

(a) Current-year meeting distributions

Investing.com Fed Rate Monitor, updated 2 Sep 2026 05:45 p.m. EDT. Format: current [prior day] [prior week]. Modal range in bold. The prior-day column is the vendor's own snapshot; see the disclosure above.

Sums are 100.0%, 100.1% and 100.0% on the published figures. Three observations. First, September's hold column rose 7.5 points on the vendor's own day and remains 24.4 points below where it sat a week ago, so the meeting is still priced for a hike but the conviction is draining rather than building. Second, the deep tail is where the day's move went, in reverse: October's +50 bucket fell to 16.7% from 21.8%, a 5.1-point unwind of Tuesday's 4.8-point build, and December's +75 bucket to 10.0% from 12.5% against Tuesday's rise to 13.5% — the market bought back exactly the tail it had sold. Third, the modal outcome at December has flipped by three-tenths of a point: two hikes at 39.4% against one hike at 39.1%, a statistical tie, after three consecutive sessions in which two hikes led by up to 6.3 points. That is the single most consequential line in this section, because it is the first time in the window that the December mode has been genuinely contested.

(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 richened at every point and undid most of Tuesday's cheapening. The eight contracts print 95.990, 95.890, 95.840, 95.775, 95.760, 95.735, 95.750, 95.775 against Tuesday's 95.960, 95.855, 95.800, 95.740, 95.725, 95.725, 95.725, 95.755 — 1.0 to 4.0 basis points richer across the strip, with the largest recovery in the January-to-July belly. Tuesday's move cheapened those same points 2.5 to 5.5 bp, so roughly two-thirds of it has been reversed in a session. July 2027's mode has flipped back to 4.00-4.25% at 30.6%, ending the one-session experiment in which 4.25-4.50% led by a tenth of a point, and the trough of the strip has moved from 95.725 to 95.735 at September 2027. Cumulative-above fell at every point: December 2027 to 93.3% from 94.7%, June to 96.1% from 97.1%. The first non-trivial cut probability rose rather than fell, to 0.8% at 3.25-3.50% in December 2027 from 0.5%. Read against Tuesday, when an oil shock repriced the terminal rate, this is a labour print repricing it straight back — which means the level of the policy rate is now the contested variable, and it is moving on data in both directions.

(c) Year-end probability ladders

Year-end 2026 — the 9 December meeting.

Cumulative above the current range: 88.5%. Cumulative below: 0.0%. Sum: 100.0%.

Year-end 2027 — the 8 December meeting.

Cumulative above the current range: 93.3%. Cumulative below: 0.8%. Sum: 100.1%.

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 October 2026 row and the December 2027, March 2027, April 2027 and July 2027 rows sum to 100.1%, and the June 2027 row to 99.9%, for exactly this reason.

(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 1 September 2026 as-of date, not the 2 September close. Same-day direction is cross-checked against the cash-market proxies underneath and against Bloomberg and WSJ credit coverage.

CDX — the full six-step ladder was worked again and is reported so the gap stays auditable. (1) Bloomberg in Chrome: /markets and /markets/rates-bonds were rendered; the quote board and fixed-income tables populate and a full-text scan returns zero occurrences of the string. (2) WSJ Market Data bonds page: rendered, zero occurrences in a full-text scan. (3) Cbonds carries dedicated CDX.NA.IG 5Y and CDX.NA.HY 5Y index pages but masks the levels behind a subscription; 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 2 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 $79.11, +0.01%, and LQD $105.35, +0.12%, against 1 September closes of $79.10 and $105.22. No CDX level is published here, because an undated third-party digest number is not a CDX level.

The tail widened for a third consecutive update while the cash market did nothing, and that divergence is the section. CCC widened 7 bp to 1,049 on the 1 September update — after Tuesday's 16 bp, so 23 bp in two sessions — taking the CCC-minus-HY differential to 784 bp from 779, a fresh extreme for the reporting window. Against that, IG widened one basis point to 81 and HY two to 265, so the aggregate is still 16 bp through where 2026 started at the HY level while the tail is 164 bp wider. What changed on Wednesday is the corroboration: on Tuesday HYG fell 0.89% against a 4 bp move in the 10-year and the cash market agreed with the index; on Wednesday HYG closed +0.01% and LQD +0.12% on an unchanged curve, so the cash market registered nothing at all. A lagged index widening at the tail while the same-day cash proxy is flat is either the index catching up to Tuesday or the cash market declining to confirm — Thursday's FRED update settles it, and it is the single most useful credit number of the week.

(b) Money-market and funding plumbing

New York Fed reference rates, published ~8:00 a.m. ET for the prior business day. The rates below carry the 1 September 2026 effective date. Reverse repo take-up is the same-day 2 September operation.

The turn cleared but not completely, and the answer to Tuesday's question is "no." This report asked whether SOFR would return below IORB on the 1 September effective date. It did not: SOFR printed 3.66%, one basis point through the 3.65% IORB, narrowing from Monday's three but still above the administered rate for a second consecutive effective date — the first back-to-back through-IORB pair of the reporting window. The distribution normalised faster than the mean: the 99th percentile fell to 3.74% from 3.77% and volume to $2,912bn from a window-high $3,056bn. Tri-party and broad general collateral both gave back the whole 4 bp turn premium, to 3.63%, so the collateralised market is clean and the residual pressure sits in the SOFR mean rather than in the repo tails. The unsecured market never moved: EFFR and OBFR unchanged at 3.63% with EFFR volume rising to $114bn from $105bn. Reverse repo take-up drained again to $525m on 2 September from $725m on 1 September and $6.726bn on 31 August, a thirteen-fold two-day drain. Reserve balances still read $2.9249tn for the week ended 26 August, $68bn below the 5 August peak; the observation covering the turn publishes in Thursday's H.4.1 and is the number to watch.

The bill strip finally said something, and it said it at four months. Section 6 records the 4-month bill at 4.02%, up 5 bp on the day and 14 on the week, on a session when nothing else on the curve moved more than two — leaving it 2 bp above the 6-month at 4.00% and 10 bp above the 3-month at 3.92%. A four-month bill matures in early January: it spans the 16 September meeting and the 28 October one from the far side without spanning December. An isolated five-basis-point cheapening in exactly that window, while the 1-month richened 2 bp and the 6-month did not move, is a supply and settlement signature — a bill auction clearing cheap into a month whose reverse repo balance has collapsed to $525m and whose reserve balances are $68bn off the peak. It is not a policy signal, and it should not be read as one; if it persists through Thursday's auctions it becomes a collateral-scarcity story instead.

(c) Rates volatility and swap spreads

The MOVE series updated again, to 77.88 on the 1 September stamp, with a day range of 75.32-77.88 and an open of 75.32 — internally consistent within its own day and reconciling exactly to the 75.32 this report published for 31 August, which makes the +2.56 point move on Tuesday publishable. What remains disqualifying is the stated "previous close" of 95.74, roughly eighteen points outside the series' own day range for a sixth consecutive session, so the 2 September change is still withheld and the level carries its vintage. The direction it does establish matters: rates volatility rose on Tuesday and equity volatility fell 6.98% on Wednesday, so the two markets are now moving apart, and the last readable MOVE print of 77.88 sits 8.29% above where the series stood a week earlier on the vendor's own weekly field, while VIX has round-tripped its two-day spike entirely. That is the configuration Bloomberg's Cameron Crise flagged on Tuesday — "most of the significant downdrafts in stock prices over the past few years have coincided with spikes in rate vol" — with the equity market pricing the calm and the bond market not yet confirming it. The swap-spread basis is unchanged in substance and unchanged in vintage: Treasuries have outperformed equivalent-maturity swaps since the 19 August buyback announcement, which Bloomberg cites as evidence the intervention did something even as the level round-tripped.

(d) Issuance, leveraged loans and private credit

Levels from the TradingEconomics currency board taken after the U.S. close. The vendor's date column reads Sep/02 for DXY, EUR, AUD, NZD, CAD, KRW, TWD and HKD and a live clock for GBP, JPY, CNY, CHF and SGD; every row's %Chg was verified against the prior edition's levels for the same vendor and reconciles to within 0.03 percentage points on the majors, so the published changes are genuine 24-hour moves and are reproduced. KRW and TWD reconcile to 0.11 and 0.14 pp and are flagged in Data Notes. USD/CNH is from Investing.com, whose session had rolled, so its move is computed against the prior edition. Quote basis: USD per unit for EUR, GBP, AUD and NZD; units per USD for JPY, CHF, CAD, KRW, TWD, CNY and CNH.

The take: the yen did the whole job, and it did it without the Bank of Japan. USD/JPY fell 0.91% to 158.724, the biggest single-day yen gain of the reporting window, taking the pair 1.44 yen below Tuesday's 160.16 close after three consecutive sessions above 160 — the level intervention watchers had been treating as live. Nothing came from Tokyo: no meeting, no statement, no operation, and the Bank of Japan does not decide until 18 September. What moved was the American front end, where the 1-year fell 2 bp and the September hike came off nearly five points, and that is the correct mechanism for a carry pair. The tell is what it says about last week: a currency that needed a 4.9-point repricing of a single meeting to move 1.44 yen was not being held up by Japanese policy expectations at all.

Sterling is the cross that will not behave, and this is the third session of it. GBP/USD fell 0.25% to 1.34821 on a day when 10-year gilts rose 4 bp in early trade and closed just 1 bp higher at 5.23% and the 30-year gilt was flat at 5.8537% — so the currency fell while the bond market did essentially nothing, after falling on Tuesday when the bond market cheapened 16 bp. A currency that declines on a rate selloff and on a rate stabilisation is not trading rates; it is trading a fiscal risk premium, and it has now given up 0.48% across two sessions with the gilt curve net 5 bp cheaper. That is the most informative signal on this board and it points at the November budget, not at the Bank of England.

The antipodean split and the Asian inversion are the second-order reads. The Australian dollar rose 0.35% on GDP beating at +2.1% year on year while the New Zealand dollar fell 0.74% — a 1.09-point divergence between two currencies that normally trade as one block, and the widest on this board. In Asia the pattern from Tuesday inverted precisely: USD/KRW fell 1.15% to 1357.91, the won's best session of the window, on the day the Kospi fell 4.00%, its worst — twenty-four hours after the won posted its worst session on the day Korea reported record chip exports. A currency that strengthens on a 4% equity decline and weakens on a record export month is, twice over, not being driven by portfolio flow or by the trade balance; the corporate-repatriation explanation this report has carried for three sessions is the one that fits both observations. USD/TWD rose 0.15% in the opposite direction again, and USD/CNH at 6.7179 firmed 0.06% with the yen rather than with the region's equities.

Settlement basis, stated, and it has changed since the prior edition. All rows are quoted from the Investing.com per-contract historical settle series read after the close, which is this report's settle series of record. WTI, RBOB, heating oil and natural gas are on the October contract; gold, silver and copper on December; Brent on the November contract, which has been the front since the 23 August rollover. The prior edition's WTI $90.94, Brent $95.21 and gold $4,375.40 were late-electronic last-trade prints, not settlements; the settle series marks 1 September at $90.22, $94.65 and $4,396.40, so all day changes below are computed on the settle basis and the reconciliation is in Data Notes. Weekly and year-to-date columns are TradingEconomics spot returns on the vendor's Sep/02 date stamp.

\*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 spot weekly changes on the same caveat. The daily settles, changes and percentage moves are on the futures basis named in each row. Mixing the two would be a basis error; they are presented in separate columns for exactly that reason.

The energy complex broke internally, and the direction of the break is the trade. EIA reported a 4.45m-barrel crude draw against a 0.4m draw expected (Investing.com economic calendar), a bullish surprise of roughly four million barrels — and WTI settled up 41 cents, or 0.45%, while heating oil fell 0.31% and RBOB 1.12%. That is a crude-bullish print producing a products-bearish tape, which only happens when the refined-product premium was carrying a risk that has just been repriced. The risk was Hormuz, and it was repriced twice on Wednesday: Energy Secretary Chris Wright told CNBC more than 17m barrels transited the strait by ship on Monday, a post-war record, with regional exports above pre-war levels once the Saudi and Emirati bypass pipelines are counted against roughly 20m bpd before the war began on 28 February; and President Trump said the strikes on Iran would "likely be short-lived," reiterating that the United States controls the strait. Middle distillate is the barrel that pays for a closed strait. An open strait is worth 31 cents a gallon less.

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

Read the two together and the position has changed character. Monday and Tuesday paid the distillate leg outright; Wednesday paid only the relative leg, and it paid it because gasoline is entering the post-Labor-Day roll-off with a winter-grade specification already in the October contract. On the TradingEconomics spot basis heating oil is still +119.67% year to date against gasoline's +80.95%, so the structural argument is intact and the tactical one has just failed its first test. EIA natural gas storage Thursday 10:30 is the next scheduled read, and it is the more interesting one now, because October natural gas settled $3.003, up 3.41%, its first close above three dollars in the reporting window, on a day nobody was watching it.

The metals bounced and copper refused, which is the same tell as Tuesday with the sign reversed. Gold rose 0.85% to $4,433.85 and silver 0.89% to $65.950, recovering roughly two-fifths and one-third of Tuesday's settle losses, with gold opening at $4,370.59, trading down to $4,329.61 and closing within eleven dollars of its high — a genuine intraday reversal rather than a drift. Copper rose 0.08%. On Tuesday copper fell with the precious complex and this report read it as the growth leg joining the monetary leg; on Wednesday the monetary leg recovered and the growth leg did not, on a session when basic materials equities rose 1.83% and steel and fertiliser led them. So the equity market bought domestic cyclicals while the industrial metal sat still, which says the materials rally is about domestic spreads and not about global demand. The gold-silver ratio is essentially unchanged at 67.23 from 67.26. Bitcoin was little changed at $77,430.85 with ether -1% at $2,396.29 (Bloomberg), so the debasement complex participated in the bounce only in the metals.

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 — the calendar spread gave back a day; hold the quarter, hold the stop

Mark first. Long ZQU6 (September 2026) against short ZQZ6 (December 2026), DV01-matched one-for-one at $41.67 per basis point per contract, entered eleven sessions ago at 96.325 / 96.160 for a spread of 16.5 bp, trimmed to a quarter on Friday, with the stop raised to 24.5 bp on Tuesday. Wednesday's mark: ZQU6 96.300, ZQZ6 96.040 — a spread of 26.0 bp. That is -1.5 bp on the day, worth -$62.50 per contract pair on the quarter, and it leaves the trade +9.5 bp from entry, a basis point and a half off Tuesday's high. The stop was not touched.

Why it lost, and the mechanism is exactly the one that made it work. The spread needs December to absorb more of every cheapening than September. Wednesday delivered a richening, and December absorbed more of that too: ZQU6 rose 1.0 bp while ZQZ6 rose 2.5 bp, because a soft ADP print unwound the deep tail faster than the front meeting — December's +75 bucket fell to 10.0% from 12.5% and its cumulative-above to 88.5% from 91.0%, while September's hike fell 4.9 points on CME. The position is short the far tail of the 2026 distribution, so it makes money when the tail extends and loses when the tail is bought back. Tuesday's oil shock extended it; Wednesday's labour print bought it back. That symmetry is now the whole risk, and Friday is a labour print.

The modal path, the base case and the tails. Modal path: a 25 bp hike on 16 September is still modal at 60.4% on Investing.com and 62.3% on CME, with ease at 0.0%; a hike is modal at 28 October too, at 54.7% against 71.4% cumulative-above; and year-end 2026 is now a genuine coin-flip between one hike and two — 39.1% against 39.4%, the tightest that contest has been in the window. The 2027 strip is modal 4.00-4.25% at all eight meetings after July flipped back. Base case: one to two hikes into year-end, with a terminal rate that has now moved on oil in one direction and on labour in the other inside forty-eight hours, which argues the level is data-dependent rather than anchored. Tail one, and it is Friday: payrolls at a consensus of +55,000 to +58,000 with unemployment at 4.1%, against a July print that fell 23,000 with the prior two months revised down a combined 103,000. A second consecutive negative headline unwinds a 62%-priced hike, and at ten percentage points of probability per basis point of ZQ price this spread can move 3 bp in an hour — against the position, because December would rally more than September. Tail two: an ISM services beat Thursday at 54.2 reversing Wednesday's front-end bid. Practical implication: the position carries 9.5 bp of profit on a quarter into a payroll it is short the wrong way, followed by a three-day weekend in which nothing can be traded. Hold the quarter, do not add, leave the stop at 24.5 bp, and buy the tail rather than reducing further — a ZQZ6 call spread struck around 96.10/96.20 funds the gap risk more cheaply than cutting size does, because the loss scenario is a December rally rather than a September one.

Expression: long ZQU6 / short ZQZ6, DV01-matched one-for-one, quarter size. Catalyst: claims and ISM services 9/3; payrolls 9/4 08:30; the buyback operation 9 September; PPI 9/10; CPI 9/11; the 16 September FOMC. Invalidation: the spread through 24.5 bp; or the September cumulative hike printing below 45% on either vendor; or any 2026 meeting showing a non-zero cut probability. Sizing: a quarter, at $41.67 per basis point per pair.

2. Long the October distillate crack against short the October gasoline crack — the first losing session; hold the quarter, move the stop up

Mark, on the restated settle basis. The two legs stand at $105.21 and $39.57 for a differential of $65.63, against $106.23 and $41.45 for $64.77 on Tuesday's settle — a session gain of +$0.86 on the relative trade and a -$1.02 loss on the distillate leg outright, the first since the position was opened. The honest reading: the thesis is a closed-strait premium in middle distillate, and on Wednesday two officials attacked it directly — 17m barrels through Hormuz on Monday, a post-war record per Energy Secretary Wright, and Trump's "short-lived." Heating oil fell 0.31% on a day crude rose 0.45% and crude stocks drew 4.45m barrels. That is the invalidation mechanism working in miniature. The position survived only because gasoline fell faster on its own seasonal roll-off, which is not the thesis. Action: hold the quarter, and raise the stop on the differential to $60 from the $56 entry level, banking most of the move rather than defending a trade whose driver is now being talked down by the administration. Catalyst: EIA natural gas storage Thursday 10:30 as a heating-fuel substitution cross-check; the post-Labor-Day gasoline roll-off; any further de-escalation language or, against it, a fresh shipping incident. Invalidation: the differential through $60, or a second consecutive session in which the distillate leg loses money outright while crude rises. Sizing: a quarter, unchanged.

3. Long the 20-year against the 30-year, on the November refunding — hold the half

Expression: long the 20-year bond against short the 30-year, DV01-neutral, half size. Mark: 20s30s at 0 bp with both tenors at 5.27% — unchanged on the day and 1 bp flatter on the week, a sixth consecutive session of no spread at all. What changed: nothing in the spread, and something in the environment. The 10-year touched 4.814%, the highest since November 2023, and then closed the official curve unchanged, so the long end absorbed a three-year-high test without cheapening — which is the first evidence in a fortnight that the buyback has a floor under it. Against that sits a $215bn September corporate calendar and an operation on 9 September whose size Treasury has described only as "at least double." The dislocation this trade waits for still needs the calendar to arrive. 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.

4. Protection on the CCC cohort funded in IG — the add-back was right; hold the half

Expression: long CCC-exposed credit protection (or short a levered-loan / CCC-heavy vehicle) against long IG cash. Mark: CCC 1,049 bp, +7 bp; HY 265 bp, +2 bp; IG 81 bp, +1 bp on the 1 September FRED update, taking the CCC-minus-HY differential to 784 bp from 779 — a 5 bp gain, on top of Tuesday's 13. The honest reading: Monday's cut to a quarter was wrong and was marked as such; Tuesday's decision to go back to a half has now been paid twice, for a cumulative 18 bp across two updates. What argues against adding further is the cash market: HYG closed +0.01% and LQD +0.12% on a flat curve, so nothing in the same-day tape corroborated the lagged index, in contrast to Tuesday when HYG fell 0.89%. A position that is only being confirmed by a one-day-lagged series is a position to hold, not to press. Action: hold the half. Catalyst: payrolls 9/4; the September IG calendar clearing into a $215bn month; CPI 9/11. Invalidation: 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.

5. Short the debasement complex against long the dollar — both legs lost; cut to a token and set a hard stop

Expression: short an equal-weighted basket of Comex gold, Comex silver and a bitcoin proxy against long the dollar index, dollar-notional matched. Mark: the short basket returned gold +0.85%, silver +0.89%, bitcoin +0.24% for an average of +0.66%, against DXY -0.10% — a 0.76-point loss on the day, and the first session in which both legs lost together. The honest reading: Tuesday's version had both legs working and this book took a third off at the top, which was correct. The reason to cut further rather than hold is that the mechanism has inverted: the short worked when the front end was cheapening, and the front end is now richening on labour data with two more labour prints inside forty-eight hours. Gold reversed $104 off its intraday low of $4,329.61 to close near the high, which is buying rather than short-covering. Action: cut to a token position and set a hard stop. Catalyst: claims and ISM services 9/3, payrolls 9/4, CPI 9/11; Chinese physical demand on the Shanghai open. Invalidation, restated onto the settle basis for consistency: gold reclaiming the 31 August settle of $4,481.50; or DXY back below 99.11, last Thursday's close. Sizing: a token, roughly a tenth.

6. New: short the beat-and-raise reaction in AI hardware into the consumer prints — long-vol, not directional

Thesis, and it is the day's clearest observation. On Wednesday evening three companies beat and raised and two of them were sold: Broadcom fell about 5% on a fourth-quarter guide $0.23bn light before recovering on fiscal-2028 promises; HPE beat by 19% on earnings, raised full-year revenue by roughly $1.5bn above consensus, and traded -0.62% after hours; NetApp posted record revenue +30%, raised the year by about $0.6bn at the midpoint, and fell 8% on free cash flow. That is not a fundamentals problem; it is a positioning and discount-rate problem, and it says the marginal buyer will not pay for guidance in a complex whose valuation is a duration bet. Expression: buy short-dated straddles or reverse collars on Lululemon into Thursday's close rather than taking a direction, sized small — and, for the hardware complex itself, own October volatility on a semiconductor proxy rather than shorting the names, because the recovery in Broadcom after the call shows the tape can go either way inside an hour. Catalyst: Campbell's and Toro before Thursday's open, Lululemon after it (Section 5); payrolls Friday; Oracle on 8 September. Invalidation: Lululemon beating and rallying more than 5%, which would say the reaction function is confined to AI hardware and not general; or implied volatility on the semiconductor proxy pricing above realised, which removes the edge. Sizing: a quarter, expressed in premium rather than in delta.

Prior closes, marked forward. The AI-halo basket against long Nvidia, closed Tuesday at +2.41 pp on the session and +0.51 pp cumulative, would have gained a further 7.22 points on Wednesday: Palo Alto -9.28%, ServiceNow -4.32%, Fortinet -4.52%, Adobe -2.21% and Synopsys +0.28% for a basket average of -4.01%, against Nvidia +3.21%. That is the largest single-session gain the pair would ever have produced, and closing it into the hardware prints was the wrong call by a wide margin — recorded here rather than quietly dropped. The gasoline-versus-distillate crack pair, closed eight sessions ago, would have gained $1.88 on Wednesday as the gasoline crack fell, its first favourable session in a week. The long-silver-against-short-gold pair, closed on thesis invalidation, would have gained 0.04 points: silver +0.89% against gold +0.85%.

The vol note. VIX closed 15.20, down 6.98%, with a low of 15.12, on a session the index gained 0.46% — a 15-to-1 ratio of volatility decline to index gain, the exact inverse of Tuesday's 13-to-1, and a complete round trip of the spike in a single day. Two sessions have taken the index -0.71% then +0.46% and VIX 14.92 to 16.34 to 15.20. What the surface is now paying for is unchanged: claims and ISM services Thursday, payrolls Friday 08:30, then a three-day weekend, then PPI on the 10th and CPI on the 11th, into a meeting priced at roughly 62%. A 15.20 handle asks for about a 0.96% daily move against a market that has delivered 0.33%, 0.71% and 0.46% in three sessions, so the near-dated surface is now cheap rather than fair — and it is cheap into a payroll that cannot be traded again until Tuesday. Reverse Tuesday's instruction: buy back index gamma dated across 4 September, keep the 10-11 September inflation block already owned, and finance the near-dated leg by selling the 18-19 September expiry, where the meeting premium is richest. Rates volatility remains unmarkable for the current session — the MOVE series is one day stale at 77.88 — and that stale print is 8.29% higher on the week while equity volatility is not, which is the divergence to hedge rather than to fade.

Crowded consensuses to stress-test, with the numbers.

The two-sided geopolitical tape. De-escalation, and it was the day's dominant signal: President Trump said the strikes on Iran would "likely be short-lived," reiterating that the United States controls the Strait of Hormuz; Energy Secretary Chris Wright told CNBC more than 17m barrels transited the strait by ship on Monday, a record since the war began on 28 February, with regional exports above pre-war levels once the Saudi and Emirati bypass pipelines are counted against roughly 20m bpd pre-war; and no new shipping incident was reported on Wednesday. Escalation risk that has not gone away: the strikes themselves continued into Tuesday night, U.S. Central Command has tied them to attempted IRGC attacks on commercial shipping and American service members, and Canada's retaliation on $20bn of U.S. goods lands 8 September. Domestically the fastest-moving risk has moved from the legislature to the balance sheet: PG&E opened a strategic and financial review into a possible reorganisation and deferred $2bn of 2027 capital spending, and the stock fell another 5.23% to $13.33 after Monday's 20% collapse.

Structural watch items. SOFR printed 1 bp through IORB at 3.66% for a second consecutive effective date, the first back-to-back pair of the window, while general collateral gave back the entire turn premium to 3.63% and reverse repo take-up drained to $525m; reserve balances at $2.9249tn with the observation covering the turn still to publish; the 4-month bill at 4.02%, up 5 bp on a day nothing else moved more than two, now yielding above the 6-month; $215bn of expected September corporate supply into a 4.79% ten-year; Japan's 10-year above 3% with the yen 0.91% stronger at 158.72 and the Bank of Japan deciding 18 September; 30-year Bunds near their highest since 1998 at 3.8468%; Uber cutting 3,300 roles, 10% of its workforce; and Nokia rejoining the Stoxx 50 with Volkswagen, down 30% year to date, likely dropping out.

Full source links and the complete Data Notes & Conflicts appendix are in the companion file US_CrossAsset_Daily_2026-09-02_DataNotes.txt, alongside the canonical Markdown report of record.