Closing Briefing — Friday, August 14, 2026

1. The strip split: the front end priced hikes out, the back end priced them in, on the same release. July advance retail sales fell 0.6% against a +0.1% consensus, and the September fed funds contract barely moved — CME's ZQU6 mid at 96.3325 against 96.3275 on Thursday, taking the September hold to 67.5% from a settled 66.1%. Meanwhile every 2027 contract fell: January 96.140 → 96.130, June 96.020 → 95.995, July 96.015 → 95.990, December 96.060 → 96.025. The 2027 year-end cumulative probability of sitting above 3.50–3.75% rose from 73.2% to 77.2% in one session and the cumulative-below fell from 4.4% to 2.8% (§8). That is the exact reversal of Thursday, and it has a name: a market that believes a weak consumer delays the hike but raises the eventual destination because the delay lets inflation re-accelerate. Bloomberg quoted Sit Investment's Bryce Doty saying "the curve-steepening trade is alive and well." The cash curve agreed to the basis point — 2s10s to 51 bp, its widest since May, 3M10Y to 82 bp, 2s30s to 108 bp (§6). Forward catalyst: the 26 August PCE deflator, and Jackson Hole positioning; nothing on the U.S. calendar before Monday 8:30.

2. Thursday's software re-rate was given back inside one session, and the week still finished as software's best since January. Twenty-four hours after nineteen software names rose 2.9–9.5% on a reported Silver Lake approach to Workday, the same names reversed: GoDaddy −5.59%, Intuit −3.53%, Gen Digital −3.69%, Fortinet −3.28%, Palo Alto Networks −2.95%, Fidelity National Information −2.75%, Paychex −2.70%, PTC −2.66%, Salesforce −2.56%, ServiceNow −2.55%, Equifax −2.55%, Paycom −2.46%, Autodesk −2.42%, Adobe −2.37%, Verisk −2.30%, Fair Isaac −2.24%, CoStar −2.03%, Fiserv −2.00%, Oracle −3.77%. Workday itself fell more than 5% and still finished the week up about 9% (CNBC). Bloomberg's own wrap notes an index of software stocks was on track for its best week since January — so the week-level re-rate survived and the day-level rumour premium did not. That distinction is the whole trade. Forward catalyst: confirmation or denial of the Silver Lake talks; a second sponsor bid would convert the read-across from rumour to regime.

3. Broadcom lost 5.95% because of a credit action, and the private-credit sponsor on the other side of it fell 3.61% with it. Bank of America's Tom Curcuruto cut Broadcom's issuer and bond ratings, with the stated concern the AI XPV financing platform built jointly with Apollo and Blackstone; separately, reports confirmed an actively exploited VMware vCenter vulnerability. The stock closed $392.95, −5.95%, a $24.87 decline, and Blackstone closed −3.61% at $143.93. This is the cleanest instance yet of the theme CNBC ran the same morning — "AI's infrastructure boom is getting more leveraged — and harder to track" — showing up as an equity price. The transmission runs credit → equity, not the other way, which is precisely the sequence a cycle turn produces. Forward catalyst: Broadcom's 2 September print, and whether any other hyperscaler-adjacent name is put on negative watch for a financing structure rather than for earnings.

4. Energy was the only sector that worked, and the refiners made all-time highs and could not hold them. Finviz Energy +1.19% against Healthcare −0.48% and Technology −0.46%; CNBC's GICS tally had energy +1.5% leading and tech and healthcare −0.6% lagging. The catalyst was geopolitical and explicit: Defence Secretary Pete Hegseth said the U.S. Navy can keep the blockade of Iranian ports going "indefinitely", and Bessent promised measures "that have never been seen on Iran." WTI settled $82.40, +1.42%, Brent $88.60, +1.76%, and Bloomberg marked energy equities at their highest since March. Halliburton +4.78%, SLB +3.28%, Devon +3.28%, Targa +3.34%, Williams +2.96%, ONEOK +2.53%, Kinder Morgan +2.31%, Baker Hughes +2.18%, ConocoPhillips +1.81%, Diamondback +1.49%, Occidental +1.14%, Chevron +1.17%, Exxon +0.94%. But Marathon Petroleum, Phillips 66 and Valero all printed all-time highs intraday and closed −0.27%, +0.43% and −0.36% — the producers and services led and the refiners did not, the first such session in eight (§11, §12). Forward catalyst: any Hormuz or blockade headline over the weekend, which is a gap risk in both directions.

5. The consumer data was bad and the consumer calendar starts on Tuesday. Retail sales −0.6% with autos −1.8% and gasoline stations −0.9% but +16.2% year-on-year, which is the point — Jefferies' Thomas Simons told Bloomberg the optimism was "dashed by the reduction in purchasing power sucked up by higher gasoline prices." Michigan sentiment fell to 51.0 from 55.2, an 8% monthly decline that ended two months of improvement, with Republicans showing the sharpest drop and only 8% of consumers expecting income growth to beat inflation, down from 18% in December 2024. Against that, the SPDR S&P Retail ETF fell about 2% on the week and Home Depot, Lowe's, Target and Walmart all report inside three sessions (§5). Bloomberg's earnings desk notes consumer-discretionary 2026 earnings growth expectations have been marked up to 31% from 12% in May. A 31% growth expectation and a −0.6% sales print are not obviously compatible. Forward catalyst: Home Depot 8/18 BMO.

6. Volatility would not bid on a consumer miss, and that is the most crowded thing on the board. The VIX closed 14.25, −2.60%, the lowest since 29 December, on a session when the index fell, the largest release of the week missed by 0.7 points and the 30-year rose 4 bp. Realised is not the problem — the S&P's last four sessions are +0.65%, −0.17% and a 0.44% high-low range on Friday — but implied at 14.25 prices a 0.90% daily move while single names moved 6% (AMD), 6% (Broadcom) and 5% (Applied Materials) in the same session. Index vol is cheap because dispersion is doing the work. Forward catalyst: the weekend geopolitical tape, which is the one thing a Friday VIX at a seven-month low is structurally short.

7. AMD raised $4.75bn, and the IG market's order book fell apart while it did. AMD closed +6.46% at $514.21 on its largest-ever dollar bond offering — $4.75bn across four IG tranches, three to ten years — priced into a week in which, per Bloomberg, investors pulled about 36% of their initial orders for high-grade deals after final pricing was squeezed, twice the prior week's rate and against a 22% average for the year, with Thursday's attrition hitting 45%, a record for 2026. AMD's own longest tranche saw orders more than halve. $136bn priced in a fortnight, August on track for its busiest ever, and next week's expected supply is only about $20bn (§9d). An equity that rallies 6.5% on a bond deal whose book halved is a market pricing the use of proceeds and ignoring the cost of them. Forward catalyst: the post-Labor Day supply burst, and whether the 36% attrition rate persists at $20bn a week.

8. Reddit joins the index on Tuesday and rose 12% on the way in, still 20%+ lower on the year. S&P Dow Jones confirmed Reddit replaces AvalonBay Communities on 18 August; the stock rose about 12% on Friday after roughly 11% after hours on Thursday, and remains down more than 20% year to date (CNBC). Loop Capital's Rob Sanderson published a $250 price target, 58% upside from Thursday's close, arguing the marketing community "is just beginning to understand the power of Reddit" and that its content is "integral to LLMs." Forward catalyst: the 18 August rebalance flow, which lands on the same morning as the Home Depot print.

Six green, five red, and a dispersion of 1.67 points — barely half Thursday's 3.09 and the narrowest of the week. CNBC's GICS-based tally reads six of eleven higher with energy +1.5% leading and tech and healthcare −0.6% lagging; Finviz ranks the same three groups at the same ends, which is the cross-check that matters. Energy +1.19% is the whole session in one row, and its composition inverted. Halliburton +4.78%, SLB +3.28%, Devon +3.28%, Targa +3.34%, Williams +2.96%, ONEOK +2.53%, Kinder Morgan +2.31%, Baker Hughes +2.18%, ConocoPhillips +1.81%, Diamondback +1.49%, Occidental +1.14%, Chevron +1.17%, Exxon +0.94% — producers and oilfield services led — against Marathon Petroleum −0.27%, Valero −0.36% and Phillips 66 +0.43%, all three of which set all-time highs intraday first. Seven consecutive sessions of refiners over producers ended on the eighth, on the session the barrel rose. Technology −0.46% is the mirror image of Thursday and it is entirely software. Twenty-four hours ago GoDaddy +9.45%, Fiserv +7.64%, Intuit +7.03% and Fair Isaac +6.33% led the group; on Friday the same four printed −5.59%, −2.00%, −3.53% and −2.24%, with Oracle −3.77%, Broadcom −5.95%, Applied Materials −5.12%, CrowdStrike −3.82%, HP −3.80%, Fortinet −3.28%, KLA −2.69% and Palantir −2.78% alongside. The offsets were memory and analog: AMD +6.46%, Seagate +5.65%, Corning +4.70%, Western Digital +4.41%, Monolithic Power +2.98%, Micron +2.42%, Synopsys +2.37%, Texas Instruments +2.27%, Analog Devices +2.16%, Microchip +1.89%, ON Semiconductor +1.35%. Exactly the dispersion inside semis that this report flagged on Thursday, with the signs flipped on logic and unchanged on memory. Healthcare −0.48% was the worst group on Eli Lilly −2.39%, Zoetis −2.70% and IDEXX −2.80%, against Molina +3.27%, Align +2.87% and Solventum +2.46%. Consumer Cyclical −0.18% on a retail-sales miss of seven tenths is the number to notice — the group barely reacted, and it is already −2.39% year to date, the second worst on the board. Source: Finviz Groups via rendered page (?g=sector&v=140&o=name); Finviz buckets are not official GICS/S&P sector indices, which is why Alphabet, Meta and Netflix sit in Communication Services and Amazon and Home Depot in Consumer Cyclical.

Reconciliation, and the Industrials break has healed. All eleven groups were checked against Thursday's published YTD compounded by Friday's 1-day move. All eleven reconcile to within 0.05 point — four exactly: Consumer Cyclical 0.9779 × 0.9982 = −2.39% vs −2.39%; Energy 1.3388 × 1.0119 = +35.47% vs +35.47%; Healthcare 1.0827 × 0.9952 = +7.75% vs +7.75%; and Communication Services 0.9940 × 1.0008 = −0.52% vs −0.53% at a rounding tick. Basic Materials 1.1460 × 1.0060 = +15.29% vs 15.30%; Consumer Defensive 1.0887 × 0.9998 = +8.85% vs 8.84%; Financial 1.0914 × 0.9994 = +9.07% vs 9.06%; Real Estate 1.1108 × 1.0022 = +11.32% vs 11.33%; Technology 1.2693 × 0.9954 = +26.35% vs 26.34%; Utilities 1.0237 × 1.0035 = +2.73% vs 2.70%. Industrials, which failed by 1.31 points on Thursday and by a similar margin on Wednesday, reconciles this session: 1.1691 × 1.0023 = +17.18% against the +17.23% shown, a 0.05-point gap and the largest of the eleven. The two-day break was therefore a constituent revision that has now settled, not a persistent data fault, and the Industrials YTD is comparable again from this edition forward.

Up

Copart (CPRT) +7.55% to $31.61 — the largest S&P 500 gain, and no fresh company announcement was identified in the reviewed material; the move reads as a rebound after a weak stretch, with the stock having lagged into it. AMD +6.46% to $514.21 on the pricing of its largest-ever dollar bond offering, $4.75bn across four IG tranches of three to ten years, funding an AI and server expansion (Bloomberg; §9d). Fox Corp B +5.70% and Fox A +5.48%. Seagate +5.65% to $973.44 and Western Digital +4.41% to $508.80, the memory and storage complex extending Thursday's leadership. NRG +5.42% to $126.24. Halliburton +4.78% to $34.41, the best of the energy complex on the blockade headlines. Corning +4.70% to $165.99, recovering the whole of Thursday's −5.32%. Albemarle +4.35%. Ford +3.42% to $14.37 on a session the auto sales line fell 1.8%. Targa +3.34%, Devon +3.28%, SLB +3.28%, Williams +2.96%, ONEOK +2.53%, Kinder Morgan +2.31%, Baker Hughes +2.18%, ConocoPhillips +1.81% to $126.78, Diamondback +1.49%, Chevron +1.17% to $200.01, Occidental +1.14%, Exxon +0.94% to $160.10. Molina +3.27%, Tyson +3.16%, Newmont +3.13% to $117.76, Celanese +3.07%, United Rentals +3.02%, Monolithic Power +2.98%, Align +2.87%, Chipotle +2.73%, Equinix +2.64%, Solventum +2.46%, Micron +2.42% to $972.78, Synopsys +2.37%, Texas Instruments +2.27%, Analog Devices +2.16% into its own print on 8/19, Constellation Brands +2.08%, Hilton +2.02%, Microchip +1.89%, ON Semiconductor +1.35%, GE Vernova +1.32%, Charles Schwab +1.23% to an all-time high, Vistra +1.18%.

PayPal +1.77% to $61.66 on a WSJ report that Stripe and Advent International have bid $60.50 a share as talks advance, valuing the company near $53bn — note the close is $1.16 above the reported bid, which is the market pricing a bump rather than a deal. Nine S&P 500 names set fresh 52-week highs, seven of them all-time: Marathon Petroleum, Phillips 66, Valero, Ameriprise Financial, Citizens Financial, Charles Schwab and CrowdStrike (CNBC).

Outside the index. Reddit +12% ahead of its 18 August S&P 500 inclusion, with Loop Capital initiating a $250 target, 58% upside from Thursday's close (Rob Sanderson). Sandisk +7.41% to $1,641.28 after JPMorgan's Harlan Sur upgraded to Overweight with a $2,250 target, 47% upside from Thursday's close, calling the company "uniquely positioned to capture the ongoing structural inflection in NAND demand"; the stock is up 544% year to date. Nebius +8.88%. Unusual Machines +24%, Red Cat +8%, AeroVironment +2% after the administration imposed a 100% tariff on imported unmanned aircraft systems and components (Bloomberg). Intuitive Machines +10% after Stifel upgraded to Buy from Hold with a $26 target, with backlog at $1.55bn. Wayfair +4% after Bernstein upgraded to Outperform from Market Perform, noting "in a furniture market that is not growing, Wayfair is putting up [high single-digit] revenue growth in the US."

Down

Broadcom −5.95% to $392.95 — the largest single-name drag on the index. Bank of America's Tom Curcuruto cut the issuer and bond ratings, citing the AI XPV financing platform built with Apollo and Blackstone, and reports confirmed an actively exploited VMware vCenter flaw. Guidance implies AI accounts for roughly 72% of expected sequential revenue growth next quarter; earnings are 2 September. GoDaddy −5.59% to $94.88, the sharpest software reversal after Thursday's +9.45%. Applied Materials −5.12% to $507.18 — the post-print verdict, after adjusted $3.50 EPS on $9.12bn revenue and semiconductor-systems sales of $7.04bn against a $6.96bn FactSet consensus; cash flow from operations of $3.04bn beat a $2.13bn estimate and the stock fell anyway, on top of Thursday's −2.48%. Two consecutive sessions of selling a beat. CrowdStrike −3.82% to $216.92 after an all-time high at $227.50 — a 4.65% high-to-close fade. HP −3.80%, Oracle −3.77%, Gen Digital −3.69%, Blackstone −3.61% to $143.93 on the same XPV read-across as Broadcom, Intuit −3.53%, Fortinet −3.28%, F5 −3.08%, Palo Alto Networks −2.95%, Pool −2.86%, IDEXX −2.80%, Norwegian Cruise −2.79%, Palantir −2.78%, Fidelity National Information −2.75%, Paychex −2.70%, Zoetis −2.70%, KLA −2.69%, PTC −2.66%, Salesforce −2.56%, Equifax −2.55%, ServiceNow −2.55%, Paycom −2.46%, Autodesk −2.42%, Eli Lilly −2.39%, Adobe −2.37%, Arista −2.36%, Verisk −2.30%, Fair Isaac −2.24%, Delta −2.16%, CoStar −2.03%, Fiserv −2.00%, Intel −1.97%, Yum! Brands −1.82%, Royal Caribbean −1.57%, Lam Research −1.36%, Nucor −1.24%, Carnival −1.07%.

Cisco −1.58% to $111.68 after HSBC's Abhishek Shukla cut to Hold from Buy and the target to $120 from $137, roughly 7.4% above the close: "We downgrade our rating to Hold (from Buy previously) as we see better value elsewhere… we believe the valuation multiple could come under pressure." That is a second consecutive down session and a cumulative −9.8% since the print. Rollins traded at its lowest since October 2023 (CNBC). Cruise lines were broadly weak, with Viking Holdings posting its worst decline since April 2025 (Bloomberg) — outside the index. Workday fell more than 5% after Thursday's 18% and still ended the week about 9% higher.

Times are ET. Every day page from Monday 8/17 through Friday 8/21 was independently re-pulled from Earnings Whispers this session (/1 = before open, /2 = after close) and screened against the Investing.com S&P-500-component board. The 8/10–8/14 rosters carry forward from prior verified pulls with reactions added from this session's boards. Re-verify times and membership against company IR before trading any date.

Mon 8/10 — completed. BMO: Berkshire Hathaway B (BRK.B) 8:00 — closed −0.57% at $504.03, a fifth consecutive decline and now −4.9% from the post-print close. AMC: Simon Property Group (SPG) 4:05.

Tue 8/11 — completed. BMO: Cardinal Health (CAH) 6:45; Amentum (AMTM) 8:00. AMC: Lumentum (LITE) 4:00; Super Micro Computer (SMCI) 4:05.

Wed 8/12 — completed. BMO: Amcor (AMCR) 6:00; Trimble (TRMB) 6:55. AMC: Cisco (CSCO) 4:05 — closed −1.58% at $111.68 on a second down session, taking the cumulative post-print move to −9.8%, after HSBC cut to Hold with a $120 target; Coherent (COHR) 4:05.

Thu 8/13 — completed. BMO: Tapestry (TPR) 6:45 — stabilised at +0.46% to $128.98 after Thursday's −16.49%. AMC: Applied Materials (AMAT) 4:00 — closed −5.12% at $507.18, a second consecutive decline and −7.5% across the two sessions, on adjusted $3.50 EPS and $9.12bn revenue with semiconductor-systems sales of $7.04bn against a $6.96bn FactSet consensus; operating cash flow of $3.04bn beat a $2.13bn estimate.

Fri 8/14 — completed. Neither page listed an S&P 500 reporter. The before-open page was micro-cap, biotech and materials; the after-close page micro-cap and gaming. None of the reviewed names is carried by the constituent board.

Mon 8/17. Neither page lists an S&P 500 reporter (ten names screened across both buckets).

Tue 8/18. BMO: Home Depot (HD) 6:00. AMC: Keysight Technologies (KEYS) 4:05, Jack Henry & Associates (JKHY) 4:15.

Wed 8/19. BMO: Lowe's (LOW) 6:00, Estée Lauder (EL) 6:00, Target (TGT) 6:30, Analog Devices (ADI) 7:00, TJX Companies (TJX) 7:30. AMC: Nordson (NDSN) 4:30.

Thu 8/20. BMO: Deere & Company (DE) 6:20, Walmart (WMT) 7:00. AMC: Ross Stores (ROST) 4:00.

Fri 8/21. Neither page lists an S&P 500 reporter (four names screened before the open; the after-close page carried a single micro-cap).

• No additions and no removals. All twelve 8/17–8/21 names re-appeared on this session's independent re-pull with identical timestamps: HD 6:00, KEYS 4:05, JKHY 4:15, LOW 6:00, EL 6:00, TGT 6:30, ADI 7:00, TJX 7:30, NDSN 4:30, DE 6:20, WMT 7:00, ROST 4:00. This is the second consecutive edition with a fully stable forward roster.

• Two names from the prior 8/18 after-close pull were not visible on this capture — Toll Brothers (TOL) 4:30 and Mercury Systems (MRCY) 4:00 — because the 8/18 after-close listing truncated mid-page. Both were conservative exclusions in any case; under this report's standing rule a single absence is not evidence of removal, and they will be treated as removed only after a second consecutive absence.

• Index membership change to diarise, not an earnings item: Reddit (RDDT) joins the S&P 500 on 18 August, replacing AvalonBay Communities (AVB), which Equity Residential is acquiring; the combined company remains in the index as Vivmark Residential. Neither name reports in this window, but the rebalance lands on the same morning as the Home Depot print.

• Membership caveat, restated rather than buried: the constituent board used as this report's screening proxy does not carry Coherent (COHR) — retained for continuity across seven editions — and does not carry Workday (WDAY) or Sandisk (SNDK), which is why both appear in §4 rather than as index constituents. Confirm with company IR. The board does carry HD, KEYS, JKHY, LOW, EL, TGT, ADI, TJX, NDSN, DE, WMT, ROST, TPR and AMAT.

• Conservative exclusions on this pull: Fabrinet (FN) 4:15, Flexsteel (FLXS) 4:15, XP (XP) 4:15, DocGo (DCGO) 4:05 and Yalla (YALA) 5:00 on 8/17; Baidu (BIDU), Klarna (KLAR), Amer Sports (AS), Hesai (HSAI), SQM 7:55, ZTO 6:00 and Auna 5:50 on 8/18; Viking (VIK) 7:00, Weibo (WB), ZIM 7:00, Full Truck Alliance (YMM), Kingsoft Cloud (KC), BILL Holdings (BILL) 4:00, Webull (BULL) 4:15, Coty (COTY) 4:30, Alvotech (ALVO), Carlyle Credit Income Fund (CCIF) 4:05, John B. Sanfilippo (JBSS) 4:10, Telix (TLX) and Unifi (UFI) on 8/19; Alibaba (BABA), Advance Auto Parts (AAP) 6:30, NetEase (NTES), Futu (FUTU), Autohome (ATHM), Daqo (DQ), ScanSource (SCSC), Aegon (AEG), Atour (ATAT), Flowers Foods (FLO) 4:05, Flux Power (FLUX), OSI Systems (OSIS) 4:00 and Intchains (ICG) on 8/20; BJ's Wholesale (BJ) 6:45, KE Holdings (BEKE), Buckle (BKE) 6:50, ZKH Group (ZKH) and OwlTing (OWLS) on 8/21. Fabrinet remains the most borderline of the group; all are listed in Data Notes.

• Timing bucket unpublished: none. Every S&P 500 name above carries a specific clock time.

• What the forward calendar hands the desk. The current week closed with the two most consequential reactions already marked, and both were negative: Applied Materials sold a beat twice, −2.48% then −5.12%, and Tapestry fell 16.49% on a double beat and a dividend raise before stabilising at +0.46%. That is the reaction function going into a block of six U.S. consumer reporters compressed into three sessions — Home Depot, Lowe's, Target, TJX, Walmart and Ross Stores — while Bloomberg-compiled expectations have consumer-discretionary 2026 earnings growth marked at 31%, up from 12% in May. Analog Devices (8/19 BMO) is the analog read into a week where analog rose and logic fell, Keysight (8/18 AMC) the test-and-measurement one, and Deere (8/20 BMO) the only industrial in the block. There is no S&P 500 reporter on either Monday or Friday, so the week's risk is concentrated in a 48-hour window from Tuesday's open to Thursday's close.

Name the shape: a long-end-led bear steepener with the bill anchored, and it is a term-premium move rather than a policy-path move. The evidence is in the gradient. The 3-month bill fell a basis point to 3.86% and the 1-year rose only 1 bp, while the 10-year rose 5 bp, the 20-year 5 bp and the 30-year 4 bp. A market repricing the policy path moves the 2-year most — it moved 2 bp, and it spent part of the morning below 4.10%, the lowest since 30 June, before a UK gilt selloff dragged it back (Bloomberg). A market repricing the compensation for holding duration moves the back end most, which is what happened: 2s10s widened 3 bp to 51 bp, the widest since May, and 2s30s widened 2 bp to 108 bp. On the week the effect is starker still — the front end fell (1-year −3 bp, 2-year −2 bp) while the long end rose (20-year +5 bp, 30-year +6 bp), so the weekly curve steepened 5 bp at 2s10s and 8 bp at 2s30s while every yield inside a year went down.

The mechanism, stated plainly. Two auctions this week set the tone: Wednesday's new 10-year drew the highest yield since 2007 and Thursday's new 30-year the highest since 2001 (Bloomberg), the latter now being written up as an investor warning to the Treasury Secretary. A weak consumer print that pushes the hike further out does not help the long end when the reason the hike is being pushed out is a slowing economy that widens the deficit, and when the supply calendar is the binding constraint. Sit Investment's Bryce Doty framed the same point from the inflation side — "weak numbers are creating a fear that the lack of rate increases will create a resurgence of inflation" — and concluded that "the curve-steepening trade is alive and well." Both readings produce the same trade and the same curve.

The bill is a separate signal, and it belongs with the funding data. The 3-month was the only tenor on the board that fell, and it fell on the day ON RRP take-up printed a third consecutive record low at $0.250bn (§9b). A bill curve easing while the coupon curve sells off is a financing signal, not a rates signal: bills are the instrument the shrinking money-fund cushion is buying. WSJ's 5:04 PM ET real-time board reads 10-year 4.697% (+5.0 bp), 30-year 5.262% (+5.3 bp), 2-year 4.182% (+2.7 bp), a few tenths above the official 3 p.m. par marks in each case — a baseline and timing artefact, not a level dispute, since the two are struck an hour apart.

Current week (Aug 10–14) — released

On the retail print itself. The $763.6bn total was 5.0% above July 2025 and the May-to-July window 6.3% above the same stretch a year earlier (Census Bureau), so the level is not collapsing — the momentum is. The composition is what makes it a stagflationary print rather than a disinflationary one: gasoline stations fell 0.9% on the month and rose 16.2% on the year, which means the consumer is paying more for fuel and buying fewer units of everything else. Jefferies' Thomas Simons put it to Bloomberg as "the reduction in purchasing power sucked up by higher gasoline prices." Bloomberg Economics' Andrew Sacher read it more benignly: "the data point to slower momentum rather than a consumer retrenchment, and remain consistent with our call for the Fed to stay on hold this year."

Next week (Aug 17–21)

There is no "Very high" release anywhere in next week's calendar. That is itself the setup.

Current target range: 3.50–3.75% (IORB 3.65%; EFFR 3.63%, NY Fed). The market prices hikes, not cuts: the probability of a target range below 3.50–3.75% is 0.0% at every 2026 meeting on both vendors.

The headline — CME FedWatch, 16 September meeting

Meeting information from CME's Current view: contract ZQU6, expiring 30 September, mid price 96.3325, prior volume 58,460, prior open interest 233,385.

Provenance of every column, and this run is unusually clean. CME's Current view published the complete four-column numeric table — NOW, 1 DAY, 1 WEEK and 1 MONTH — with the footer stamped "Data as of 14 Aug 2026 05:02:44 PM CT", i.e. 6:02:44 p.m. ET. Every cell above is therefore a published vendor figure; for the first time in this report's window the 1 MONTH column is not a bar-height estimate, and no cell in this section is chart-read. One correction is owed from the prior edition, as promised. Thursday's report published CME's September hold at 65.2%, read live at about 6:20 p.m. ET; CME's settled figure for 13 August is 66.1%, a 0.9-point difference. The live-read caveat this report attaches to any post-5 p.m. CME pull did what it was written to do, and the settled figure is what appears in the 1 DAY column above.

The cross-check — Investing.com Fed Rate Monitor, and the gap reconciled

Investing.com's card for 16 September, updated 5:45 PM EDT on 14 August, shows hold 67.7% (previous day 69.0%, previous week 56.6%) and +25 bp 32.3% (31.0%, 43.4%), on a September fed funds contract price of 96.330.

The gap is 0.2 percentage points — the smallest this report has recorded — and it does not reconcile the usual way. CME's mid is 96.3325 against Investing.com's 96.330, a difference of 0.25 basis points, with CME quoting the higher price. Work the sensitivity: the 16 September meeting falls on day 16 of a 30-day contract month, so a certain 25 bp hike lifts the September average effective rate by 25 × (14/30) = 11.7 bp, which makes one basis point of ZQU6 worth about 8.6 points of headline probability. A 0.25 bp higher price should therefore make CME's hold about 2.1 points higher than Investing.com's; it is instead 0.2 points lower. The residual is the assumed base rate. CME anchors on the prevailing EFFR of 3.63%, half a basis point above the 3.625% range midpoint, and half a basis point of base is worth roughly 4.3 points of headline probability in the opposite direction. Net the two and you get a fraction of a point — which is what the board shows. The two vendors are not disagreeing about the Fed; they are quoting the same contract to different conventions inside a quarter of a basis point. The one caveat to carry: Investing.com's own previous-day column reads 69.0% for the September hold, against the 67.0% this report read live from the same vendor at 5:45 p.m. EDT on 13 August — the vendor revises its daily snapshot after the fact. All triplets below use Investing.com's own columns end-to-end so the changes are internally consistent; see Data Notes.

(a) Current-year meeting distributions — current / [prev day] / [prev week]

Modal ranges in bold. Contract prices: ZQU6 96.330, ZQV6 96.280, ZQZ6 96.165.

The one-day and one-week changes, and the 2026 strip did almost nothing. In one session the September cumulative hike rose 1.3 points (31.0% → 32.3%), the October cumulative fell 0.5 points (46.8% → 46.3%) and the December cumulative rose 0.5 points (66.7% → 67.2%). All three are inside the rounding noise of a single card, on a session whose headline release missed by seven tenths of a percentage point. That is the finding. A −0.6% retail print moved the 2026 meetings by half a point and moved the 2027 strip by up to 4 points (section b) — the market did not change its mind about whether the Fed hikes this year, it changed its mind about what a slowing consumer means for 2027 inflation. Over one week the picture is entirely different and entirely front-loaded: September −11.1 points, October −12.8, December −9.6, with the weekly gradient now humped in the middle rather than decaying with tenor. The multi-day read from CME's own four columns is the cleanest version: the September hold has gone 41.9% a month ago → 55.6% a week ago → 66.1% on Thursday → 67.5% now, and the +50 bp September tail, which was 8.0% on 14 July, is 0.0%. Bloomberg's framing of the same move: only about nine basis points of tightening remain priced for September against nearly nineteen on 31 July.

(b) Next-year meeting path

Every 2027 contract fell, and that is where the retail print was actually traded. The lowest price on the strip remains 95.990 at the 28 July 2027 meeting, but it now implies 4.010% against 3.985% on Thursday — the terminal rate rose 2.5 basis points, reversing about two-fifths of Thursday's 6.5 bp rally in a single session. The move was uniform and it grew with tenor: January 2027 96.140 → 96.130 (−1.0 bp), March 96.080 → 96.065 (−1.5 bp), April 96.050 → 96.030 (−2.0 bp), June 96.020 → 95.995 (−2.5 bp), July 96.015 → 95.990 (−2.5 bp), September 96.025 → 95.995 (−3.0 bp), October 96.030 → 96.000 (−3.0 bp), December 96.060 → 96.025 (−3.5 bp). The modal range is 3.75–4.00% at every 2027 meeting and the modal probability still decays monotonically from 42.8% in January to 34.9% in December, so the shape is unchanged and the whole distribution has shifted up. The first non-zero probability of a target range below 3.50–3.75% has moved back out to the 15 September 2027 meeting at 0.6%, from 28 July 2027 on Thursday — thirteen months of futures with no easing priced at all, up from eleven.

(c) Year-end probability ladders

Year-end 2026 (9 December 2026 meeting), relative to the current 3.50–3.75%:

Year-end 2027 (8 December 2027 meeting), relative to the current 3.50–3.75%:

The week-over-week arc, and the divergence reopened one day after it closed. Thursday's edition recorded the first session in this report's window in which both ends of the strip priced hikes out together, with the 2027 year-end cumulative-above falling 5.8 points to 73.2%. Friday reversed most of it: cumulative-above at year-end 2027 rose to 77.2%, a 4.0-point one-day move, and the cumulative-below fell to 2.8% from 4.4%. Inside the ladder, hold at year-end 2027 fell 2.5 points to 19.9% while +50, +75 and +100 rose 1.9, 1.6 and 0.6 points. At year-end 2026 the same day produced a 0.5-point move in the cumulative hike. So the week ends where it began in structure — a front end pricing hikes out and a back end pricing them in — with the difference that the front end is now 11 points further along and the back end has recovered most of one day's dovish shock. Over the week the 2027 year-end cumulative-above is 77.2% against 77.7%, essentially unchanged, while September's hike probability fell 11.1 points. The entire week's repricing was a re-timing, not a re-levelling.

Rounding, stated transparently. Investing.com's cards do not always sum to exactly 100.0%: the September, October and December 2026 cards and the January, September and October 2027 cards each sum to 100.0; March, June, July and December 2027 each sum to 99.9; and April 2027 sums to 100.2. CME's September headline sums to 100.0 in every one of its four columns. Cumulative figures above are computed by summing the vendor's published cells without re-normalising, so a cumulative may carry the same ±0.1 to ±0.2 rounding as its parent card. No cell in this section is estimated or chart-read.

(a) IG and HY credit spreads

As-of date, stated rather than implied: the ICE BofA option-adjusted spread series below are FRED's 13 August prints, published on the morning of 14 August. FRED carries these with a one-business-day lag, so there is no 14 August index credit spread available at the time of writing; the same-day direction is cross-checked against Bloomberg below.

Same-day cross-check. Bloomberg's 14 August coverage of the primary market puts "the average spread on investment-grade bonds at roughly 0.78 percentage point over Treasuries" — 78 bp, one basis point inside FRED's 13 August print, which corroborates a flat-to-marginally-tighter IG session and is the ninth consecutive print in a 78–79 bp band. The IG index has not moved more than a basis point since 3 August. The movement is all in the tail: the CCC credit spread widened 4 bp to 1,024 bp, its widest since 31 July, and the CCC-minus-HY differential widened 4 bp to 753 bp, a new high for the reporting window and 149 bp above January's 604 bp.

CDX retrieval note — the six-step ladder was worked in full and the index level remains publicly unobtainable. (1) Bloomberg in Chrome: /markets and /markets/rates-bonds were both rendered; the latter publishes the Bloomberg Fixed Income Indices grid — Global Aggregate 500.17, +0.50, MTD +0.53 — and was verified programmatically to contain no CDX string anywhere in the page text. (2) WSJ Market Data Bonds & Rates: the 5:04 PM ET Treasurys grid, consumer rates and money-rate tables were captured; no CDX, verified programmatically. (3) Cbonds: the CDX.NA.IG 5Y page was rendered directly and displays the level as *** bps with the previous value masked to *** on 12/08/2026, calculating organisation IHS Markit — entitlement-gated. (4) FT Markets Data: markets.ft.com/data/search?query=CDX returned no instrument page for the index. (5) Search across ICE, S&P Dow Jones and TradingView product pages: these return methodology and index-family documentation, and the most recent third-party level found anywhere was an undated-context 81 bp reference from 27 July — under this report's standing rule an undated digest number is not a CDX level and is withheld. (6) Cash-market proxy, labelled as such: HYG closed $80.14, +$0.28 / +0.35% — a proxy for HY credit direction only, not the CDX HY 5y index, which is quoted in price points rather than spread. The LQD close could not be retrieved from a primary source this session. Quoting conventions restated so no reader mis-signs the field: CDX IG 5y is quoted in basis points of spread; CDX HY 5y in price points, where a rising price means tightening credit spreads.

(b) Money-market and funding plumbing

NY Fed reference rates publish at roughly 8:00 a.m. ET for the prior business day, so the rates below are the 13 August fixings released on the morning of 14 August; the ON RRP figure is same-day.

Read this together with §6, because the two are the same story. The ON RRP facility has effectively emptied — $0.250bn is a rounding error against a $2.4tn peak — and the take-up has now printed a fresh record low on three consecutive sessions. Reserves are $198.6bn below the July peak. And yet there is no daily stress anywhere in the numbers: SOFR at 3.62% is 3 bp through IORB, the 1st-to-99th percentile band is 11 bp, tri-party and broad general collateral both clear at 3.60% on $1.2tn a day, and EFFR sits mid-range with a 5 bp band. The bill curve is where the adjustment is showing up — the 3-month eased a basis point on a day the 10-year rose five, which is what happens when the marginal cash that used to sit at the facility has to buy something. There is no cushion left for September quarter-end. The Fed publishes its own Reserve Demand Elasticity read on 20 August at 10:00 (§7), which is now the most informative scheduled item on the funding question. No month-end or quarter-end distortion is present in the 13 August fixings; mid-month is the quiet part of the cycle, which is precisely why the RRP number is notable.

(c) Rates volatility and swap spreads

The MOVE index is 69.23 on a 13 August vintage, down 2.86 points, −3.97%, from 72.09 on 12 August. The Investing.com series is delayed and published no 14 August value at the time of writing, so the ratio below pairs two different dates. MOVE/VIX is 4.86× — 69.23 on 13 August against a VIX of 14.25 on 14 August — and it should not be traded on; it is quoted for continuity with prior editions, which ran 4.93× on the same mismatched basis. What can be said without a date mismatch is directional and it is striking: rates vol fell 4.0% into a week in which the 30-year auction cleared at the highest yield since 2001 and the 10-year at the highest since 2007, and equity vol fell 2.6% to a seven-month low on the day the largest consumer release of the month missed by seven tenths. Both vol markets are pricing the absence of an event, in a week that delivered two. 2y, 10y and 30y swap spreads could not be retrieved from a primary source this session — the Bloomberg rates board, the WSJ money-rates table and the FT markets search were all worked and none publishes the series; this is a gap and it is recorded as one rather than filled.

(d) Issuance, leveraged loans and private credit

The primary market is the most interesting thing in credit this week, and it is showing indigestion for the first time this year. Per Bloomberg, investors pulled about 36% of their initial orders for high-grade deals on average this week after final pricing was squeezed — twice the prior week's rate, and well above the 22% average for the year — with Thursday's attrition reaching 45%, a record for 2026. AMD's $4.75bn four-tranche offering saw orders for its longest tranche more than halve after it cut the concession. The context is volume: $136bn priced in the last two weeks alone, putting August on track for its busiest ever, $1.4tn of IG sold year to date, 9% above the pace of 2020's record $1.75tn year, and global issuance reaching $5tn on Monday, more than a month faster than last year's record. Next week's expected supply is only about $20bn by Bloomberg's syndicate survey, and the market usually goes quiet until after Labor Day. Demand at the fund level is still healthy — an inflow of almost $3bn for the week ended Wednesday.

Private credit is where the equity market started pricing the risk. Bank of America's downgrade of Broadcom's issuer and bond ratings turned on the AI XPV financing platform built with Apollo and Blackstone, and the equity market marked both sides: Broadcom −5.95%, Blackstone −3.61%. CNBC's own morning piece — "AI's infrastructure boom is getting more leveraged — and harder to track" — describes the mechanism: hyperscalers and their backers are funding the buildout through bond markets, joint ventures, leases and other structures, while hedge funds amplify the exposure through prime brokerage and derivatives. On the leveraged loan side, the most recent published Morningstar LSTA index analysis available is June 2026, when the index returned 0.08%, its weakest month since February, on a renewed selloff in software loans and a softer technical backdrop; no verified August level was obtainable from a primary source this session, and it is flagged rather than estimated. Named private-credit watch items: the XPV vehicle itself; Apollo and Blackstone as its sponsors; and the software loan complex, which has now had a rumour-driven re-rate and an immediate reversal in the space of two sessions (§2).

Quote basis — read this before the table. All pairs are spot in the market convention shown; a positive move on a USD/XXX pair means the dollar strengthened, and on EUR/USD, GBP/USD, AUD/USD and NZD/USD it means the dollar weakened. The TradingEconomics pull landed at approximately 18:30 ET on 14 August, before that vendor's daily boundary rolled — every row carries an Aug/14 stamp with a live %Chg column ranging from −0.43% to +3.31%, so unlike the prior two editions the vendor's own %Chg column is usable and is reproduced directly. It has been independently corroborated against the same vendor's 13 August levels published in the prior edition; worked examples are in Data Notes, where the agreement is within 0.03 point on ten of thirteen pairs and within 0.08 on twelve, with USD/MXN the single outlier.

The take — the dollar fell on the one session this week when the rate inputs argued for it to rise, and that is the story. Run them: the 10-year rose 5 bp, the 30-year 4 bp, the 2027 terminal contract 2.5 bp, and the September hold moved only 1.4 points. Every one of those is dollar-supportive on a rate-differential model, and the dollar index fell 0.33%, its largest daily decline in a fortnight. Compare the two preceding sessions: on Wednesday the index rose 0.13% on dollar-negative inputs; on Thursday it was inert at +0.01% on the largest dollar-negative rate move of the month. Three consecutive sessions in which the dollar has done the opposite of what the front end told it to do is no longer noise — it is the market trading the growth leg rather than the rate leg, and Friday's release was unambiguously a growth miss. That reading also explains why the move was broad rather than concentrated: eleven of the thirteen pairs above moved against the dollar.

The contrarian cross has flipped, and it is now the Canadian dollar rather than the Korean won. USD/CAD fell 0.40% on a 1.42% crude gain — the first correctly-signed petro-currency session in four, after −0.10% on a 1.30% gain, +0.15% on a flat barrel and −0.08% on a 2.47% decline. This report has argued for three editions that anyone using CAD as a crude proxy was running an unhedged rates position; on Friday the rates leg and the oil leg pointed the same way for the first time, so the pair is uninformative as a test rather than newly reliable. The won remains the stubborn one. USD/KRW moved −0.08% on a session when the Kospi rose 2.42% to 6,977.94 with SK Hynix +3.26% and Samsung +2.43%, and the index is roughly 26% above its 30 July low. Meanwhile USD/TWD fell 0.41%, the largest Asian move, on a Taiwan market with no comparable rally. Four consecutive sessions in which the currency of the strongest equity market is the weakest Asian currency and vice versa. The inference has not changed and has now survived four tests: the Kospi rally is domestically levered rather than foreign-funded, so it generates no currency demand, while Taiwan's flows are the repatriation kind.

And the one nobody is looking at: the rouble. USD/RUB rose 3.31% in a session and 4.78% on the week, with the pair up 10.74% on the month. Hegseth's "indefinitely" and Bessent's "never been seen" were priced in exactly one asset on Friday, and it was not crude — WTI rose 1.42%, which is a normal day. A currency that falls 3.3% while the commodity its economy sells rises 1.4% is pricing the payment channel, not the barrel, and it is the cleanest available read on how seriously the market takes the sanctions threat. Watch it over the weekend: it is the instrument most likely to gap on a Sunday headline.

Basis, stated before the table. The settlement column is the 14 August front-month close from the Investing.com continuous futures historical boards, corroborated against Bloomberg's own quote board (crude 82.40, +1.42% on both, to the cent and the basis point) and against TradingEconomics' independent Aug/14 board. Note a base change from prior editions: those used the per-contract-month boards, which mark Thursday's crude at $81.21 and gold at $4,407.10; the continuous boards used here mark the same session at $81.25 and $4,420.40. The two are not interchangeable and the change is disclosed rather than smoothed — every change figure in this table is computed within the continuous series. The week / month / YTD columns are TradingEconomics' own spot-basis series and are not on the same basis as the settlement column; the vendor's header order was verified programmatically as Price | Chg | %Chg | Weekly | Monthly | YTD | YoY | Date before any YTD was quoted.

The take — the complex rose on a supply headline, and the composition says the market believes it. WTI settled $82.40, +1.42%, at the very top of an 80.77–82.99 range, and Brent $88.60, +1.76%, after the U.S. Defence Secretary said the naval blockade of Iranian ports can continue "indefinitely" and the Treasury Secretary promised measures "that have never been seen on Iran." Closing at the high of the day on a geopolitical headline is the tell — Thursday's session, by contrast, opened +1.15% on genuinely bullish supply news and slid 2.52% from the high, which is what a market with no marginal buyer does. The corroborating price is not in oil at all: the rouble fell 3.31% (§10). The refined products are where the year's move actually is. Heating oil is +101.92% year to date against crude's +43.50% and rose 9.77% on the week against crude's 5.40%; gasoline is +85.79% on the year and +6.48% on the week. Both product cracks widened again on a session the barrel rose, which is the harder configuration for a crack trade to survive and it survived at the product level.

The positioning read, and the crack expression failed at the equity level for the first time in eight sessions. The products widened, but the refiner equities did not: Marathon Petroleum −0.27%, Valero −0.36% and Phillips 66 +0.43% — an average of −0.07% — against a producer-and-services basket that averaged +2.4% (Halliburton +4.78%, SLB +3.28%, Devon +3.28%, ConocoPhillips +1.81%, Diamondback +1.49%, Occidental +1.14%, Chevron +1.17%, Exxon +0.94%, Baker Hughes +2.18%). And all three refiners printed all-time highs intraday before closing flat to lower. Seven consecutive sessions of refiners over producers ended on a session when the crack itself kept widening — so this is a positioning unwind in the equity expression, not a fundamental break in the trade (§12 idea 2). The honest reading: the refiner names had absorbed the whole crack story and more, and the marginal dollar went to the beta.

Gold recovered and the franc went with it, which restores the hedge reading. The metal closed $4,432.00, +0.26%, with TradingEconomics' spot series +0.60% to $4,375.50 on its own basis, after Thursday's −1.35% and a 2.26% intraday fade. USD/CHF fell 0.07% on the same session, so the two legs of the inflation-hedge pair moved together again after four sessions apart (§10). Two caveats before anyone reads too much into a 0.26% day. First, the session's rate inputs were gold-negative — the 10-year rose 5 bp and the terminal rose 2.5 bp — and gold rose anyway, which is the inflation-hedge signature rather than the real-rate one, and is consistent with Michigan's year-ahead inflation expectation ticking up to 4.3%. Second, silver fell 0.26% and copper was unchanged, so this was not a metals-complex bid; platinum +1.41% and palladium −0.79% went in opposite directions on the same day. A precious-metal rally with no silver and no copper is a monetary trade, not an industrial one.

1. The rates trade — the first winner in three attempts; hold it and raise the stop

Mark on the prior edition's rates idea, and it worked. The book is long ZQZ6 against short ZQZ7, DV01-matched one-for-one at half of half size, entered at 96.165 / 96.060 — a spread of 10.5 bp. Friday: ZQZ6 was unchanged at 96.165 while ZQZ7 fell to 96.025 (−3.5 bp), so the spread widened to 14.0 bp and the position made +3.5 bp, or +33% on the entry spread, in one session. At $41.67 of DV01 per basis point per contract, that is $145.85 per contract pair. The stated invalidation was the spread through 7.5 bp; it moved 6.5 bp the other way. Neither of the other two triggers fired — no 2026 meeting shows a non-zero cut probability, and the PCE deflator is not until 26 August.

Why it worked is the reason to keep it. The thesis was that hikes come later in 2026 than the market thought but the destination is not lower, and that Thursday's 6.5 bp dovish repricing of the terminal was a misreading of a producer print whose core line had accelerated fourfold. Friday tested it directly: a consumer release missed by seven tenths, and the 2026 meetings moved by half a point while every 2027 contract fell 1.0 to 3.5 bp (§8). Two structures were stopped out in two sessions trading the slope of this strip; the third worked because it stopped trading the front and started trading the destination.

Modal path, base case and the tails — this is what the position expresses. Modal path: hold on 16 September (Investing.com 67.7%, CME 67.5%, ease 0.0%); hold on 28 October at 53.7% modal with the cumulative hike at 46.3%; one 25 bp hike delivered by 9 December, 3.75–4.00% modal at 45.1% with the cumulative at 67.2%; terminal 4.010% drawn by the 28 July 2027 contract at 95.990; year-end 2027 modal 3.75–4.00% at 34.9%, with the first non-zero easing probability anywhere on the strip at 0.6% in September 2027. Base case: demand is slowing (retail sales −0.6%, Michigan 51.0) while the supply-side inflation impulse is intact (gasoline stations +16.2% y/y, heating oil +101.9% YTD, core PPI ex food, energy and trade +0.4%). That combination delays the hike and raises the terminal, which is exactly the shape the position is long. Tail one, dovish: claims through roughly 240,000 on 20 August would start pricing a 2026 cut into a strip that shows 0.0% at every meeting — the front leg gains but the back leg gains more and the spread compresses. Tail two, hawkish: a core PCE deflator at +0.4% on 26 August puts September back to a coin flip, the front leg loses and the back leg is unchanged. Both tails compress the spread, which is why the stop matters more than the target.

Expression: long ZQZ6 (December 2026 fed funds) against short ZQZ7 (December 2027 fed funds), DV01-matched one-for-one at $41.67 per basis point per contract, unchanged. Catalyst: initial claims 8/20 8:30; the 26 August PCE deflator; Jackson Hole in the closing week of the month. Invalidation, tightened: the spread back through 11.0 bp, which gives back the whole week's gain and no more; or the core PCE deflator at or below +0.2% m/m on 26 August; or any 2026 meeting showing a non-zero cut probability. Sizing: unchanged — do not add. One winning session after two stops earns a tighter stop, not more capital.

2. Own the crack, not the barrel — downgrade to small; the equity expression broke while the crack widened

Expression: long a basket of Valero, Phillips 66 and Marathon Petroleum against a short in front-month WTI, sized so the futures leg matches the basket's crude beta. Mark: the worst session the idea has had, and it lost on the leg that was supposed to be safe. VLO −0.36%, PSX +0.43%, MPC −0.27% — an average of −0.07% — against crude +1.42%, so the pair lost roughly 1.49 points. All three names printed all-time highs intraday and closed flat to lower. What is intact: the crack itself widened again — heating oil +9.77% on the week against crude's +5.40%, and +101.92% year to date against +43.50% (§11). What broke: the equity expression. On a session when the producers and services averaged +2.4%, the marginal energy dollar went to beta rather than to margin, and seven consecutive sessions of refiner outperformance ended. Catalyst: the EIA weekly report; any Hormuz or blockade headline over the weekend. Invalidation, unchanged: a confirmed Hormuz reopening; or a distillate build above 3m barrels. Sizing: downgraded from medium to small. A trade whose thesis is intact and whose expression has stopped tracking it should be smaller, not re-argued.

3. Sell the rumour premium, buy the delivered miss — hold, effectively flat

Expression: long an equal-weight pair of Cisco and Applied Materials against short an equal-dollar basket of GoDaddy, CoStar, Paycom and Fair Isaac, beta-adjusted. Mark: a loss of 0.27 points, and the direction was right. The short basket collapsed — GDDY −5.59%, CSGP −2.03%, PAYC −2.46%, FICO −2.24%, an average of −3.08% — exactly as the thesis required. The long leg fell further: CSCO −1.58% and AMAT −5.12%, an average of −3.35%, because Applied Materials was marked down a second time on its print and Cisco took an HSBC downgrade to Hold with a $120 target. The uncomfortable detail: the rumour premium came out of the short basket in a single session and the pair still lost money, because the long leg had its own idiosyncratic news on the same day. That is a warning about the construction, not the idea. Catalyst: confirmation or denial of the Silver Lake–Workday talks. Invalidation, unchanged: a second confirmed sponsor bid in enterprise software; or the short basket underperforming the long by more than 4% within five sessions — it is 3.08 points the right way after one. Sizing: small, unchanged.

4. Own downside in the consumer block, not in the index — working; hold into Tuesday

Expression: a one- to three-week put spread on a consumer-discretionary basket or the sector ETF, struck beneath Thursday's level and financed by selling the wing. Mark: the first of three consecutive downside structures to be on the right side of its catalyst. Retail sales printed −0.6% against +0.1%, Michigan fell to 51.0 from 55.2, and the SPDR S&P Retail ETF fell about 2% on the week, its first down week in three (CNBC). Finviz Consumer Cyclical was −0.18% on the day and is −2.39% year to date, the second-worst group on the board. Neither invalidation triggered: the stated triggers were a control-group print above +0.5% m/m and Home Depot rising more than 3% on its print — the headline was −0.6% with ex-autos −0.3%, and Home Depot closed −0.83% three sessions ahead of reporting. What to be honest about: the structure is working on the data and has not yet been paid by the tape, because a 0.18% sector decline does not move a put spread much. The payoff is in the 48 hours from Tuesday's open to Thursday's close, when six of these names report. Catalyst: Home Depot 8/18 BMO; Lowe's, Target, TJX 8/19 BMO; Walmart, Ross Stores 8/20. Invalidation, unchanged. Sizing: small, defined-risk, premium-at-risk only — do not roll it out past 8/20; the whole thesis expires with the block.

5. Long gold against the Swiss franc — hold at a quarter

Expression: long spot gold funded in CHF, notional-matched, at quarter size. Mark: a gain of about 0.19%. Gold closed $4,432.00, +0.26% on the continuous board (TradingEconomics spot +0.60%), and USD/CHF fell 0.07%, so the short-franc leg cost 0.07% and the long-gold leg made 0.26%. Thesis, confirmed on its narrowed terms: Thursday's edition kept only the inflation-hedge reading and discarded the real-rate one. Friday tested precisely that: the 10-year rose 5 bp and the terminal rose 2.5 bp — both gold-negative on a real-rate model — and gold rose anyway, on the same morning Michigan's year-ahead inflation expectation ticked up to 4.3% from 4.2%. Catalyst: the 26 August PCE deflator; any weekend escalation, which is gold-positive and franc-positive simultaneously and would make this a poor hedge. Invalidation, unchanged: spot gold below $4,300; or the franc strengthening more than 0.75% against the dollar in a session where gold is flat or lower; or a second consecutive session of gold falling while the terminal rate falls. Sizing: a quarter, unchanged. One confirming session does not restore capital that two disconfirming ones removed.

6. Protection on the CCC cohort funded in IG — working; the differential made a new window high

Expression: long CCC-exposed credit protection (or short a levered-loan/CCC-heavy vehicle) against long IG cash. Mark: the best print the idea has had. The CCC credit spread widened 4 bp to 1,024 bp, the HY credit spread was unchanged at 271 bp, and the CCC-minus-HY differential widened 4 bp to 753 bp — a new high for this reporting window and 149 bp above January's 604 bp. On the week the differential is +6 bp. The IG funding leg did what it is there to do: nothing, for a ninth consecutive print. Thesis, strengthened by an unexpected route: Friday added a named transmission — Bank of America cutting Broadcom's issuer and bond ratings over an off-balance-sheet AI financing vehicle, with Blackstone −3.61% alongside (§9d). Tail credit stress is no longer only a CCC-index abstraction. Catalyst: the post-Labor Day supply burst against a 36% order-attrition rate; the September quarter-end with ON RRP at $0.250bn. Invalidation, unchanged: the differential back through 735 bp. Sizing: small, raised from a half to three-quarters of the small allocation — and note the FRED series lags a business day, so the position is never sized on a same-day number.

7. New — on-balance-sheet AI funding against off-balance-sheet AI funding

Expression: long AMD against short Blackstone, dollar-neutral, small. Thesis: Friday priced the same theme twice in opposite directions and told you which financing structure the market will pay for. AMD rose 6.46% on the day it priced $4.75bn of senior notes across four IG tranches — debt on its own balance sheet, rated, publicly marked and used to fund capacity it will own. Blackstone fell 3.61% because a rating agency action against a third party — Broadcom — turned on the AI XPV vehicle Blackstone co-sponsors with Apollo. One structure is transparent and was rewarded with a 6.5% equity move; the other is, in CNBC's words that morning, "harder to track", and was punished without any change to Blackstone's own earnings. The second-order detail that makes it interesting: AMD's own bond book more than halved on the longest tranche amid a record 45% single-day attrition rate, so the market is not indiscriminately funding either — it is discriminating on disclosure. Catalyst: Broadcom's 2 September print; the post-Labor Day IG calendar and whether attrition normalises toward the 22% year average; any further rating action referencing a financing vehicle rather than earnings. Invalidation: BofA restoring Broadcom's ratings, which retires the premise; or AMD's IG spread widening more than 25 bp from new issue, which would say the market is discriminating on leverage rather than on structure and the long leg is wrong; or the pair losing more than 5% in five sessions. Sizing: small — this is a thematic pair with one week of evidence behind it.

8. The consumer split by ticket size — hold; the first winning session

Expression: long Darden, Yum! Brands and Hilton against short Home Depot, Lowe's, DR Horton, Lennar and PulteGroup, beta-adjusted. Mark: the pair gained 0.84 points. The long basket averaged +0.33% (DRI +0.79%, YUM −1.82%, HLT +2.02%) against a short basket averaging −0.51% (HD −0.83%, LOW +0.11%, DHI −0.71%, LEN −0.81%, PHM −0.31%). Cumulatively the pair is −2.11 + 0.84 = −1.27 points, and the stated invalidation — the short basket outperforming the long by more than 3% before 8/18 — is now 1.27 points away rather than 0.89. What worked: the homebuilders finally traded the long end rather than the belly, with the 30-year at 5.25%, up 6 bp on the week (§6), and the retail print took the big-ticket names down with it. Catalyst: Home Depot 8/18 BMO; Lowe's and Target 8/19 BMO. Invalidation, unchanged: the short basket outperforming the long by more than 3% before 8/18; or a retail control-group print above +0.6% m/m, which the July release did not deliver on any published line. Sizing: small, unchanged.

Volatility note. VIX 14.25 (−2.60%), range 14.18–14.72 — the lowest close since 29 December (Bloomberg), on a day the index fell and the largest release of the month missed by seven tenths. MOVE is 69.23 on a 13 August vintage, −3.97%; the delayed series did not update for Friday, so MOVE/VIX at 4.86× pairs two dates and should not be traded on (§9c). At 14.25, one-month S&P implied prices a daily move of about 0.90% against three-session realised of −0.06%, +0.65% and −0.17%, so the short-vol carry is still comfortably positive and the crowding is rational on the arithmetic. The specific objection is calendar, not valuation. This is a Friday close into a weekend carrying a naval blockade described as open-ended, a currency down 3.3% on it, an index rebalance on Tuesday and six consumer prints inside 48 hours. A seven-month low in index vol is a poor place from which to be short gamma over a weekend with a live geopolitical bid in crude. The cheapest expression of that concern is not long VIX — it is the put spread in idea 4, which is already owned.

The crowded consensuses, with the numbers that would break them.

1. "September is done and there is no chance of a cut." The strip prices a 67.5–67.7% September hold, a 46.3% cumulative October hike, a 67.2% cumulative December hike and 0.0% probability of any 2026 cut for a twentieth consecutive session (§8). Only about nine basis points of tightening remain priced for September against nearly nineteen on 31 July (Bloomberg). Both tails are therefore effectively unpriced, and Friday handed each of them an argument: retail sales −0.6% and Michigan at 51.0 on the dovish side; year-ahead inflation expectations at 4.3%, gasoline +16.2% y/y and core PPI ex food, energy and trade at +0.4% on the hawkish one. The stress test: initial claims through roughly 240,000 on 20 August, or a core PCE deflator at +0.4% on 26 August. Note that the market has already shown you which it fears — the 2027 strip sold off on Friday even as September rallied. 2. "Credit is fine." IG at 79 bp, unmoved for nine prints and flat on the year; HY at 271 bp, 10 bp inside January; $1.4tn of IG issued year to date, 9% above a record year's pace; $136bn in the last fortnight alone. The stress test, and it now has three parts: the CCC credit spread at 1,024 bp against 885 bp in January with the CCC-minus-HY differential at 753 bp against 604 bp; order attrition at 36% this week and 45% on Thursday against a 22% year average, which is the first sign the bid is price-sensitive; and the Broadcom rating action, which was about a financing structure rather than about earnings. Index credit is priced for no defaults and every one of those three is moving the other way. 3. "The AI trade is broad." It is not, and Friday's dispersion was extreme in both directions inside one index: AMD +6.46%, Seagate +5.65%, Corning +4.70%, Western Digital +4.41%, Micron +2.42% against Broadcom −5.95%, Applied Materials −5.12%, Oracle −3.77%, CrowdStrike −3.82%, Palantir −2.78%, KLA −2.69%, Arista −2.36% — and the SOX finished −0.31%, which tells you nothing about either group. Nvidia closed −0.06%. The stress test: Applied Materials has now been marked down on two consecutive sessions after a beat, and Broadcom's guidance implies AI is roughly 72% of expected sequential revenue growth next quarter. If the marginal buyer will not pay for a beat and will sell a financing structure, the theme's funding leg is where it breaks first. 4. "The consumer is fine because earnings are fine." Bloomberg-compiled expectations have S&P 500 consumer-discretionary 2026 earnings growth at 31%, up from 12% in May, and Seaport's Jonathan Golub calls the current reporting season the best on record, with Q2 EPS growth above 50%. Against that: retail sales −0.6%, ex-autos −0.3%, Michigan 51.0, and only 8% of consumers expecting income growth to beat inflation — down from 18% in December 2024. The stress test: six of these companies report inside 48 hours starting Tuesday, into a tape that took 16.5% out of Tapestry on a double beat and a dividend raise. 5. "Volatility is cheap because nothing happens." VIX 14.25, the lowest close since 29 December, and it fell on a down day with a big miss. The stress test: implied prices a 0.90% daily index move; in the same session AMD moved 6.5%, Broadcom 6.0%, Applied Materials 5.1%, Copart 7.6% and Reddit 12%. Index vol is cheap because dispersion is doing the work, and dispersion is not a hedge over a weekend. 6. "Small caps are confirming the rally." The Russell 2000 rose 0.53% to 3,069.96 and closed 0.02% below a fresh 52-week high, a fourth consecutive gain and the longest run since June, on the only day this week the three large-cap gauges fell. That is either genuine breadth or it is the most rate-sensitive index in America rallying because September's hike was priced out. The stress test: the front end has already done its work — the September hold cannot go much above 67.5% without pricing October too, and the long end went the other way all week. A small-cap rally that needs 2s10s to keep steepening is a different trade from one that needs earnings.

The two-sided geopolitical tape. Escalation: Defence Secretary Pete Hegseth says the U.S. Navy can keep the blockade of Iranian ports going "indefinitely", rotating ships "as we have, and we'll continue to"; Treasury Secretary Scott Bessent promises economic measures "that have never been seen on Iran"; the war is approaching its sixth month; and the market's response was to take the rouble down 3.31% in a session and 4.78% on the week while crude rose only 1.42%. Separately, the administration imposed a 100% tariff on imported unmanned aircraft systems and components, which lifted the domestic drone complex and is a second front in the tariff regime. De-escalation, or at least containment: crude is $82.40, still below its 7 August level on a monthly basis of only +3.52%, and the VIX fell. The asymmetry is that the escalation channel is priced in one currency and one commodity, and the de-escalation channel is priced in none.

Structural watch items. (1) The funding cushion is gone and there is no daily stress — yet. ON RRP take-up printed $0.250bn, a third consecutive record low; reserves are $198.6bn below the July peak at $2,944.1bn; SOFR trades 3 bp through IORB and repo clears $1.2tn a day at 3.60%. The September quarter-end will be the first in years with no buffer, and the Fed publishes its own Reserve Demand Elasticity read on 20 August. (2) Term premium, not policy, is setting the long end. The 30-year auction cleared at the highest yield since 2001 and the 10-year at the highest since 2007; 2s10s at 51 bp is the widest since May; and Bank of America's Michael Hartnett is publicly arguing that surging national debt and higher bond yields are what derail the equity run. (3) Off-balance-sheet AI financing is now a named equity risk factor, not a think-piece: one rating action moved two large caps 5.95% and 3.61% on the same day. (4) Index-level supply and structure: Reddit joins the S&P 500 on 18 August as AvalonBay leaves; Bloomberg reports OpenAI's annualised revenue topping $40bn ahead of an IPO and Anthropic in talks to buy Decart for $6bn — private-market marks that will eventually need public-market clearing prices.

Full Data Notes & Conflicts and the complete categorized Source Links appendix are in the companion file US_CrossAsset_Daily_2026-08-14_DataNotes.txt, saved alongside this report.