Closing Briefing — Thursday, August 27, 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 — CNBC's GICS tally counts information technology as the session's only positive sector, which is consistent with Finviz's basic materials and energy sitting within 21 bp of unchanged.

YTD reconciliation, and two groups drifted. Compounding each group's 26 August YTD by Thursday's one-day move reproduces the published YTD to within 0.01 percentage points for nine of the eleven groups. Worked examples: technology 1.2183 × 1.0309 = 1.2560 → +25.60% against a published +25.60%, deviation 0.005 pp; consumer defensive 1.0785 × 0.9861 = 1.0635 → +6.35%, deviation 0.001 pp. The two exceptions are real estate, where 1.1108 × 0.9905 = 1.1003 → +10.02% against a published +9.83%, a 0.195 pp gap, and basic materials, where 1.2297 × 1.0021 = 1.2323 → +23.23% against +23.12%, a 0.108 pp gap. Both drifts are downward revisions to the published YTD rather than to the one-day figure, which is the signature of a constituent change or a dividend adjustment inside the group rather than a data error; they are flagged and carried, not corrected.

One group did the work and ten did not. Technology +3.09% is the widest single-session sector gap of the reporting window, and the distance to second place — basic materials at +0.21% — is 288 basis points. Underneath it the leadership is narrow even by technology's standards: Synopsys +13.39%, Palo Alto Networks +12.83%, ServiceNow +10.04%, Fortinet +9.67%, Nvidia +8.74%, Autodesk +6.21%, Adobe +5.73%, CDW +5.31%, Broadcom +4.49%, Intel +4.36% and IBM +3.88%. That is a software-and-security list with the accelerator at its centre, not a semiconductor list. Consumer defensive -1.39% was the laggard on Hormel -10.25%, Kroger -2.70% and Costco -2.24%, and the four consumer names that reported before the open produced three declines.

The weekly column has flipped back toward risk and the yearly one has not moved. Over five sessions basic materials +3.28%, technology +2.83% and financial +1.76% lead, while energy is -2.24% despite Thursday's crude rally — the equity has not followed the barrel. On the year the shape is unchanged and extreme: energy +35.82% and technology +25.60% against consumer cyclical -4.03% and communication services -1.91%, still the only two negative groups. Note the Finviz composition trap that matters this session: Amazon (-1.54%) sits in consumer cyclical and Alphabet (-0.39% on the A line) and Meta (-0.87%) in communication services, so both negative-YTD groups carry megacap weight a GICS reader would file under technology or discretionary — and all three fell on the day technology gained 3.09%.

Higher

Lower

After the close

Analyst 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.

Fri 8/28. No S&P 500 reporter on either bucket. The Nasdaq capture returns twelve names for the date, none of them constituents.

Mon 8/31. No S&P 500 reporter on either bucket.

Tue 9/1. BMO: Medtronic (MDT). AMC: Palo Alto Networks (PANW), Dell Technologies (DELL).

Wed 9/2. BMO: Brown-Forman (BF.B). AMC: Broadcom (AVGO), Hewlett Packard Enterprise (HPE), NetApp (NTAP).

Thu 9/3. BMO: Campbell's (CPB), Toro (TTC). AMC: Lululemon Athletica (LULU). Timing bucket not published: Copart (CPRT) — the reviewed calendar carries no before-open or after-close designation for this date; confirm with company investor relations.

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

Changes vs. the prior calendar (8/26 report):

Source: U.S. Department of the Treasury Daily Treasury Par Yield Curve Rates, Text View for August 2026, read after publication. Changes are versus the 26 August official row (1-day) and the 20 August official row (1-week).

Off-table bills, extracted and reported here because they carry the financing story. 1.5 Mo 3.79% (+1 bp on the day, +2 bp on the week), 2 Mo 3.81% (+1 bp, +2 bp), 4 Mo 3.88% (unchanged on both), 6 Mo 3.94% (unchanged on both). The very front of the bill strip cheapened into month-end while the 3-month richened a basis point — a 7 bp spread between the 2-month and the 3-month moving 3 bp wider in a week with no policy content. See Section 9 block b.

The read: a one-basis-point parallel shift on the day, and a front-end-led bear flattener on the week. Take the day first, because its uniformity is the diagnostic. Every coupon tenor from two years to thirty rose exactly one basis point; the 1-year rose two and the 3-month bill fell one. A curve that moves in parallel after a labour print that beat, on a day two regional presidents argued publicly for a hike, is a curve with nothing left to reprice at the level — the whole argument has moved into the distribution, which is where Section 8 finds it. 2s10s, 2s30s and 20s30s were all unchanged; the only spread that moved was 3M10Y, 2 bp wider to 83, and it moved from the bill side.

The week is where the shape lives. On a five-session view the 1-year is 5 bp cheaper at 4.04% — the largest weekly move on the entire curve — the 3-year 4 bp cheaper, the 2-year 1 bp cheaper, while the 5-year and 7-year are 1 bp richer, the 10-year and 20-year 2 bp richer and the 30-year 4 bp richer at 5.19%. That is a bear flattener driven from the one-year point, and it has a precise translation: the market has spent the week pulling forward the timing of the next hike into the twelve-month window while simultaneously buying duration beyond five years. 2s30s flattened 5 bp to 99 and 2s10s 3 bp to 47 on exactly that mechanism. The corroborating evidence is the bill: the 3-month fell 3 bp on the week to 3.84% while the 1-year rose 5 bp, widening the 3M-to-1Y segment from 12 bp to 20 bp. Nothing happens to the Fed inside three months; a great deal is now priced inside twelve.

The auction and the vendor gap. The Wall Street Journal's bond desk headlined the day "Yields on Seven-Year U.S. Treasurys Sold at Auction Near Two-Year High" — a seven-year that cleared near a two-year cheap on the day it rose only a basis point in the secondary market is a supply datapoint, not a rates one, and it sits directly on the segment the buyback programme has not yet touched. Bloomberg's live board marks the 10-year at 4.68%, +3 bp, against the official par 4.67%, +1 bp; the level gap is one basis point and the change gap is two, because Bloomberg's mark is a Friday-Asia print taken at 10:39 p.m. ET while the official par is a Thursday 3:30 p.m. bid-side construct. It is a timing artefact and it has now run at one basis point of level difference for three consecutive sessions.

What Friday's keynote does to the long end, per the people who own it. Bloomberg's Michael MacKenzie and Ye Xie reported that JPMorgan, Apollo and Morgan Stanley all see Warsh with a chance to convince the market that inflation is his priority, and that a credible rebuke would trigger buying of 30-year bonds, whose yields "hit the highest since 2007 last week." JPMorgan Investment Management's Priya Misra: if he can, "some of the angst on Fed credibility will reduce." Apollo's Torsten Slok framed the risk directly: "He will have to deliver something that is clearer than the July press conference," and "if he does not give any framework guidance, the risk is that it will involve a much higher move in long rates." Bloomberg's own rates strategist Alyce Andres supplied the trading instruction: "The composition of any bond selloff after Federal Reserve Chairman Kevin Warsh's comments Friday is crucial — it will signify the difference between a market adjusting to a new Fed regime and a market beginning to question whether the Fed has one." The 20-year at 5.18% against the 30-year at 5.19% — one basis point of pickup for ten fewer years of duration — is the position that resolves either way on that composition. See Section 9 block c and Section 12 idea 2.

Source: Federal Reserve Bank of New York Economic Indicators Calendar for August and September 2026 (all times Eastern). Consensus figures are carried only 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 24-28) — still to come

Not on the statistical calendar but the week's largest scheduled risk: the Kansas City Fed's Jackson Hole Economic Policy Symposium runs through Saturday 29 August, themed "Financial Innovation: Implications for Payments and Policy." Chair Kevin Warsh delivers his first Jackson Hole keynote on Friday 28 August at approximately 10:00 a.m. ET, livestreamed by the Kansas City Fed. It lands nineteen days before the 16 September FOMC. Sensitivity: Very high.

Next week (Aug 31 - Sep 4)

Current target range: 3.50%-3.75%. Two vendors, two snapshot times twelve hours apart, and a 0.5 percentage-point gap that is entirely explained by the gap between them.

CME FedWatch headline — 16 September 2026 meeting. Data as of 27 Aug 2026, 09:19:39 a.m. CT (10:19 a.m. ET), read from the FedWatch probability table.

Provenance of every column, stated — and note that NOW is a morning read this session. CME's NOW column carries a 09:19:39 CT timestamp, which is 10:19 a.m. ET, roughly two hours after the claims print and six hours before the close. It is therefore neither a settlement snapshot nor an evening indicative read, and it is used here only for direction and for the reference columns beside it. 1 DAY carries the legend date 26 August and prints 36.6% for the hike — against the 36.5% this report published as an indicative live CME read on Wednesday evening, a 0.1 pp difference that retrospectively validates that read. 1 WEEK (20 August) and 1 MONTH (27 July) are genuine reference dates rather than chart reads and are used in calculations. The Investing.com matrix below is timestamped 27 Aug 2026, 10:05 p.m. EDT — six hours after the close — and is the primary source for every number in parts (a), (b) and (c).

The CME-versus-Investing.com gap, quantified. CME puts the September hike at 33.9% at 10:19 a.m. ET; Investing.com at 34.4% at 10:05 p.m. ET — a 0.5 percentage-point difference across a twelve-hour snapshot gap, against 0.5 pp across nine minutes on Wednesday. The two vendors agree on the level and agree on the direction; what the gap measures this session is the afternoon, in which the hike probability rose half a point off the morning low while equities were adding 0.72%. Investing.com publishes the underlying September future price at 96.325, unchanged from Wednesday, which is the cleanest single statement of the day: the September contract did not move at all, and every basis point of repricing happened further out the strip.

One-day, one-week and multi-day momentum. The September hike fell 2.7 pp on CME's own columns (36.6% → 33.9%) and 1.3 pp on Investing.com's prior-day column (35.7% → 34.4%). On a one-week view it is roughly flat: 33.9% against 36.1% on CME and 34.4% against 35.3% on Investing.com, so the week has taken about 1 to 2 pp out of September. The one-month column still holds the structural move: on 27 July CME priced 55.7% at 375-400 plus 25.8% at 400-425, a cumulative 81.5% chance of at least one hike by September, now 33.9% — a 47.6 pp collapse in a month. The day's shape is the opposite of the headline. October's cumulative hike rose to 52.5% from 52.0%, December's to 73.6% from 71.7% and from 66.8% a week ago, January 2027's to 79.3% from 78.3%, and — the session's genuinely new fact — the modal range at the June, July, September and October 2027 meetings stepped up from 3.75-4.00% to 4.00-4.25%. September down, December up, 2027 up a whole range: the strip is no longer pricing one hike late, it is pricing two. The probability of a cut at any 2026 meeting remains 0.0%.

(a) Current-year meeting distributions

Investing.com Fed Rate Monitor, updated 27 Aug 2026 10:05 p.m. EDT. Format: current [prior day] [prior week]. Modal range in bold.

All three rows sum to 100.0% on the published figures. October's hold has fallen below 48% for the first time, at 47.5% against a cumulative 52.5% above, having been 48.0%/52.0% on Wednesday and 51.3%/48.7% a week ago — a 3.8 pp swing in five sessions. December's modal range stays +25 bp at 45.0%, essentially unchanged, while the entire week's gain is in the tail: +50 bp went from 19.3% to 24.4% and +75 bp from 2.6% to 4.2%, a combined 6.7 pp transferred out of the hold column.

(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 modal range moved up a step, and that has not happened before in this reporting window. Twenty-four hours ago every 2027 meeting had 3.75-4.00% as its most likely outcome. Thursday's strip puts 4.00-4.25% at the top from June through October 2027, and leaves December 2027 an exact 31.0% / 31.0% tie between the two. ZQ prices fall from 96.095 in January to a trough of 95.920 in September 2027 — 1.5 bp lower than Wednesday's 95.935 trough and now sitting one meeting later — before recovering to 95.945 by December. Cumulative-above peaks at 88.1% in September 2027, from 87.0% on Wednesday, and the first non-trivial cut probability, 1.2% at 3.25-3.50%, still appears only at December 2027, down from 1.4%. A strip that raises the modal number of hikes while lowering the odds of the first cut is not shifting a distribution; it is extending one tail and deleting the other.

(c) Year-end probability ladders

Year-end 2026 — the 9 December meeting.

Cumulative above the current range: 73.6%. Cumulative below: 0.0%. Sum: 99.9%.

Year-end 2027 — the 8 December meeting.

Cumulative above the current range: 85.0%. Cumulative below: 1.2%. Sum: 100.0%.

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 the vendor does not publish are shown as 0.0% only where the vendor's own card omits the range entirely, which under CME methodology means a probability below the rounding floor. The December 2026 ladder sums 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 are as of 26 August 2026, not the 27 August close. Same-day direction is cross-checked against the cash-market proxies underneath.

CDX — the full six-step ladder was worked again and is reported so the gap stays auditable. (1) Bloomberg in Chrome: the /markets and /markets/rates-bonds boards were rendered and carry global government yields and the Bloomberg Fixed Income Indices, but no CDX line; the day's credit coverage on the site was issuance and private-credit reporting, none of it quoting an index level. (2) WSJ Market Data bonds page: rendered, and its credit tables again did not populate a CDX row — the page's lead item was the seven-year auction. (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 and methodology documentation, not the daily running spread. (4) FT Markets Data and Reuters credit wraps returned no dated CDX quote for 27 August. (5) TradingView and Barchart symbol searches for CDX resolve to the Simplify High Yield ETF, an unrelated instrument. (6) Cash-market proxies, labelled as proxies: HYG closed at $79.87, -0.04%, and LQD at $106.73, -0.05%, both on 27 August per Bloomberg. Two ETFs moving four and five hundredths of a percent on a day the coupon curve rose a uniform basis point says credit took the duration and gave back nothing on spread — consistent with the tightening FRED picture through 26 August. 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 it has done in three weeks. CCC and lower tightened 8 bp to 1,031, the largest single-session tightening in the series this month, taking the CCC-minus-HY differential to 764 bp from 769 — the first narrowing of the reporting window. HY tightened 3 bp to 267, now 14 bp tighter than where 2026 started, and IG tightened 1 bp to 80, leaving it just 1 bp wider on the year. On the week the picture is still split: HY -6 bp, IG -1 bp, CCC +1 bp. One session does not undo +146 bp of year-to-date CCC widening against +1 bp for IG, but the direction of the tail is the single most informative variable in this section, and on 26 August it turned.

(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 26 August 2026 effective date.

The squeeze lasted exactly one day. SOFR printed a basis point through the 3.65% IORB on the 25 August effective date; on the 26th it fell 2 bp to 3.64%, back a basis point below it, with the 99th percentile down to 3.72% from 3.74%, volume $57bn lighter at $2,859bn, and both tri-party and broad general collateral 2 bp lower at 3.62%. The facility emptied in step: overnight reverse repo take-up fell to $456m on 27 August from $702m on the 26th, though it remains more than double the $200m of 21 August. What did not reverse is the structural drain — reserve balances fell $10.4bn to $2.9249tn for the week ended 26 August, the fourth consecutive weekly decline and now $68bn below the 5 August peak of $2.9933tn.

The bill strip is the corroborating tell and it disagrees with the overnight market. The 2-month bill rose a basis point to 3.81% and is 2 bp cheaper on the week, the 1.5-month rose to 3.79%, also 2 bp cheaper on the week, while the 3-month fell a basis point to 3.84% and is 3 bp richer on the week and the 4-month and 6-month did not move at all (Section 6, off-table tenors). Overnight funding relaxed and the two bills that straddle 31 August month-end cheapened anyway. That is a calendar effect being priced in term while the overnight market takes a day off, which is the ordinary sequence two sessions before a turn — and it arrives with a 9 September bills-for-bonds operation behind it. Watch whether SOFR goes back through IORB on the 31st.

(c) Rates volatility and swap spreads

The MOVE series failed its own consistency check for a second consecutive session. The Investing.com page marks 69.86, +0.42 (+0.60%) with an opening print of 69.44 — internally coherent — alongside a stated "previous close" of 95.74, which is 26 points away from the day's range and cannot be reconciled with the rest of the page. The level is published with its vintage and the change is withheld. What can be said without the vendor is that the level is roughly two points below the ≈71.9 this report published on Wednesday, and that rates volatility falling while VIX fell 4.60% to a 14 handle is a two-market statement of calm going into the single largest scheduled communication event of the quarter. The swap-spread basis is unchanged in substance: Treasuries have outperformed equivalent-maturity swaps since the buyback announcement, compressing the 30-year spread to a six-month extreme, with Fed researchers putting hedge-fund swap-spread positions at a record $305bn last year against under $50bn in 2022. That basis now faces a Warsh keynote in which, per Bloomberg, three of the largest houses on the Street are explicitly positioned for a long-end rally.

(d) Issuance, leveraged loans and private credit

Levels from the Investing.com boards taken after the 5:00 p.m. ET close, cross-checked against TradingEconomics. Basis warning: both vendors' daily boundaries had rolled to 28 August by the time of this pull, so their published %Chg columns measure only the new Asian session and are not reproduced. The change column below is computed against the prior edition's levels for the same vendor, and therefore spans approximately twenty-eight hours rather than a clean session. 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 dollar did nothing on a day the equity market did everything, and that is the second consecutive session it has failed a test. DXY closed 99.11, up four hundredths of a percent, on a session in which the S&P rose 0.72%, the Nasdaq 1.57%, two Fed presidents argued publicly for higher rates and the December hike probability rose to 73.6%. Société Générale's Kit Juckes framed the standing problem earlier in the week: dollar strength "will only return when (if) domestic data turn stronger and put pressure on the Fed to tighten." The pressure to tighten is visibly building in the strip — and the dollar index has now moved a cumulative 26 basis points across two sessions in which it built.

The Asian crosses inverted, and the mechanism is a central bank. USD/KRW fell 0.48% to 1378.78 and USD/TWD 0.55% to 31.68 — the two largest moves on the board, and both in the direction of local-currency strength — after the Bank of Korea raised its policy rate 25 bp to 3%, its second consecutive hike, on core inflation at 2.6%. Twenty-four hours earlier this report flagged the opposite configuration: a won that weakened on the Kospi's best session. The distinction matters for how the region's semiconductor rally is being funded. Wednesday's tape was domestic leverage with hedged foreign flow; Thursday's is a rate differential doing the work, and it is worth noting that the won strengthened on the day the Kospi's own gain was 1.53% and then the Kospi gave three-quarters of it back on Friday. Currency first, equity second, is the ordering that usually survives.

The renminbi keeps grinding and the yen keeps not caring about its own bond market. USD/CNH closed 6.7185, a fresh closing extreme that takes out Wednesday's 6.7216 without taking out Wednesday's 6.7151 intraday print — a currency being walked, not run. And USD/JPY rose 0.10% to 159.440 on the day Japanese 10-year yields rose 5 bp to 2.91%, the largest sovereign move on Bloomberg's global board. A bond market that sells off five basis points while its currency weakens is a bond market repricing domestic inflation without attracting foreign capital, and it is the cleanest reminder available that the carry trade is still the dominant flow in the pair. The franc firmed 0.08% on a risk-on tape, which is the exact inverse of Wednesday's behaviour and consistent with the return of the debasement bid in silver and bitcoin (Section 11).

Settlement basis, stated: all rows are the Investing.com per-contract close taken from the instrument pages after the 27 August settlement, read during the Friday electronic session as each contract's "previous close". Gold and WTI are independently corroborated against Bloomberg's GC1 and CL1 marks; the Investing.com front-month historical table disagrees with its own quote page on gold and on silver, and that reconciliation is worked in Data Notes. All contract months are named where the vendor names them. Day changes are computed against the prior edition's published 26 August settles.

\*YTD figures marked with an asterisk are TradingEconomics spot year-to-date returns, not futures returns on the contracts quoted above, and are shown for direction and magnitude only. The daily settles, changes and percentage moves are on the futures basis named in each row. Weekly columns are TradingEconomics spot weekly changes on the same caveat, and were read after the vendor's daily boundary had rolled to 28 August. Mixing the two would be a basis error; they are presented in separate columns for exactly that reason.

The crack spreads went again, and the gasoline crack is now $5.69 above where it was on Tuesday. On the same 42-gallon basis this report has used all month:

The differential moved $2.33 in the direction it moved $3.34 the day before, and it did so with the underlying barrel rising rather than falling — which is the part that changes the read. For three sessions the gasoline crack expanded because crude was collapsing faster than the product; on Thursday crude rallied 2.10% and RBOB rallied 3.29% anyway. A product that outperforms on the way down and on the way up is not being driven by the barrel at all; it is being driven by its own balance. The seasonal caveat remains and is now the whole risk: this is the last week of August, and the September gasoline calendar works structurally against the long leg.

The metals split, and silver took the debasement bid on its own. Silver settled $69.431, up 1.92%, against gold $4,664.00, up 0.32% and copper $6.5915, down 0.18% — the first session of the week in which silver outperformed both its monetary and its industrial reference at the same time. The gold-silver ratio compressed to 4,664.00 / 69.431 = 67.18 from 68.25, a third consecutive session of compression and, at 1.07 points, the largest of the three. Alongside it bitcoin retook $80,000, trading $80,002, +1.63%, with Bloomberg reporting more than $2.6bn of inflows into U.S. bitcoin ETFs over eight trading sessions and a Coinbase-versus-Binance premium for the first time in roughly three months. Bloomberg's own analysis is worth quoting against the consensus framing: the bitcoin advance "came as technology shares surged after Nvidia Corp.'s outlook revived confidence in the artificial-intelligence trade," which "cuts against the idea that Bitcoin needed investors to rotate out of AI winners to find its next leg higher." Both are the same retail risk appetite, not a rotation between them — and if that is right, the debasement trade and the AI trade share a single funding source and will unwind together.

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 made a new high on the exact mechanism it was underwritten for; hold the half, raise the stop

Mark first, honestly. The book is long ZQU6 (September 2026) against short ZQZ6 (December 2026), DV01-matched one-for-one at $41.67 per basis point per contract, entered a week ago at 96.325 / 96.160 for a spread of 16.5 bp, marked 18.5 bp, then 20.5 bp where a second quarter came off and the stop went to 18.0 bp, then 18.5 bp, then 19.0 bp on Wednesday. Thursday's mark: ZQU6 96.325, ZQZ6 96.130 — a spread of 19.5 bp. That is +0.5 bp on the day, worth +$20.84 per contract pair on the retained half, and it leaves the trade +3.0 bp from entry, its best mark since Monday, with 1.5 bp of cushion above the stop.

Why it worked, and it is the mechanism, not luck. The trade needs December to absorb more of every repricing than September. Thursday delivered the purest version yet: ZQU6 did not move at all — 96.325 on Wednesday, 96.325 on Thursday — while ZQZ6 fell half a tick to 96.130 as December's cumulative hike rose from 71.7% to 73.6%. The September contract was inert through a claims beat and two hawkish Fed presidents; every basis point of the day's repricing landed in December and 2027, which is precisely the asymmetry this spread is long.

The modal path, the base case and the tails. Modal path: hold on 16 September (Investing.com 65.6%, CME 66.1%, ease 0.0%); hold still modal on 28 October at 47.5% against 52.5% cumulative above; one 25 bp hike delivered by 9 December, modal 3.75-4.00% at 45.0%, cumulative 73.6%; and then the genuinely new fact — the modal 2027 range steps up to 4.00-4.25% from June through October, with the terminal geometry troughing at 95.920 in September 2027, one meeting later and 1.5 bp lower than Wednesday. Base case: the strip is pricing the first hike later and the second hike more, which is the exact configuration in which a September-versus-December calendar spread pays and keeps paying. Tail one, dovish: a Warsh keynote that treats inflation as solved and pulls December back under 65%, which compresses the spread through the stop. Tail two, hawkish but adverse: a keynote so explicit that the market pulls the hike forward into September, which compresses it from the other side — note that this is the tail two sitting Fed presidents were arguing for on Thursday, and the market ignored them. Practical implication: the trade is three basis points wide after six sessions and its stop has never been tested. Raise the stop to 18.5 bp, hold the half, do not add.

Expression: long ZQU6 / short ZQZ6, DV01-matched one-for-one, half size. Catalyst: Warsh 8/28 ~10:00; ISM manufacturing and JOLTS 9/1; payrolls 9/4 08:30; the first buyback operation 9 September; the 16 September FOMC. Invalidation: the spread through 18.5 bp; or the September cumulative hike printing above 50% on either vendor; or any 2026 meeting showing a non-zero cut probability. Sizing: one-for-one DV01 at $41.67 per basis point per pair; risks 1.0 bp to make a further 2.0.

2. Long the 20-year against the 30-year, on the November refunding — hold, and Friday is the test

Expression: long the 20-year bond against short the 30-year, DV01-neutral, half size. Mark: 20s30s at 1 bp, unchanged on the day — the 30-year at 5.19% and the 20-year at 5.18%, both a basis point higher — and 2 bp flatter on the week, which is a small loss on the position. What changed: Bloomberg reported JPMorgan, Apollo and Morgan Stanley all arguing that a credibly hawkish Warsh would buy the long end, with 30-year yields having hit their highest since 2007 last week. That is a directional long-end call, not a curve call, and it cuts both ways for this trade: a rally led by the 30-year flattens 20s30s further, while Alyce Andres's framing — that "the composition of any bond selloff" is what matters — is exactly the variable this position is expressing. The issuance thesis is unchanged: Citi has pushed its forecast for larger auctions to 2028 and raised the tail risk that Treasury eliminates the 20-year, and the seven-year auction clearing near a two-year high yield says the middle of the curve is where the concession is being paid. Catalyst: Warsh Friday, and specifically the composition of the move that follows; 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.

3. Protection on the CCC cohort funded in IG — cut to a quarter; the differential just made its first new low

Expression: long CCC-exposed credit protection (or short a levered-loan / CCC-heavy vehicle) against long IG cash. Mark: CCC 1,031 bp, -8 bp; IG 80 bp, -1 bp on the 26 August FRED update, taking the CCC-minus-HY differential to 764 bp from 769 — the first narrowing of the reporting window and a 5 bp loss on the position. The honest reading: this trade has worked for three weeks on a single mechanism, that the tail widens while the index does not. On 26 August the tail tightened eight basis points, the most in a session this month, and it tightened more than the index. That is not noise at that magnitude; it is the first evidence against the thesis. What is still intact: on the year CCC +146 bp against IG +1 bp, and on the week CCC +1 bp against IG -1 bp, so the structural dispersion has not closed — only the momentum has. Action: cut from a half to a quarter. Keep the option on month-end and on Warsh; stop paying full carry for a thesis that just took its first counter-signal. Catalyst: 31 August month-end; the September IG calendar clearing; Broadcom on 2 September, because the AI capex line is what the calendar funds. Invalidation, tightened: the differential back through 750 bp (from 735 bp), or IG widening beyond 85 bp. Sizing: a quarter.

4. On-balance-sheet AI funding against off-balance-sheet AI funding — close it

Expression: long customer-funded suppliers against short self-funded ones, cut to a quarter into Marvell's print. Mark: Marvell -1.49% to $241.45 against Broadcom +4.49% to $371.54 — a 5.98-point loss, the worst single session of the trade's life, and then Marvell fell a further 2% after the close on Q3 EPS guidance of $1.05-$1.10 against a $1.08 consensus, despite revenue guidance of $3.15bn against $3.04bn. Action: close it. The honest reading, and it is the reason rather than the excuse: the quarter was carried into the print deliberately, and the print was the right kind — revenue guidance well above consensus on custom AI ASIC demand, which is the thesis in one line — and the stock fell anyway while the self-funded leg rose 4.49% on nothing but sympathy. When the fundamental evidence arrives, confirms the thesis, and the position loses six points on the day and two more after the bell, the market is telling you the spread is not trading on the variable you underwrote. Total damage from the quarter carried through the event is roughly 8 points; the trade still made money over its life on the two sessions before. What replaces it: nothing in this pair. See idea 8.

5. Long silver against short gold — hold, small; three sessions of compression and the largest yet

Expression: long Comex silver against short Comex December gold, notional-matched, small. Mark: silver +1.92% against gold +0.32% is a 1.60-point gain, the best session of the trade, and the gold-silver ratio compressed to 67.18 from 68.25, a third consecutive session of compression and the largest of the three. The thesis is now confirmed twice over: it worked on a day both legs fell (Wednesday) and on a day both legs rose (Thursday), which is what distinguishes a ratio trade from a levered long. The catalyst identified on Wednesday — that silver is the only asset levered to both the industrial and the monetary leg while participating in neither — resolved in exactly the predicted direction, with bitcoin retaking $80,000 on the same tape. Copper at $6.5915 is comfortably above the $6.40 invalidation despite a 0.18% decline. Action: hold, small, do not add into the keynote. The position has made 1.07 ratio points in three sessions on a thesis about monetary debasement, and the person who can most quickly invalidate that thesis speaks at 10:00 on Friday. Catalyst: Warsh Friday ~10:00; the copper tape into the 8 September Canadian tariff date. Invalidation, unchanged: the ratio through 71, or copper closing below $6.40. Sizing: small, unchanged.

6. Long the gasoline crack against the distillate crack — take another third off; the trade has made its case

Expression: long the RBOB crack against short the heating-oil crack, both on the standard 42-gallon basis against front-month WTI, barrel-for-barrel, currently two-thirds of a small position. Mark: the gasoline crack rose $2.82 to $58.61 while the distillate crack rose 49 cents to $95.90 — a $2.33 move in the differential, on top of Wednesday's $3.34, for a two-session total of $5.67. Why it matters more than Wednesday: the first session worked because crude collapsed and the product held. This one worked because crude rallied 2.10% and RBOB rallied 3.29% anyway, which removes the "it is just the barrel" explanation entirely. Action: take another third off, hold a third. The pair has made roughly its expected quarterly range in two sessions and now faces the September gasoline calendar with a full geopolitical premium re-entering the barrel via Goldman's Gulf-export note. Catalyst: enforcement detail on the Iran sanctions programme; weekly EIA product inventories; the Labor Day driving-season roll-off. Invalidation, unchanged: a re-escalation headline that adds more than $3 to the distillate crack in a single session, or the gasoline crack falling below $50. Sizing: a third of a small position.

7. New — short the AI-halo basket against long the name that actually reported

Expression: short an equal-weighted basket of Synopsys, Palo Alto Networks, ServiceNow, Fortinet and Adobe against long Nvidia, beta-adjusted, quarter size. Thesis: on Thursday those five rose 13.39%, 12.83%, 10.04%, 9.67% and 5.73% without reporting anything, on the read-through from a Salesforce and CrowdStrike print and an Nvidia guide. The four S&P 500 names that did report into the same tape — Autodesk, Workday, Marvell and, on Wednesday, HP — were sold on their own numbers, by 6%, 5%, 2% and 9% respectively, and Rubrik was sold 10% on a beat-and-raise. The halo is being paid for in advance and the evidence is being sold on arrival. Nvidia is the long leg because it is the one name whose numbers are already known and whose 8.74% move is the only one in the group backed by a $108bn ±2% guide and a 70% fiscal-2028 growth forecast. Catalyst: Palo Alto Networks reports after the close on Tuesday 1 September — the largest name in the short basket, with a 12.83% pre-print gain to defend; Broadcom, HPE and NetApp Wednesday 2 September. Invalidation: Nvidia closing below $209.66, its pre-print level, which would mean the long leg's own evidence has been rejected; or the basket outperforming Nvidia by more than 6 percentage points cumulatively before 1 September. Sizing: a quarter, and note the obvious risk — this is a short of five names that just gapped, into a semiconductor-tariff headline that would hurt the long leg more than the short one.

Prior closes, marked forward. The short-utilities-versus-S&P pair, closed on Wednesday at a loss, would have made 1.42 points on Thursday: utilities fell 0.70% while the index rose 0.72%. That is the fourth consecutive edition in which a correctly-executed exit cost money in the following session, and it is being recorded rather than re-entered, for the reason given when it was closed — the stops were set on the long end and the variable that mattered was the discount rate on regulated cash flow, which is not the same thing. The equal-weight-versus-cap-weight pair, closed two sessions ago, would have lost heavily: 141 advancers against 351 decliners under a 0.72% index is the single worst possible tape for a long-median-stock position, which is the vindication of that exit rather than a regret. The memory-versus-platform pair, stopped a week ago, would have lost 3.1 points (MU -0.32%, WDC -1.47%, STX +0.10%, SNDK -0.96% against AAPL +0.36%, MSFT +1.75%, GOOGL -0.39%, AMZN -1.54%).

The vol note. VIX closed at 14.51, down 4.60%, a 14 handle for the first time this month, on the eve of the first Jackson Hole keynote by a new Federal Reserve Chairman whose communication style Bloomberg describes as having "baffled investors." A 14.51 handle asks for roughly a 0.91% daily move. Consider what is in front of it: Warsh at ~10:00 Friday, with three of the largest houses on the Street publicly positioned for a long-end rally and Apollo's chief economist warning that the absence of framework guidance means "a much higher move in long rates"; 31 August month-end into a funding market where the two-month bill is 2 bp cheaper on the week; ISM, JOLTS, ADP and ISM services across 1-3 September; and August payrolls on 4 September. Rates volatility, on a delayed and internally inconsistent vendor series, sits near 69.9 — roughly two points below Wednesday. Both volatility markets are pricing calm into the single largest scheduled communication event of the quarter, and the equity index that they are pricing has just delivered a 0.72% gain on 141 advancing stocks. Own gamma dated 28 August through 4 September, and prefer it in the index over single names, because the single-name premium has already been paid twice this week and given back both times.

Crowded consensuses to stress-test, with the numbers.

The two-sided geopolitical tape. Escalation: the USS Theodore Roosevelt deploys to the Middle East for at least seven months as the war enters its sixth; Trump told Al Jazeera the U.S. is "achieving a very big victory" but that the war "will continue for as long as necessary," with no timetable for talks; and Goldman puts Gulf oil exports at two-thirds of pre-war levels, which is what took crude up 2.10%. Canada's retaliation on $20bn of U.S. goods lands 8 September, and Washington is weighing semiconductor tariffs that would reach data-centre servers. De-escalation: Iran and Oman are finalising a temporary shipping route through the Strait of Hormuz, with Qatar's foreign minister in Tehran and Pakistan's army chief there two days earlier; ING's caveat is the one to hold — normalisation requires the U.S. to "lift its blockade on Iranian ports and ease sanctions." The market spent three sessions trading the de-escalation and one trading the escalation, and the escalation was worth $1.72 on the barrel.

Structural watch items. SOFR back below IORB after one day through it, with reverse repo take-up down to $456m from $702m but the 2-month bill 2 bp cheaper on the week into a 31 August month-end; reserve balances at $2.9249tn, $68bn below the 5 August peak and falling for a fourth week; seven-year auction yields near a two-year high in the one part of the curve the buyback programme has not addressed; a record ~$1.4tn of 2026 IG issuance funding a capex cycle that is now also a tariff target; the Bank of Korea's second consecutive hike to 3% with core inflation at 2.6%; Japanese 10-year yields +5 bp to 2.91% on a day the yen weakened; and USD/CNH setting a fresh closing extreme for a second consecutive session.

Full Data Notes & Conflicts, the Overnight / Asia & Europe read-through and the categorised source links are in the companion file US_CrossAsset_Daily_2026-08-27_DataNotes.txt, alongside the canonical Markdown report of record.