The week ahead — September 6, 2026: An oil shock the meeting cannot see

Producer prices Thursday, Oracle and Adobe after Thursday's close, August CPI Friday 9/11 with core consensus at +0.3% m/m and 3.1% y/y. All of it lands four sessions before the Sept 16 FOMC, which futures price near 60/40 for a hike. Levels, scenarios and what would change them.

Last week, graded

Grading last Sunday: we were defensive into payrolls on two-sided risk. The hawkish tail printed, +162k against a 53k consensus, and the tape gave back 0.39% Friday for +0.11% on the week. Right tail, wrong size, and neither invalidation level traded. Verdict: mixed.

• This week — mixed. We said: Defensive into a payrolls week on asymmetry, not on trend: no volatility cushion carried into a dense calendar, with both tails unfriendly. What happened: The hawkish tail is the one that printed. August payrolls printed 162k against a consensus near 53k, unemployment held at 4.1%, and July was revised from a 23k decline to a 21k gain. September hike odds moved from 49.4% on Thursday to 60.4% on Friday on CME FedWatch. SPY gave back 0.39% on Friday and still finished the week +0.11% at 770.19. Neither invalidation level traded: no daily close above the range high and no close below the 50-day. The bull branch did not print either, because IWM finished at 296.01 and never reclaimed its own.
• This month — too early to call. We said: Neutral on September, because the month turns on one two-sided scheduled event with a dot plot attached. What happened: One of five weeks is in and neither invalidation level traded. The distribution did move: hold-versus-hike went from close to even to roughly 60/40 for the hike after payrolls.
• This quarter — too early to call. We said: Neutral on Q4, with rising earnings estimates set against stretched positioning and an election inside the window; the range was the uncertain part, not the direction. What happened: Seven days into a thirteen-week horizon, nothing in the invalidation set has resolved: IWM has not held 305.18, SPY has not lost its 200-day, and 2027 estimates have not turned down.
• This year — too early to call. We said: Constructive on twelve months, on contracted capex funding double-digit 2027 earnings growth with the multiple assumed flat. What happened: Not gradeable in seven days, and the tracked variable has not moved: 2027 estimates have not been revised down. The rate backdrop did change, with the ten-year touching its highest since October 2023.

Week horizon: 0 right, 0 wrong, 1 mixed over 1 graded note. Month, quarter and year: 0 graded, 3 too\_early over 1 archived note. Archive holds 1 prior note (2026-08-30).

The setup

Range. SPY has held a box of about 2.6% for twenty sessions and the 20/50 spread is narrowing on all three indices, so the edges of that box are what decide the next move. SPY sits above every major average with a positive weekly MACD histogram. QQQ closed a hair above its 20-day with the weekly histogram still negative. IWM is under its 50-day and under its 10-week average. Compression with split internals is a range, and the edge that gives way first is the information.

What is actually going on

The reflex that has governed every dip since 2024 stops being useful into this week, and that is the single most important thing to carry into September. For two years, weak data meant help was coming. Friday broke that link.

August payrolls landed Friday at 162,000 against a consensus near 53,000, per the BLS release and CNBC's coverage of it. The unemployment rate held at 4.1%. July, first reported as a 23k decline, was revised to a 21k gain. Participation rose to 61.6% from 61.4%, per MT Newswires, so the labour force grew and the rate still held. Odds of a quarter-point hike at the September meeting went from 49.4% on Thursday to 60.4% on Friday on CME's FedWatch tool. A day earlier, remarks from a Fed governor conditioning a hold on continued progress toward 2% had pushed the same number down from roughly 70% to about 50%. The path moved twenty percentage points in two sessions on two pieces of information. That is not a market with a view. That is a market waiting to be told.

Here is the part that has not been priced, because it cannot be yet. Crude had its strongest week since mid-July, with WTI settling up 9.7%, after US strikes on Iranian tankers followed attempted IRGC attacks on commercial shipping in the Strait of Hormuz, per Axios and EnergyNow. Roughly a quarter of seaborne crude and about a fifth of LNG transit that strait, per the Congressional Research Service. The move began on September 1. August's inflation data was collected in August. So the print that arrives Friday cannot contain any of it, and the September data that would contain it does not publish until after the committee has already voted. The committee decides with an energy impulse it can read in the newspaper and not in its own numbers.

That matters more than usual because the services side is already running hot underneath. August ISM services printed 55.4, the strongest expansion in six months, and its prices-paid subindex spiked to a four-year high of 72.6, per Benzinga's coverage of the release. Energy costs feed services costs with a lag. A committee that already has three members who dissented in July preferring a hike does not need much more to justify one.

So the reaction function has inverted, and that is the regime change worth stating plainly. Since 2024, a weak number was good news for stocks because it pulled forward easing. Now a strong number raises hike odds, a hot inflation number raises them further, and a weak number arrives at a committee that has publicly put price stability first and cannot answer it quickly. Both surprises push the rate path the same way. The old cushion is gone.

What the tape did with that is the second story. The index barely moved. The internals were violent. Semiconductors and memory ripped on Friday while software was gutted, per Benzinga's session wrap: Astera Labs added 12%, SanDisk 10.3%, KLA 8.2%, while Guidewire fell 21%, UiPath 16.7%, Autodesk 7.9% and Adobe 6.3%. That is a market absorbing a hawkish surprise by rotating instead of de-rating. It is a sign of capacity, not of complacency. It is also a warning, because when the rotation valve runs out of places to rotate into, the absorption stops.

The live argument among strategists is no longer about AI demand. It is about how the buildout is financed. Oracle reports Thursday and it is the cleanest test case available. Remaining performance obligations stood at 638bn at the last report, per MT Newswires. Nobody disputes that the demand is there. Against that, the company raised 43bn in debt and 5bn in equity last year, plans another 40bn, and RBC's note this week describes it funding the buildout partly through successive headcount reductions, flagging execution risk in cloud applications. Both readings are honest. Contracted demand of that size is real, and financing it with debt into a rising rate path is a different proposition from financing it out of cash flow. Adobe reports the same evening carrying a different version of the same anxiety, over whether AI disrupts the incumbent or enriches it. Its stock fell 6.3% on Friday after it named an insider chief executive days before the print.

The last thread is breadth, and it has quietly reversed. US small caps returned 22.93% in the first half of 2026 against 9.55% for large caps, their best first half on record, per LSEG's FTSE Russell data. The share of small caps outperforming their own index sat in the 97th percentile. That was the broadening everyone had asked for since 2023. It has since stalled. Small caps are the only one of the three major US indices sitting below their own intermediate averages going into this week, and they have given back ground for the quarter while still leading for the year. Whether that broadening survives a hiking cycle is the question the fourth quarter answers, and nothing before September 16 settles it.

The week ahead

• 2026-09-10Producer Price Index (August), 08:30 ET
• 2026-09-10Oracle fiscal Q1 2027 and Adobe fiscal Q3 2026, both after the close. Consensus for Oracle near 1.74 EPS on 19.13bn revenue per FactSet, with RBC modelling 1.65 and 19.10bn. Remaining performance obligations were 638bn at the last report. Earnings dates move; treat as scheduled.
• 2026-09-11Consumer Price Index (August), 08:30 ET. Core consensus +0.3% m/m and 3.1% y/y, with Goldman at 0.36% and a 3.13% annual rate. NOTE: our local econ list carries this as 09-10; the BLS release schedule says 09-11 and it is the authority on its own release.
• 2026-09-16FOMC decision, 2:00pm ET (meeting Sept 15-16), with the Summary of Economic Projections and dot plot. Priced near 60/40 for a 25bp hike on CME FedWatch as of the Sept 4 close; Kalshi has carried a lower number, so the venue matters when quoting it.
• 2026-09-18September quarterly options expiration
• 2026-11-03US midterm elections

The next five sessions

This week

The week's information is stacked into two days. Producer prices land Thursday morning, Oracle and Adobe report after Thursday's close, and August CPI publishes Friday at 8:30 ET, per the BLS release schedule. Core CPI consensus sits at +0.3% month over month and 3.1% year over year, with Goldman modelling 0.36% and a 3.13% annual rate, per Morningstar. One housekeeping note, because it matters for anyone planning around the print: our own local economic calendar carries CPI as Thursday. The BLS schedule says Friday, and the BLS is the authority on its own release.

What makes Friday's number awkward is what it excludes. The August survey period closed before crude moved, so a hot print is a tariff-and-services print with no energy in it, and a cool print buys the committee nothing about what comes next. Either way, four trading sessions separate the release from the decision.

Price has compressed into that. SPY has spent the last twenty sessions between 759.48 and 779.37, a box of about 2.6%, and closed the week at 770.19 near the middle of it. Its 10-day at 767.33 and 20-day at 769.05 now sit within half a percent of spot, which is what a coiled tape looks like on the daily frame. Daily ATR is 4.49, or 0.58% of price, so the box is roughly five average sessions wide in either direction. QQQ closed at 718.96, a hair above its 20-day at 717.51, and printed a doji. IWM closed at 296.01, still under its 50-day at 297.01, and it is the only one of the three that has to reclaim a level instead of defending one.

Open interest for Monday's expiry, as of the OCC's September 3 file, stacks the largest call line at 770 and the largest put line at 737. That is open interest and nothing else. We carry no greeks, and this is not a gamma read.

The next four weeks

This month

The month ahead is decided on the 16th at 2:00pm ET, when the decision lands with the Summary of Economic Projections attached. That last detail is what lets a hold reprice the path as much as a hike would. The committee can leave the rate alone and still move the whole 2027 dot path, so a reader watching only the decision line will miss half the information in the release.

The distribution has tilted since last Sunday. A week ago futures and prediction markets treated hold against hike as close to even. It now sits near 60/40 for the hike on CME's FedWatch tool, and it got there through two opposite shocks in two days: remarks from a Fed governor conditioning a hold on continued progress toward 2%, which pushed the odds down toward 50, and then the payrolls print, which pushed them past 60. An unanchored path is what widens the reaction to Friday's CPI.

The month also carries an asymmetry in the data calendar itself, and this is the part that has not been priced because it cannot be yet. August's inflation data closed its survey window before crude moved. September's data, which would carry the energy pass-through, publishes in October. The committee's decision is therefore made with a clean view of a period that no longer describes the present, and the October and November communications are where the impulse actually shows up.

Seasonality is the weakest argument available here and it is worth naming as such. September's reputation is real in a long sample and close to useless in any single year. The month's actual signal is whether SPY resolves outside 759.48 or 779.37 on a weekly close, because a twenty-session compression that resolves on a scheduled catalyst tells you which side of the argument was carrying the larger position.

The next quarter

This quarter

Q4 opens with the earnings line still doing the work. FactSet's third-quarter estimate calls for 28.3% index earnings growth, after a second quarter that ran above 50%, per Benzinga's summary of the data. Full-year consensus sits near 11% revenue growth and 23% earnings growth. That is the constructive side of the quarter and it is not in dispute.

The contested side is the discount rate and who pays for the buildout. The ten-year touched 4.81% this week, its highest since October 2023. A quarter that hikes into that is a quarter where the longest-duration earnings in the index get marked against a higher curve, and those are the same AI-infrastructure earnings carrying the growth. Oracle's balance sheet on Thursday is the first public read on whether the market treats debt-funded capacity as an asset or a liability.

Sector leadership already reflects some of this. Energy is the year's best sector, up roughly 43% through August, per Investing.com's sector tracking, and it has earned that on cash flow generated today, not capacity built for tomorrow. Industrials have joined it. That is a value-leaning rotation, and it is the shape a market takes when the cost of future money rises.

The breadth question is the one that decides the quarter's range. US small caps returned 22.93% in the first half against 9.55% for large caps, their best first half on record, per LSEG's FTSE Russell data. They have since given ground and now sit below their 50-day. IWM holding above 305.18 while the large-cap index holds its own highs would say the broadening survived a hiking cycle. IWM failing to reclaim 297.01 says the opposite, and it says it early. Midterms land November 3.

The next twelve months

This year

Into the next twelve months the argument has not changed shape. It has changed price. Published year-end targets now run from 7,100 at BofA, which has flagged too many red flags on US equities, to about 8,250 at the bullish end, with a median near 7,850 across 19 firms tracked by Investing.com. Citigroup lifted its own to 8,100 from 7,700 on earnings and on AI spending broadening past technology. That spread is not a disagreement about whether earnings grow. It is a disagreement about whether the multiple survives a rising cost of money.

That is the reason to carry this horizon at lower confidence than a rising-earnings story would suggest on its own. A year built on flat multiples and real profit growth works while the discount rate is still. The discount rate is not still. The ten-year is at its highest in almost three years and the September meeting is priced closer to a hike than a hold.

The variable to track is not price. It is 2027 index EPS. If those estimates keep rising, the constructive read stands even through a hiking quarter, because price can lag earnings for a long stretch and still be right. If they roll over for two consecutive months, the whole argument loses its foundation, and the level of the index would be the last thing to tell you.

The financing question sits underneath all of it. Contracted AI demand is visible and large. How much of it is funded from operating cash flow, and how much from debt raised against a rate path moving the wrong way, is what the next two reporting seasons answer. Thursday night is the first data point of that season, and it arrives from a company whose remaining performance obligations and whose debt stack are both among the largest in the complex.

The levels

Into the week SPY carries a twenty-session box of 759.48-779.37, closing at 770.19, with 756.86 the shelf beneath it. QQQ sits just above its 20-day at 717.51. IWM at 296.01 is the outlier, still under its 50-day. Which of the three resolves first carries the information.

• SPYsupport 759.48resistance 779.37The pivot pair sits on top of spot: 10-day 767.33 and 20-day 769.05 are both within half a percent, and Friday printed a harami inside the prior body with the close under the open. That is compression, not direction. The box is 759.48 to 779.37 and ATR14 is 4.49, so either edge is about two average sessions away. The 20/50 spread is +1.61% and narrowing by 0.12pp on the week. That is the mechanical version of the same statement.
• QQQsupport 711.09resistance 717.51It closed a doji a hair above its 20-day at 717.51, which makes that line the hinge for the week: above it the 50-day at 711.09 is a cushion, below it that same line is what the tape has to answer to. The weekly MACD histogram is still negative at -2.85, so the weekly frame has not repaired even though the daily structure is intact. ATR14 at 6.71 is 0.93% of price, the widest of the three, so it needs the least help to resolve.
• IWMsupport 289.97resistance 297.01The only index that has to reclaim rather than defend. It sits under the 10-day at 296.14, the 50-day at 297.01, the 20-day at 298.87 and the 10-week average at 296.83, so those form a single cluster about one percent wide that has to be taken back before any breadth-repair argument can be made. Friday's candle engulfed the prior body and closed above the open. That is where reclaim attempts start. It is still 8.84% above its 200-day at 271.98, so this remains an intermediate break.

What would change it

Watch five things, none of them a forecast: SPY closing outside its twenty-session box; IWM reclaiming and holding its 50-day; QQQ's weekly MACD histogram turning positive; Friday's CPI moving September hike odds off 60/40; 2027 index EPS estimates starting to fall.

Leadership

• Leading: Semiconductors and memory: SMH closed Friday at 566.99 against a 552.60 prior close, with Astera Labs +12%, SanDisk +10.3% and KLA +8.2% on the session, Energy on the year, the best S&P sector at roughly +43% through August, though XLE itself closed lower on Friday as crude paused, Industrials: XLI closed at 175.24 against a 174.56 prior close, the only other sector green with technology on Friday
• Lagging: Software: Guidewire -21%, UiPath -16.7%, Autodesk -7.9%, Adobe -6.3% on Friday, the clearest de-rating in the tape, Small caps: IWM below its 10-day, 20-day, 50-day and 10-week average, and -1.48% on the quarter while still +20.75% on the year, Equal weight: RSP closed at 219.01 against a 220.05 prior close, so the average stock lost more than the cap-weighted index on Friday

The leadership story this week is inside the index, not between indices. September dispersion is only 0.40pp between IWM (+0.71%) and QQQ (+0.31%), the narrowest gap in weeks, while the single-session spread between semiconductors and software ran above thirty percentage points at the name level. A tape that rotates that hard under a quiet index is expressing a rate view: hardware with contracted backlog is being bought and long-duration software multiples are being sold. That has a direct read for the week ahead. A cool CPI print releases the software complex, which has the most compressed valuations after Friday. A hot one extends the same rotation until the hardware side runs out of buyers. Small caps sit outside both trades and are the honest test of whether any of it is broadening.

Cross-asset

• USO (WTI proxy, NOT spot crude): week +9.45% · MTD +6.18% · vs 200-day +32.07%. The most important cross-asset number in the note. A 9.45% week on the proxy puts crude's move at the top of everything else on the board, and it started on September 1, after August's inflation survey window had already closed. The pass-through is therefore a fourth-quarter data story, not a Friday story.
• VIXY (VIX FUTURES ETF, NOT spot VIX; we carry no spot vol): week -3.74% · MTD -1.62% · vs 200-day -34.67%. Volatility kept compressing through a hawkish payrolls surprise, which is the clearest evidence that the tape absorbed the print by rotating. It also means the cushion going into CPI and the FOMC is thinner than it was a week ago, not thicker.
• TLT (20y+ Treasuries): week -0.42% · vs 200-day -2.81%. The long end gave ground on the week and sits below both its 50-day and 200-day. Watch the belly against the long end for whether the market is repricing the policy path or the term premium; the two carry different implications for equity duration.
• IEF (7-10y Treasuries): week -0.29% · vs 200-day -1.31%. The belly sold less than the long end on the week, which is a mild term-premium tilt rather than a pure policy-path move. A CPI surprise that reverses that ordering would be the cleaner policy signal.
• GLD (gold proxy, NOT spot gold): week -0.52% · vs 200-day -2.09%. Gold gave ground as the dollar firmed on the jobs data, which is the textbook reaction and a useful confirmation that Friday was a real-rate move and not a growth scare.
• UUP (DXY proxy, NOT the dollar index): week -0.35% · vs 200-day +1.66%. The dollar proxy is still net lower on the week despite Friday's firming, so the hawkish repricing has not yet shown up as a sustained currency move. That divergence resolves one way or the other on CPI.

Sources

• CNBC: US payrolls rose much more than expected in August (2026-09-04)
• BLS: Employment Situation Summary, August 2026
• MT Newswires, 2026-09-04, via the Robinhood news feed: pre-bell wrap, participation 61.6% from 61.4%
• BLS: CPI release schedule — August CPI publishes 2026-09-11, 08:30 ET
• Morningstar: August CPI forecasts point to sticky inflation and tariff pressures
• CME FedWatch via CNBC: September hike odds 60.4% Friday from 49.4% Thursday
• DeFiRate: September Fed decision odds, Kalshi and Polymarket tracker
• EnergyNow: oil posts its biggest weekly gain since July as the war premium rebuilds
• Axios: US strikes Iranian oil tankers for the first time (2026-09-02)
• Congressional Research Service R45281: the Strait of Hormuz and its share of seaborne crude and LNG
• Benzinga, 2026-09-03, via the Robinhood news feed: ISM services 55.4, prices paid at a four-year high, 10-year at 4.81%
• Benzinga, 2026-09-04, via the Robinhood news feed: Friday session internals and sector moves
• MT Newswires, 2026-09-04, via the Robinhood news feed: RBC on Oracle's AI buildout, layoffs and RPO
• Investing.com: S&P 500 sector performance, energy leading 2026
• Investing.com: the updated S&P 500 year-end analyst consensus across 19 firms
• LSEG / FTSE Russell: small caps step up in a broadening US equity rally
• Yahoo Finance: Adobe to announce Q3 FY2026 results on Sept 10, 2026

Prices, ranges, moving averages and open interest in this note come from our own market-data bars through 2026-09-04. Every external claim names its source above. This note is forward-looking market research: it describes scenarios and the conditions that would confirm or invalidate them. It is not investment advice, not a recommendation, and states no forecast, price target or trade.