A hundred-point corridor, a magnet at its floor, and a keynote at ten
by 80 points
- The corridor closed to a hundred points overnight, and the market is standing in the middle of it.The dealer put wall jumped two hundred points, from 7,500 — where it had not moved in five sessions — to 7,700. The call wall held at 7,800. That leaves a hundred-point band, the narrowest this desk has recorded, with Thursday’s cash close inside it: thirty-one points off the floor, sixty-nine under the ceiling. The floor did not drift up. It teleported.
- Three separate readings all nominate the same price, and it is beneath the market.Today’s expiration magnet, the dealer put wall and the single most crowded strike in the book — roughly 62,000 calls against 59,000 puts — are all the same line: 7,700. Expiration arithmetic, gamma arithmetic and raw positioning agree, which they almost never do. In futures terms that is 7,712, and Cannon’s own S1 for today is 7,709.
- The rates market is not arguing about September. It has already voted on December.The Street is writing about a one-in-three September hike, and that is right: CME pricing has 33.7% for a move to 3.75–4.00% on September 16 against 66.3% on hold. The number nobody is quoting sits three meetings out. For December 9 the same curve prices 74.2% that rates are higher and 0.0% that they are lower. There is no cut priced at any meeting this year.
- Thursday’s rally was paid for by the rest of the index.The S&P rose 0.72% and the Nasdaq 100 1.43%, but the equal-weight S&P fell 0.29% and RSP fell 0.30%. Software carried everything: IGV +7.74%, with Salesforce +22.58%, Okta +28.63%, CrowdStrike +20.50%, Synopsys +13.39%, Palo Alto +12.83%. Semis managed +1.95%. A one-complex session, and the other four hundred and ninety names funded it.
- Overnight, the leadership is being handed back.NQ is the weakest board this morning at −0.30% while the Dow future is green — the exact mirror of Thursday. Korea’s Kospi fell 1.4% on the chip complex. Marvell beat, raised fiscal 2027 and fiscal 2028, and traded down 7.8% after hours. Rubrik beat and raised and fell 10%. Beat-and-raise stopped being enough somewhere between four o’clock and midnight.
What ties it together is a scheduling accident. There is no 8:30 data today — July PCE landed Wednesday, advance trade Thursday — so the session has exactly one event in it, and it arrives at ten o’clock when Kevin Warsh delivers his first Jackson Hole keynote as chair. The symposium’s published theme is Financial Innovation: Implications for Payments and Policy. Rates are not on the programme. Morgan Stanley’s Michael Gapen and Matthew Hornbach expect him to outline strategic direction and skip the September and December questions entirely; Solus’ Dan Greenhaus put it more bluntly — Warsh “isn’t likely to do much of anything.” The index option market agrees: VIX closed at 14.51 and the nine-day at 12.10, down ten percent on the day. A market that has already declared an event benign is the market with the least room if it is not.
Where everything sits going into the keynote
| Instrument | Last | Change | Note |
|---|---|---|---|
| ES Sep S&P 500 E-mini, live | 7,732.50 | −0.13% | Settled 7,742.50, +11.51 over cash. Trading a quarter of a point off Cannon’s daily pivot of 7,732.25 — Section 04 |
| NQ Sep Nasdaq 100 E-mini, live | 29,605.25 | −0.30% | Settled 29,695.75. The weakest board overnight after being the strongest on Thursday — the rotation is running backwards |
| YM Sep Dow E-mini, live | 53,663 | +0.08% | Green while NQ is red. The Dow gained 0.20% Thursday against the Nasdaq’s 1.57% — now it is collecting the difference back |
| RTY Sep Russell E-mini, live | 3,020.9 | +0.06% | Settled 3,019.0. Cash Russell closed 3,014.34, +0.28% — more than the equal-weight S&P managed |
| S&P 500 cash prior close | 7,730.99 | +0.72% | Range 7,689.89–7,741.27. The record close of 7,816.70 on August 13 is 1.1% above here |
| S&P 500 equal weight prior close | 8,961.42 | −0.29% | Down on a day the cap-weighted index rose 0.72%. A hundred-basis-point divergence in a single session — the number that defines Thursday |
| VIX prior close | 14.51 | −4.60% | Nine-day VIX 12.10, down 10.2%. Front-end volatility priced for nothing to happen at ten o’clock |
| WTI Oct live | 83.19 | −0.41% | Settled 83.53 after +2.15% Thursday — the barrel took back two days of decline in one session. Cannon’s pivot is 82.87 |
| Brent Oct live | 89.15 | −0.61% | Settled 89.70. Spread to WTI above six dollars |
| Gold Dec live | 4,625.90 | −0.82% | Settled 4,664.00. Managed money carries a net long near 142,000 contracts into the keynote |
| Silver Dec live | 68.695 | −1.06% | Settled 69.431 after +1.81% Thursday — the best day on the whole metals board, and it happened quietly |
| US 10Y live yield | 4.680% | +0.8 bp | Grinding up rather than breaking out. Sonders’ tripwire — comfortably above 4.75, quickly — is seven basis points away, and not quick |
| US 2Y live yield | 4.232% | UNCH | 2s10s +44.8 bp. Unchanged into the keynote — the front end has stopped moving and is waiting to be told |
| US 30Y live yield | 5.201% | +1.0 bp | Back above 5.20 despite the Treasury doubling its buyback size nine days ago. Section 07 |
| DXY live | 99.187 | +0.03% | EUR/USD 1.1644, USD/JPY 159.51. Nothing is being pre-positioned in FX |
| Bitcoin live | 79,691 | −0.46% | Settled 80,061 after +1.91%. Cannon’s pivot is 79,749 — price is sitting on it |
| Overnight Asia / Europe | — | — | Nikkei +0.38%, Hang Seng +0.39%, ASX +0.50%, Shanghai flat — and Kospi −1.4%, the only red major, on the chip complex. Europe opens after Thursday’s FTSE −0.79% |
Sentiment & flow gauges
| Gauge | Reading | Prior | What it says |
|---|---|---|---|
| CNN Fear & Greed | 57.9 | 58.2 | GREED Unchanged through a 0.72% index day. A month ago it was 37.9 |
| — the split inside it | 85 / 28 | — | Junk-bond demand at extreme greed, stock-price strength at fear. Credit says risk-on; breadth says the opposite |
| AAII bulls | 32.9% | 35.5% | Against a long-run average of 37.5%. Week to August 26 — taken before Thursday’s software session |
| AAII bears | 44.4% | 39.9% | −11.5 Bears sit 12.9 points above their historical average of 31.5% with the index 1.1% off a record. Sixth straight week bears over bulls |
| Dealer gamma regime | POS | POS | LONG GAMMA Cash closed 80.4 points above the flip — roughly +$47bn of aggregate exposure per one percent move, the largest positive reading in this desk’s log. Levels in Section 04 |
| VIX term structure | +2.24 | +2.15 | Spot 14.51, September future 16.75, October 18.51. Contango steepened again — cheap now, expensive later, which is the market pricing the event risk after today |
| SKEW | 144.05 | 142.96 | Rising while the nine-day VIX fell 10% and the equity put/call printed 0.39. Nobody is buying the at-the-money; somebody is buying the wing |
| Sept FOMC — hike | 33.7% | ~38% | Cut priced at 0.0%. Section 09 |
| Dec FOMC — hike | 74.2% | — | The number the Street is not quoting. Three-in-four that policy is tighter by December 9, and zero that it is easier |
| E-mini S&P, leveraged funds | −281,402 | — | Net short as of August 18 against asset managers +960,595 long. This afternoon’s report covers the week the tape broke out |
The flow read
The gauge table has stopped agreeing with itself. The parts of the market that price credit and options are at extreme greed — junk-bond demand 85, the put/call leg 82, an equity put/call of 0.39, a nine-day VIX at 12.10 that fell ten percent on the day. The parts that measure participation are in fear — stock-price strength 28, momentum 44, the share of the index above its fifty-day average down to 54.5% from a fifty-two-week high of 73.9%. Retail is the tiebreaker and votes bearish: AAII bears 44.4% against a 31.5% average, a sixth week over bulls, surveyed before the software session.
That configuration is not a contradiction; it is a description. What has been bought is index and credit — the instruments you use to own beta cheaply and to fund it — while the underlying breadth has been thinning for weeks. Thursday made it literal: money went into eight software names and came out of everything else, and the index printed a gain anyway. The positioning data agrees from the other direction: leveraged funds were net short 281,000 e-minis as of August 18 against asset managers 960,000 long, so the marginal seller in a squeeze is a hedge that has to be lifted rather than an investor with a view. Global equity funds took in $22.0 billion in the week to August 19 and bond funds $15.4 billion, a twentieth straight week. Nobody is de-risking. They are just not buying breadth.
Yesterday’s calls graded
One event, and it is not a data release
| When | Event | Consensus | Why it matters |
|---|---|---|---|
| Fri 8/28 | Warsh Jackson Hole keynote 10:00 | — | His first as chair, nineteen days before the September 15–16 decision. The symposium’s billed theme is payments innovation, not rates — which is exactly why the reaction function is the only thing anyone will listen for |
| Fri 8/28 | Chicago PMI 09:45 | 57.6 | Prior 57.0. Fifteen minutes ahead of the keynote and about to be forgotten by 10:01 |
| Fri 8/28 | UMich final 10:00 · Baker Hughes 13:00 | 55.2 | Sentiment from 51.0; one-year inflation expectations 4.2% from 4.3%, five-year 3.3%. Released into the keynote — the inflation-expectations line is the only part that survives the collision |
| Fri 8/28 | No 08:30 data | — | July PCE printed Wednesday (headline 3.7%, core 3.3%), advance trade Thursday. The morning is empty until 9:45 — unusual, and it concentrates everything into one hour |
| Mon 8/31 | Month end · Dallas Fed 10:30 | 1.3 | Prior 0.7. The last session against Tom Lee’s end-August target, and the close that frames September’s seasonal argument |
| Tue 9/1–9/3 | ISM Manufacturing · PANW, DELL · ADP, Beige Book · AVGO · ISM Services | — | The week after month-end reloads immediately. Broadcom on Wednesday is the AI complex’s next referendum |
Scenario language describes how desks and pricing frame outcomes. It is not a recommendation.
Every level in today’s session comes from two methods that share no inputs
| Gamma level | SPX | ES Sep · +11.51 | Role in today’s tape |
|---|---|---|---|
| Call wall · ceiling | 7,800.00 | 7,812 | Unmoved for a second session and the heaviest single gamma line on the board, backed by roughly 76,000 calls — the largest open interest at any strike. Sixty-nine points above cash. Cannon’s R3 for today is 7,814.50, three points away from the same price by a completely different method |
| Max pain | 7,700.00 | 7,712 | Today’s expiring contract, the only same-date series on the board. The magnet has crossed back under the price: thirty-one points below Thursday’s close, where it was thirty-nine above on Thursday morning. Reads as a fifteen-point zone, not a point, and it recomputes all session |
| Put wall · floor | 7,700.00 | 7,712 | Jumped two hundred points overnight, from 7,500 where it had sat for five straight sessions. It lands on the same strike as max pain and on the most crowded line in the book — about 59,000 puts against 62,000 calls. Cannon’s S1 is 7,709.00. Untested, and only nine days old as a level |
| Gamma flip | 7,650.61 | 7,662 | Cash closed 80.38 points above the boundary — 1.05%, the widest cushion of this run after two sessions decided by ten points. This is the regime line, and it is now a long way beneath the floor. Cannon’s S3 is 7,656.25 |
Gamma levels from a public dealer-gamma (GEX) model, computed on the August 27 close.
What Cannon’s own levels say
Start with the number that should not happen. The September contract’s daily pivot is 7,732.25. ES is trading 7,732.50. A quarter of a point — one tick — between where the pivot sheet says the day is balanced and where the market actually opened its overnight session. That is not a forecast, it is a description: the tape closed Thursday exactly where the arithmetic says equilibrium is.
From there the two systems keep agreeing. Cannon’s S1 at 7,709.00 sits three points under the magnet and the floor in futures terms; R3 at 7,814.50 sits three points over the ceiling; S3 at 7,656.25 sits six points under the regime line. Pivot arithmetic comes from Thursday’s high, low and close; the option levels come from open interest across four expirations. They share no inputs and this morning they bracket the session at both ends and mark its centre. The practical consequence: R1 at 7,761.75 and R2 at 7,785.00 are the only prices in the upper half of the day with no option structure behind them.
The book behind those levels is stacked on two strikes rather than spread across the chain: roughly 76,000 calls at the ceiling, and at the floor about 62,000 calls against 59,000 puts — more total contracts than any other line, and the reason one price is simultaneously the magnet and the floor. Split by expiry, today’s contract carries about $8.3 billion of gamma, Monday’s $7.6 billion and the September monthly $12.3 billion. Roughly a third of the structure dampening this tape expires at the close.
Trend and structure
Cannon’s trend board is more constructive than it has been in a month. The thirty-year Treasury has held the short-term up arrow it flipped yesterday — long-term still down, but the row that spent weeks pointing down on both horizons no longer does. Both index contracts carry long-term up arrows; gold, silver, copper and bitcoin all carry short-term up arrows; crude’s short-term uptrend survived the decline and has the price back above it after Thursday’s 2.15% settle. Live cattle is now the only row pointing down on both. For a board that has been this letter’s bearish exhibit all month, that is a real thaw — and it arrived in the same week the equal-weight index started falling on up days.
One number for scale before the keynote: the E-mini’s fifty-two-week high is 7,838.50, which is 1.4% above this morning’s futures price and twenty-four points above the call wall in futures terms. The all-time high and the top of the option map are essentially the same place.
The Street-low bear defends her number, and Goldman raises a target without raising a rating
She sat for a full interview on Thursday afternoon, hours after the tape’s best session of the quarter, and did not move. Her 7,100 year-end target is the lowest of more than twenty strategists surveyed, against a street average of 7,901 — and she is keeping it. The quote is the honest part: “I get a lot of flak for my target, more than I’ve gotten in years. Clients used to kind of caution me and say, ‘you don’t want to be too bullish,’ and now they’re like, ‘what is going on, Savita? When are you going to raise your target?’” When the street-low bear is being asked by her own clients to capitulate, that is a positioning datapoint independent of whether she is right.
The substance underneath it is more specific than the headline suggests, and it is not an AI-is-a-bubble argument. On the megacaps: “Some of these stocks are going to be great, some are not. But I think the entire basket warrants an equal weight, if not an underweight.” What would change her mind is narrow and testable — she wants “inklings of real monetization across the capex build-out,” not more capex. And the risk she names is not equity risk at all: rising leverage tied to AI infrastructure financing at a moment when credit spreads are unusually tight. That is the same seam Thursday’s CNN gauge marked at extreme greed on junk-bond demand. Where she wants to be paid instead: large-cap value in energy, financials and software with clean balance sheets — “a great way to screen for areas where you’re buying quality, but you’re not overpaying.”
Rasgon took Nvidia to $400 from $315 and put the reasoning in one sentence: “This quarter should remind NVIDIA investors why they own the stock. Demand remains not only off the charts but continues to apparently accelerate amid what looks to be (once again) the largest impending product cycle in the company’s history as the Rubin ramp kicks off.” He also credited management with getting “out in front of gross margin worries” — the specific thing that had been capping the multiple. On air the same morning he called the guide “a massive raise relative to prior expectations.”
What makes the number worth carrying is where it sits. The same twenty-four hours produced RBC at $330, Wedbush at $345, JPMorgan at $320, Stifel at $315 and a consensus near $321, against a $227.98 close. The Street is not arguing about whether the quarter was good. It is arguing about how many years of it to underwrite.
Goldman raised its Nvidia target to $300 from $285 and left the rating at Neutral. That combination is the most informative single line in the whole post-print sweep. The desk expects the stock to hold its gains, flags fiscal-2027 guidance well ahead of the Street, a medium-term gross margin guide of 72–73%, and better disclosure on customer financial commitments — three genuinely constructive observations — and still will not put a Buy on it at $228 against a $300 target.
Read structurally, that is a house saying the earnings are fine and the entry is not — the same conclusion Subramanian reaches from the top down, the opposite of the one Bernstein reaches from the bottom up, all three looking at the identical quarter. When a target raise and a rating hold arrive in the same note, the disagreement has moved from the numerator to the denominator.
Luria is the sweep’s non-raiser. He reiterated Buy and held his target at $300 — unchanged — while conceding that fiscal-2028 guidance is significantly above consensus. In the same window his published note for CNBC ran under the headline that Nvidia’s blowout earnings “contained some red flags,” with the observation that the numbers “are so big and they’re making really big commitments way out into the future.”
Yesterday this letter recorded him calling the buildout read-through “very positive” hours before the print made him right. He has now been right and stayed put — a rarer posture than capitulating or doubling. His unchanged number against a Street that moved is the cleanest available measure of how much of Thursday was earnings and how much was relief.
Newton’s gate cleared and he immediately widened the frame: his Thursday note argues the triangle breakout “on expanding volume should carry indices back to new highs,” with 7,727 on the S&P and 712 on QQQ now behind the market rather than in front of it. The call is graded in Section 02; what belongs here is the sector observation he attached to it, because it is the one piece of technical work that describes Thursday accurately.
“Software continues to lead Semis,” he wrote, with IGV breaking out to its highest level of the year and Salesforce, CrowdStrike, Fortinet and Palo Alto all up more than seven percent. He also flagged the other half of the tape, which almost nobody printed: utilities, staples and REITs have all hit new yearly lows in the last month, and he does not consider energy or healthcare defensive right now either. A market where the leadership is software and there is no functioning defensive sector is a market with one trade in it — which is a technician’s way of arriving at the same place the equal-weight index arrived on Thursday.
The roster, and where it has landed
| Voice | Firm | Stance | Where it stands this morning |
|---|---|---|---|
| Savita Subramanian | Bank of America | BEAR 7,100 | Street-low target defended on the record; megacap basket equal weight “if not an underweight.” Section 05 |
| Stacy Rasgon | Bernstein | BULL | Nvidia to $400 from $315 — the highest post-print number on the board. Section 05 |
| James Schneider | Goldman Sachs | NEUT | Target to $300, rating held at Neutral. The earnings are fine; the entry is the argument. Section 05 |
| Gil Luria | D.A. Davidson | BULL | Buy reiterated, target unchanged at $300 — the sweep’s only non-raiser. Section 05 |
| Mark Newton | Fundstrat | BULL | 7,727 gate cleared; software leading semis, and no working defensive sector. Section 05 |
| Ed Yardeni | Yardeni Research | BULL 8,400 | Thursday’s note argues the AI buildout is stimulating the whole economy — non-hyperscaler revenue growing 138% against hyperscalers’ 102%. Section 08 |
| Tom Lee | Fundstrat | BULL 8,000 | Strong finish to August, then “might see a correction after” — and he calls today the month’s final clearing event. Graded in Section 02 |
| Ben Reitzes | Melius | BULL | Street-high $420 set Thursday on a ~$279B purchase-commitment backlog. Carried, not restated |
| Matt Bryson | Wedbush | BULL | Nvidia to $345 from $330. The firm’s note is his, not Dan Ives’ — a byline this desk has seen misassigned twice this week |
| Michael Gapen & Matthew Hornbach | Morgan Stanley | NEUT | Expect Warsh to outline strategic direction and skip the September and December questions. Section 09 |
| Dan Greenhaus | Solus | NEUT | Warsh “isn’t likely to do much of anything” today. Section 09 |
| Mike Wilson | Morgan Stanley | CAUT 7,800 | His named near-term risk was oil rising; it settled up 2.15% Thursday. Carried, no new note since August 24 |
| Michael Hartnett | Bank of America | CAUT | Sell signal still on and the trade still long gold. The Flow Show expected last night had not surfaced on any carrier by this morning — a second consecutive week without a citable print |
| Jeremy Siegel | WisdomTree / Wharton | CAUT | His 5% ten-year threshold is intact and the thirty-year is above 5.20%. New commentary Monday |
| Ankur Crawford | Alger | NEUT | The AI trade’s biggest risk is the midterms, via the data-center backlash. Carried from Wednesday |
| Jonathan Krinsky | BTIG | BEAR | Seasonal-peak call graded OPEN in Section 02. No citable note for a sixth consecutive run |
| Scott Rubner | Citadel Securities | DARK | Nothing published since August 11. His August checklist described a re-leveraging tape; this is the fourth week without a number to hold him to |
| Tony Pasquariello | Goldman Sachs | DARK | Third straight run with no trusted carrier. His last weekend note argued the breakout was intact but had work to do after Labor Day — that week starts Tuesday |
Two seats on this roster remain vacant with no fresh call attributed to either: Christopher Harvey, formerly of Wells Fargo, and Jonathan Golub, formerly of UBS.
Two institutions, one long end, and a curve that has stopped pricing cuts
Begin with the number that reframes everything else. The Fed funds curve prices a 33.7% chance of a September hike — the figure the Street has been quoting all week — and a 74.2% chance that policy is tighter by the December 9 meeting. At every single 2026 meeting, the probability of a cut is zero. This is not a market debating the direction of the next move. It is a market that has settled the direction and is arguing about the date. Equity multiples at 1.1% off a record high are being set against that curve, and the two facts have not been in the same paragraph often enough this month.
The data underneath it explains why. July core PCE ran 3.3% year over year with headline at 3.7%, and the Cleveland Fed’s nowcast has August core PCE accelerating to 3.40% — while the Atlanta Fed’s GDPNow has third-quarter real growth at +4.6%. Hot inflation is one problem; hot inflation on top of four-and-a-half percent real growth is a different one, because it removes the argument that patience is cheap. Two regional presidents said so on the record Thursday: Cleveland’s Hammack called inflation “running too hot,” and Kansas City’s Schmid said inflation is “too hot” and rates “too accommodative.” The Fed is not in blackout until September 5, so those were deliberate.
The second axis is institutional, and it is why this keynote is not routine. On August 19 the Treasury doubled its ten-to-thirty-year buybacks to $4 billion per operation. The long end erased the entire move inside twenty-four hours. Scott Bessent’s response was to promise more and to say the quiet part: “Part of it is signalling here, and to show that we believe that the yields don’t reflect the underlying fundamentals.” Critics call it an Operation Twist that cannot work against 2026’s money growth; the simpler observation is that the thirty-year sits at 5.201% this morning, nine days and one doubling later. The intervention has been tested and the price has not moved.
Which sets up the collision. Warsh’s stated preference is for market-set yields with the Fed out of the way — he has argued the long end was doing some of the central bank’s work for it. He walks to the podium this morning with the Treasury actively managing that same long end, a hot inflation print behind him, two of his own regional presidents on the record calling policy too loose, and a front end that has stopped pricing cuts entirely. Loretta Mester, the former Cleveland president, put it on CNBC on Thursday: the Treasury’s bond intervention “complicates things a bit” for him. That is the whole macro story of today in six words.
On the commodity side the freshest item is geopolitical. Reuters reported Thursday evening that the administration is negotiating long-term US access to seventeen Venezuelan oil fields through a lease-then-auction structure, and Bloomberg separately reports Venezuela is weighing leaving OPEC after sixty-six years as a founding member. At roughly 1.25 million barrels a day the supply arithmetic is small; the signal is not, because a producing member contemplating exit changes how quota discipline gets priced. It lands in the week WTI took back two days of decline in a single 2.15% session.
The software re-rating, and what the after-hours tape did to it
The session. Thursday was one complex and almost nothing else. Salesforce closed +22.58% at $252.05, Okta +28.63% at $172.91, CrowdStrike +20.50% at $227.96, with Veeva, Synopsys, Palo Alto, ServiceNow and Fortinet all between +9.7% and +15.2%. Against that, semiconductors managed +1.95% — AMD down 0.89%, Micron down 0.32%, even as Broadcom added 4.49% and TSMC 2.30%. Nvidia closed +8.74% at $227.98, its first day-after-earnings gain in five quarters, and slipped 0.80% after the bell.
What happened underneath. The equal-weight S&P fell 0.29% on a day the cap-weighted index gained 0.72%. That gap is the story: not risk-on, but a rotation violent enough to move the headline while the median stock declined. Newton supplies the other half — utilities, staples and REITs have all made new yearly lows in the past month. There is no sector doing the job defensives are supposed to do, which means the only way out of the AI complex right now is cash or Subramanian’s value screen.
The after-hours tape is where the discipline showed up. Marvell reported a record quarter — revenue $2.739 billion, up 37%, with fiscal-2027 guidance raised by $500 million and fiscal-2028 by $1.5 billion — and traded down 7.8% to $222.62. Rubrik beat and raised and fell 10%. Autodesk beat and fell 5.15%. Workday beat and fell 5%. SentinelOne beat and fell 4.1%. Earlier in the regular session Best Buy beat and raised and closed −4.44%, and HP printed record revenue and earnings and closed −2.92% on a sixteen percent decline in PC shipments and memory-cost pressure. Seven beat-and-raises, seven negative reactions. The market spent Thursday paying up for eight software names and spent Thursday night refusing to pay for anything else, which is a more selective tape than the index gain suggests.
Two items to keep on the desk. Nvidia has reportedly agreed to buy Hugging Face for $12.9 billion — reported by The Information, confirmed by neither party, described elsewhere as unsigned, and being priced as done. And Yardeni’s Thursday note carried the most useful figure in the whole print for sizing the cycle: non-hyperscaler revenue grew 138% year over year against hyperscalers’ 102%. Demand is broadening away from the five customers everyone monitors — either the best news in the quarter, or, depending on who is financing the new buyers, the thing to check first.
A first keynote with the wrong theme printed on the programme
Warsh delivers his first Jackson Hole keynote as chair at ten this morning, nineteen days before the September 15–16 decision. Rates are formally not on the symposium’s agenda, which is why the reaction function is the only thing the market will listen for. Morgan Stanley’s preview expects the balance sheet, the inflation-targeting framework, communications, AI and productivity — and nothing at all on September or December. Greenhaus expects the same and less: he “isn’t likely to do much of anything.”
The pricing into it is the asymmetry. The target range is 3.50–3.75% and has not moved all year; the last change was a cut in December 2025. September prices 33.7% hike, 66.3% hold, zero cut. December prices 74.2% hike, zero cut. The Fed is not in blackout, and two regional presidents used that freedom on Thursday to say inflation is too hot and policy too accommodative. Against that, the nine-day VIX closed at 12.10 — down ten percent on the day — while the September future sits 2.24 points over spot and SKEW rose to 144. The index option market has priced today as a non-event and is paying up for what comes after it. If he says nothing, that positioning is correct and expensive. If he supplies a framework hawkish enough to make the December number look conservative, the cheapest thing in the room this morning was the thing nobody wanted.
The closer
The dampening that makes today feel safe expires at the close, and the floor holding it up is nine days old.
Positive gamma is being read this morning as a cushion, and for the next six hours it is one — roughly $47 billion of aggregate exposure per one percent move, the largest positive reading this desk has logged, with dealers mechanically selling strength and buying weakness inside a hundred-point band. But look at where that structure lives. About a third of it, some $8.3 billion, sits in contracts that expire at four o’clock this afternoon, and the strike doing most of the work is both today’s expiration magnet and a floor that arrived overnight from two hundred points lower after five sessions of sitting still. Nothing tested it. It was built, not defended. The market’s stability today is therefore being supplied by open interest that stops existing tonight, on a level with no history, into a long month-end weekend — with the next real structure fifty points below it and a thin book in between. The stillness is real. It is also scheduled to end.
Everyone is trading the September meeting. The curve settled that argument and moved on to December.
The one-in-three September hike has been quoted in every preview this week, and it deserves to be — it is the highest odds of a tightening at a scheduled meeting since the cycle turned. But it is the wrong number to position around, because it is nearly a coin flip and the market knows it. Three meetings out the same curve is not ambivalent at all: 74.2% that rates are higher on December 9, and zero — not low, zero — that they are lower at any meeting this year. That is a rates market that has stopped hedging the easing case entirely, and equities have not priced the same thing. A three-quarters probability of tighter policy by December does not sit comfortably beside a nine-day VIX at 12.10 and an equity put/call at 0.39, and the reconciliation does not require Warsh to be hawkish today. It only requires him not to be dovish, which is exactly what his own previewers expect. The point the consensus is missing is not a view on the speech; it is that the speech may not matter and the December number will anyway.
The index went up on Thursday because the median stock went down, and the money for the next leg is not on any income statement that got re-rated.
The cap-weighted S&P rose 0.72%; the equal-weight S&P fell 0.29%. That hundred-basis-point split in one session is not a curiosity, it is the mechanism — capital left four hundred and ninety names to buy eight, and the index printed a gain on a day the median stock declined. The share of the index above its fifty-day average is down to 54.5% from a fifty-two-week high of 73.9%, there is no working defensive sector, and CNN’s stock-price-strength gauge sits at 28 in fear while its junk-bond-demand gauge sits at 85 in extreme greed. Now add the part the equity desks re-rated around rather than through: off-balance-sheet AI commitments across big tech have reached roughly $3.1 trillion against something near $600 billion of reported capital expenditure. Subramanian named exactly this — AI-infrastructure leverage arriving while credit spreads are unusually tight — and her own clients are asking her to capitulate on it. The equity market has decided the growth is real. Nobody has yet had to decide who is funding it, and that decision gets made in spreads, on a day when the credit gauge is the most greedy reading on the board.
eli@cannontrading.com · cannontrading.com
Legal Disclosure & Risk Warning — Cannon Trading Company
This publication is provided by Cannon Trading Company for informational and educational purposes only. Content may include market commentary, technical observations, analyst opinions, and aggregated material derived from publicly available sources. While such information is believed to be reliable, Cannon Trading Company does not author, independently verify, endorse, or guarantee the accuracy, completeness, or timeliness of any third-party information referenced or summarized herein.
The information, opinions, market data, and commentary contained in this publication are subject to change at any time without notice and do not constitute investment advice, a solicitation, or a recommendation to buy or sell any security, futures contract, option on futures, foreign currency transaction, or any other financial instrument.
Past performance is not indicative of future results.
Trading Futures, Options on Futures, retail off-exchange foreign currency transactions, and other derivatives involves substantial risk of loss and is not suitable for all investors. You may lose all or more than your initial investment. Carefully consider whether trading is appropriate for you in light of your experience, objectives, financial resources, and other relevant circumstances.
Cannon Trading Company does not guarantee any profits and makes no representation that the strategies, ideas, analyses, or information presented will result in profitable trades or avoid losses. Any market views, analyst calls, forecasts, or third-party commentary referenced reflect the opinions of their respective authors and may or may not align with the views of Cannon Trading Company.
HYPOTHETICAL PERFORMANCE RESULTS HAVE MANY INHERENT LIMITATIONS, SOME OF WHICH ARE DESCRIBED BELOW. NO REPRESENTATION IS BEING MADE THAT ANY ACCOUNT WILL OR IS LIKELY TO ACHIEVE PROFITS OR LOSSES SIMILAR TO THOSE SHOWN. IN FACT, THERE ARE FREQUENTLY SHARP DIFFERENCES BETWEEN HYPOTHETICAL PERFORMANCE RESULTS AND THE ACTUAL RESULTS SUBSEQUENTLY ACHIEVED BY ANY PARTICULAR TRADING PROGRAM.
ONE OF THE LIMITATIONS OF HYPOTHETICAL PERFORMANCE RESULTS IS THAT THEY ARE GENERALLY PREPARED WITH THE BENEFIT OF HINDSIGHT. IN ADDITION, HYPOTHETICAL TRADING DOES NO INVOLVE FINACIAL RISK, AND NO HYPOTHETICAL TRADING RECORD CAN COMPLETLEY ACCOUNT FOR THE IMPACT OF FINANCIAL RISK IN ACTUAL TRADING. FOR EXAMPLE, THE ABILITY TO WITHSTAND LOSSES OR TO ADHERE TO A PARTICULAR TRADING PROGRAM IN SPITE OF TRADING LOSSES ARE MATERIAL POINTS WHICH CAN ALSO ADVERSELY AFFECT ACTUAL TRADING RESULTS. THERE ARE NUMEROUS OTHER FACTORS RELATED TO THE MARKETS IN GENERAL OR TO THE IMPLEMENTATION OF ANY SPECIFIC TRADING PROGRAM WHICH CANNOT BE FULLY ACCOUNTED FOR IN THE PREPARATION OF HYPOTHETICAL PERFORMANCE RESULTS AND ALL OF WHICH CAN ADVERSELY AFFECT ACTUAL TRADING RESULTS.
Cannon Trading Company is registered solely as a commodities broker. Nothing contained herein constitutes the provision of investment advisory services.
© 2026 Cannon Trading Company, Inc. All rights reserved.
cannontrading.com • 1-800-454-9572 • (310) 859-9572