Weekly outlook · week of September 8, 2026

Sixty Percent, the Other Way

Three weeks after markets priced a September Fed hold as the base case, they now price a rate hike. Stronger payrolls, Chair Warsh's inflation framework, and oil near $97 have shifted the burden of proof.

Author Brandon Leon Posted Tuesday, September 8, 2026 Data Equities through Friday, September 4; oil through Monday, September 7 (Labor Day — U.S. stock markets closed, oil traded); Tuesday-morning levels noted where used Coverage Fed policy, oil, Nvidia, Dell, and this week's market tests

Three weeks ago, markets priced roughly a 60% chance that the Fed would hold rates steady in September. We argued that confidence was misplaced. Today, futures price roughly a 60% chance of a quarter-point hike at the September 16 meeting.¹

The reversal came from two developments. First, Chair Kevin Warsh used his Jackson Hole speech on Friday, August 28 to focus the market on PCE inflation, which he said was still running at 3.7%, and to argue that financial conditions were not restrictive. Then August payrolls, reported Friday, September 4, came in at 162,000 versus roughly 53,000 expected, while July's initially negative result was revised to a gain.²

Oil adds another complication. Brent is trading near $97 after renewed attacks around the Strait of Hormuz over the holiday weekend, but our actual test remains unchanged: it must settle above $95 on Friday to show that the move has staying power.⁴

This note also closes the loop on the calls already on the record. Nvidia cleared its published revenue threshold. Dell disproved the margin concern at the center of our original framework. Walmart remains mixed. And the Vistra order was never placed. That last point matters: a research process should be judged not only by its forecasts, but also by whether it does what it says it will do.

The record, kept straight

The tests in this series are wires — pre-committed, observable triggers published before the event and graded against their exact wording afterward. The August 26 Nvidia note promised its grade “in Sunday's outlook.” That Sunday was August 30. No note ran August 30, and none ran September 6. This note is late, and the missed publishing cadence is part of the record as well — logged here the same way a missed wire would be.

1. The reversal: 60% hold became 60% hike

The path is worth tracing, because it shows the market talking itself around in eleven days. On August 12, futures put the odds of a September hike near 40%. On August 19 the Fed's July meeting minutes showed that support for a rate hike extended beyond the three dissenting votes — the trigger for the test we had published two days earlier — and the market barely moved: hike odds sat near 35% as late as August 20, with Goldman Sachs' chief economist on record calling even that level of pricing “too hawkish.” Our published wire said the complacency ran the other way.⁶

Then the repricing came in two steps: Warsh's speech took the odds from roughly one-in-three to 60.4% by the following Monday, and the September 4 jobs report held them near 60% after a brief wobble. The next section explains why each mattered.¹

Chart 1: The Path of September Hike Odds

From one-in-three to roughly sixty percent in eleven days.

CME FedWatch readings as reported in dated press coverage; points are approximate and the path between them is interpolated.¹

What it means: A repricing this fast means the hike itself is now mostly reflected in market prices. The bigger risk sits with whoever is still positioned for the old story — and the next two sections cover the two forces that could still move the odds before the decision.

2. Why September rate expectations reversed

Warsh's first Jackson Hole speech as Chair, on Friday, August 28, did three things in plain language. It named his gauge: “The Fed's preferred measure of inflation, the 12-month change in the PCE price index, stands at 3.7 percent, while the six-month change is 4.1 percent.” (PCE is the inflation measure the Fed officially targets; it runs on a different basket than the better-known CPI.) It dismissed the restrictiveness defense: “I would be hard pressed to describe broad financial conditions as restrictive.” And it set the bar for holding still: “We must be confident that underlying inflation is moving to our objective, clearly and at sufficient speed. Otherwise, we have work to do.”²

Then the labor market took away the doves' best argument. The September 4 report printed 162,000 jobs against roughly 53,000 expected — the strongest month since spring — with unemployment steady at 4.1% and wages up 3.1% over the year. The sharper fact sits in the revision: July, first reported as the expansion's first negative month and the anchor of August's economy-is-rolling-over argument, was revised to plus 23,000. The two-year Treasury yield touched its highest level since January 2025, and the S&P 500 closed Friday at 7,718.60, down 0.4%.³

One asymmetry matters for this Friday's CPI report. Consensus expects core CPI — inflation excluding food and energy — near 2.4%, tame by that index's math. But the Chair has already told the market his number is PCE at 3.7%. A cool CPI print is therefore a weaker case for holding rates than it would normally be: the two indexes are telling different stories, and Warsh has pre-committed to the hotter one.¹³

Chart 2: The Payrolls Reversal

July's weak payroll print was revised positive — and August tripled expectations.

Per the BLS Employment Situation report for August 2026 (Sep 4) and dated press coverage of the consensus and revisions.³

What it means: The market now sees a September rate hike as more likely than not. That shift reflects two things: Chair Warsh has emphasized PCE inflation, which remains well above the Fed's 2% target, and August hiring was much stronger than expected. A softer CPI report on Friday could still change the odds, but it may matter less than usual because the Chair has already emphasized a hotter inflation measure.

3. Oil moves closer to the inflation debate

Over the September 5–7 holiday weekend, the tanker war stopped being a metaphor. The United States said it struck three Iranian oil tankers, destroying one, in retaliation for ballistic-missile attacks on Navy warships; Iran threatened a restricted zone extending beyond the Strait of Hormuz, through which roughly a fifth of the world's oil moves in peacetime. Brent rose to around $97 in Monday's holiday session — U.S. stock markets were closed for Labor Day, but oil traded — up roughly 9% in five days and about 19% in a month, near its highest level since late July.⁴

Our standing wire, set August 17, asks one specific question: does Brent settle a Friday above $95? Three Fridays have now said no — including September 4 at $94.71, a 29-cent miss — and twice this summer the premium collapsed within days when talks resumed, which is exactly why the wire demands a weekly close rather than an intraday spike. This Friday, September 11, the test arrives with Brent prices already above the line. A settle above $95 puts the war-oil echo into the fall inflation math our wire described. A settle below $95, after a weekend like that one, would tell us the escalation premium still cannot hold a week — and that is worth knowing too.⁵

Chart 3: Four Fridays vs. the $95 Wire

Three settles under $95 — then a weekend of tanker strikes pushed Brent through the line.

Settles per dated press coverage; the Aug 21 settle is derived from the following week's reported decline and marked as such. Monday's print is not a Friday settle — the dashed segment is the open question.⁵

What it means: The U.S. and Iran are now attacking tankers in the world's most important oil chokepoint, and Brent is near $97. Our tripwire only counts if oil finishes a week above $95, because spikes like this have faded within days twice this summer. If Friday's close clears the line, energy pressure re-enters the inflation picture right as the Fed decides on rates; if it doesn't, the market is saying the war premium still isn't durable.

4. The scorecard

Five tests were on the record coming into this note — four from the August 17 outlook, one from the August 26 Nvidia piece. Here is each one, graded against its wording as written.

Chart 4: The Grade Sheet

Five published tests, graded against their wording. One is still live — and resolves this Friday.

Wording per the Aug 17 outlook and the Aug 26 note; resolutions per fn6–fn8.

Nvidia cleared the threshold we published before earnings. On Tuesday, August 26 — hours before the release — we published the test: a fiscal-Q3 revenue guide (guidance is management's own forecast for the coming quarter) of roughly $100 billion or more would support the acceleration case; below roughly $95 billion would trigger a formal review of the position. The guide came in at $108.0 billion, plus or minus 2% — above the acceleration line by $8 billion, and above the roughly $104 billion consensus. Revenue for the reported quarter was $96.2 billion, up 106% from a year earlier. The formal review threshold was not triggered; the result supports the acceleration case, though it does not settle the full thesis.⁸

Two details from the release matter beyond the headline. The guide assumes no data-center compute revenue from China — “We are not assuming any Data Center compute revenue from China in our outlook” — so the number clears our bar without the market everyone argues about. And supply commitments jumped from $119 billion to $279 billion, “primarily related to the procurement of memory”: Nvidia effectively underwrote the memory complex for years ahead, a direct read-through to the memory positions this site has carried since early August. The $105 billion lease-guarantee question we flagged on August 26 stayed quiet; per management's commentary there is no near-term recognition, and the accounting test arrives with the October 10-Q, where we said that chart would be graded.⁸

Chart 5: Nvidia's Guide vs. Published Thresholds

We drew the lines before the print. The actual landed $8 billion above the upper one.

Thresholds as published in the earnings-eve note; the actual guide and China language per NVIDIA's release.⁸

Walmart graded in the middle, and we grade it that way rather than force a verdict. The company raised all three full-year targets on August 20, which is not what a demand crack looks like. But U.S. comparable sales slowed to +2.6%, the weakest in more than six years, and the stock fell almost 9%. Demand is slowing, but the evidence does not yet point to a broad consumer downturn.⁷

Dell, graded against our own number. Our initiation rated Dell a Hold, with a probability-weighted reference value near $340. Its central bear case was that AI-server mix would dilute operating margins (operating profit as a share of revenue) in the Infrastructure Solutions Group.

Dell's September 1 results invalidated that concern. Revenue reached $47.0 billion, up 58% year over year, while EPS of $7.04 came in roughly 43% above consensus. Infrastructure Solutions Group operating margin expanded to 15.0%, up 620 basis points from a year earlier. Full-year guidance moved to $192 billion in revenue and $25.50 in EPS — the fifth straight raise.

The stock, around $394 at initiation, now trades near $528 — above the roughly $475 bull-case scenario we published. The conclusion is straightforward: the central risk in the original framework did not occur, the Hold rating limited participation, and the $340 reference value is retired here, in public. We will rebuild the thesis after this week's 10-Q filing, including a closer look at financing extended through Dell's credit arm — the one number in the print we have not yet seen audited.⁹

Chart 6: The Dell Margin Call We Got Wrong

We said AI mix would dilute this margin. It did the opposite.

Margins per Dell's FQ2’27 release; the bear-case wording per our initiation.⁹

Synopsys answered its question — for a quarter. The August 17 note asked whether this summer's open-source design tools had dented the franchise. The August 26 report said not yet: EPS of $3.91 against roughly $3.47 expected, the full-year outlook raised to about $9.7 billion in revenue and $15.07 in EPS, and the design-IP segment back to growth. One quarter does not resolve the longer-term competitive-risk question; we re-ask it at the December report.¹⁰

Bloom Energy joins the index. After Friday's close on September 4, S&P Dow Jones announced Bloom Energy will join the S&P 500 on September 21; the stock, already up over 200% this year, opened Tuesday sharply higher. The desk has been building a full Bloom framework since late August — the initiation publishes after the company's October 29 report, when the numbers can be tested rather than admired. A position disclosure related to this name appears in the housekeeping note below.¹¹

Vistra: the order was never placed. The stock traded back through the published buy zone in early September, but no resting order was active. That is an execution failure, not a market outcome.

We are retiring the standing wire rather than repeating an authorization that was never exercised. The Vistra framework will return only with a live order or a clearly stated decision not to place one. Constellation deserves the same treatment: it has closed above $300, its August ceiling question remains open, and we will address it with updated numbers in an upcoming note.¹²

What it means: Nvidia passed the test we set before its earnings, so we keep holding. Dell disproved the exact fear our framework was built around, so we are retiring our stale price reference and rebuilding it — the value of the process is identifying where the framework was wrong and explaining how it will be rebuilt. Synopsys passed for now. Bloom Energy made the S&P 500, and our full report on it comes after its late-October earnings. And we finally answer the question we kept deferring: the Vistra order was never placed. Every one of these judgments was written down in advance, so you can check our work rather than take our word.

5. This week's tests

The calendar is compressed, and it all points at Wednesday the 16th. The Treasury sells 3-year notes Tuesday, 10-year notes Wednesday, and 30-year bonds Thursday, with the long bond near 5.25% — close to August's 5.216% auction stop, the highest 30-year auction yield since 2001 — while running up to $14.5 billion of bond buybacks over the same two days. Producer prices print Thursday, and August CPI lands Friday at 8:30 a.m. ET, with consensus near 3.4% headline and core near 2.4%. The Fed is in its pre-meeting blackout — no speakers — until the decision at 2:00 p.m. ET on Wednesday, September 16, with a quarter-point hike priced near 60%.¹³

This week's wires — set before the events

CPI (Friday, September 11): a headline print of 3.5% or higher and we expect the hike to be effectively fully priced — 75% or better — by Friday's close. At 3.2% or lower, the hold case revives and the decision would become genuinely uncertain again.
Brent (standing, final form): the wire resolves as written — a Friday settle above $95 puts war oil into the fall inflation math. A settle below $95 after this weekend means the escalation premium still cannot hold a week, and we will say so.
The decision (Wednesday, September 16): on the record — we expect a 25-basis-point hike (a basis point is one-hundredth of a percentage point). The Chair named his gauge at 3.7% and left himself “work to do.” If the committee holds, this wire is wrong and gets graded like the others.
The long bond (Thursday, September 10): a 30-year auction stop at or above 5.30%, or a meaningful reported tail (a tail means the auction cleared at a higher yield than pre-auction trading implied — a sign of weak demand), says strain at the long end is back. A clean sale near current levels retires that alarm for the month.

What makes this note wrong

The hike call is a contrarian call with meaningful risk: at least one governor has signaled a preference for holding if disinflation resumes, 40% is not nothing, and committees walk back from doorsteps. A cool CPI on Friday would test our claim that the Chair's PCE framing has demoted it. Oil has faded from spikes twice this summer; a single diplomatic headline could take Brent from $97 to $89 in days and make this note's escalation section look overstated. The payrolls revision cuts both ways — a series that can revise a month by 46,000 jobs can revise it back. And the most immediate risk is operational: we missed two scheduled outlooks, and rhythm is the fix, not the promise of rhythm.

What it means: Everything this week funnels into next Wednesday's Fed decision. Watch four things: Friday's inflation report, whether oil can finish the week above $95, Thursday's 30-year bond auction, and the decision itself — where we are on record expecting a hike. Each has a pass/fail line written down above, so you can grade us next week. If we are right, expect continued pressure on bond yields; if we are wrong, the relief rally likely starts in the most rate-sensitive corners of the market.

6. Something to think about

The last time the U.S. Navy fought a tanker war in the Strait of Hormuz, the oil price went down. From July 1987 to September 1988, Operation Earnest Will — the largest naval convoy operation since World War II — escorted reflagged Kuwaiti tankers through the strait while Iran and Iraq attacked shipping.

The escorts were not quiet work. The very first convoy's lead tanker, the Bridgeton, struck an Iranian mine and kept sailing. The frigate Samuel B. Roberts was nearly sunk by another mine in April 1988, and the U.S. answer, Operation Praying Mantis, destroyed a third of Iran's major surface fleet in a day.

And across those fourteen months, crude fell — from roughly $19 a barrel on average in 1987 to roughly $15 in 1988 — because Saudi Arabia and its neighbors were pumping into the teeth of the war, and supply overwhelmed the fear.¹⁴

That is the question hiding inside every Hormuz headline this week: not whether the strait is dangerous — it is — but whether 2026 has a 1987-style cushion behind it. Then, the cushion was measured in millions of idle barrels a day; today's market is arguing about whether it exists at all. The Strait of Hormuz can create the shock, but available spare production capacity will determine whether oil stays elevated.

If there's something you want this desk to dig into, reply to the email — everything gets read.

Housekeeping — executions, disclosures, and a correction

Per the process we published on August 17: when an order enters or fills after a note publishes, the next note reports it. One execution since the August 26 note. On Tuesday, September 2, we bought 25 shares of Bloom Energy at $214.87 — disclosed here ahead of the coverage the desk is preparing; the initiation, including how that entry compares with the desk's own accumulation framework, runs after the October 29 report and will be graded against our own entry.

One more execution report, and it is not flattering. The desk's framework carried a partial-trim order for Nvidia at $233–$237. That order was never placed. The stock traded through the zone to a high of $236.54 last week, and no shares were sold; the position is unchanged from the August 26 disclosure. That is the second execution failure on this page, and it is logged the same way as the first.

And a correction, owed plainly: the disclosure lines in prior notes did not match the actual account. A long Micron position and a long Alphabet position — 7 shares at a $343.44 average cost — were both held while recent notes stated we had no position in either name, and a quantum-computing ETF listed in earlier disclosures is no longer held. The Alphabet position also sits above the $318 alert level we published on August 17; the full write-up reconciling that position with the published framework runs with our coverage of Alphabet's October 27 report, and until then the alert framework should be treated as superseded. These errors were ours. The disclosure line below now reflects the account as it actually stands, and the miss is logged on the same scorecard as everything else.

Get the next note by email → All writing

Sources & footnotes

Hike-odds path per CME FedWatch readings in dated press coverage: ~40% implied on Aug 12 by the ~60% hold pricing cited in the Aug 17 outlook (fn5 there); ~35% as of Aug 20 (centralbank.watch); roughly one-in-three before the Aug 28 speech (PBS News); 60.4% by Mon Aug 31 (CNBC analyst roundup); ~58% after the Sep 4 jobs report (CNBC); near 60% in week-ahead coverage Sep 8 (fn13). ↩
Keynote remarks by Chair Warsh, Jackson Hole Economic Policy Symposium, Aug 28, 2026 — all quotes verbatim from the published text (federalreserve.gov); coverage per CNBC and NPR. ↩
BLS Employment Situation, August 2026, released Sep 4 (bls.gov): payrolls +162,000 vs. ~53,000 consensus; July revised to +23,000; unemployment 4.1%; average hourly earnings +0.3% m/m, +3.1% y/y. Coverage: CNBC; the 2-year at its highest since January 2025 (CNBC bonds). S&P 500 close Fri Sep 4: 7,718.60, −0.38% (Yahoo Finance); 10-year ~4.78%, 30-year ~5.25% (CNBC bonds, above). ↩
Holiday-weekend escalation per Al Jazeera, Sep 7: U.S. strikes on three Iranian tankers (one destroyed) in retaliation for ballistic-missile attacks on Navy warships; Iran's threatened restricted zone; Energy Secretary Chris Wright on maintaining the blockade; roughly one-fifth of world oil transiting Hormuz in peacetime. ↩
Brent Friday settles: $89.31 Aug 28, with the week down more than 5% (CNBC, Aug 28) — the Aug 21 settle of roughly $94 is derived from that reported weekly decline and marked as derived; Brent traded $95.29 the morning of Aug 21 (Fortune, Aug 21) without holding it into the settle. $94.71 Sep 4 (Forbes Advisor, Sep 4). ~$97.29 Mon Sep 7, +9% over five days, +19% over a month (Trading Economics). ↩
Minutes of the FOMC, July 28–29, 2026, released Aug 19 (federalreserve.gov) — “Several participants favored an increase of 25 basis points in the target range at this meeting,” verbatim; the breadth reading is an interpretation of the Fed's counting language, not a headcount, as we noted when we set the wire. Vote 9–3 per the July 29 statement. Post-minutes pricing near one-in-three per CME FedWatch trackers (Investing.com, Aug 19); Goldman's Jan Hatzius, “too hawkish,” Aug 17 (Yahoo Finance). Wire wording per the Aug 17 outlook. ↩
Walmart Q2 FY27 (Aug 20): adj EPS $0.81, revenue $187.9B; U.S. comps +2.6%, slowest in 6+ years; FY constant-currency sales growth raised to 4–5%, operating income to 7–8.5%, EPS to $2.80–$2.87; stock −8.7% (company release; Globe & Mail). Test wording per the Aug 17 outlook. ↩
Thresholds and wording per our Aug 26 note, published before the release. NVIDIA FQ2’27 results, Aug 26 after the close: revenue $96.2B (+106% y/y), FQ3 guide $108.0B ±2%, the China-outlook sentence verbatim, and supply commitments of $279B (from $119B) “primarily related to the procurement of memory,” per the company's release and CFO commentary (NVIDIA newsroom). Guide consensus ~$104.2B per CNBC pre-print coverage. Lease-guarantee treatment per management commentary on the call; accounting recognition, if any, appears first in the October 10-Q. ↩
Dell FQ2’27, reported Tue Sep 1 after the close: revenue $47.0B +58%, non-GAAP EPS $7.04 (~43% above consensus), ISG revenue $31.8B +89% with 15.0% operating margin (+620bps y/y), AI-server revenue $16.4B, backlog $95.0B (~2× the prior quarter's $51.3B), FY27 guide $192B ±$2B revenue / $25.50 EPS — the fifth straight raise — per the company's release (via Business Wire mirror) and press coverage; the stock rose 13.8% the following day to ~$484 [per press accounts] and trades near $528 at this writing (quote feed, levels only). Initiation terms — HOLD, PW ~$340, bull ~$475 (30%), first-listed bear = ISG margin dilution — per the initiation. The financing-arm receivables figure in the press release is single-sourced until the 10-Q posts; we publish it, if it holds, in the re-underwrite. ↩
Synopsys FQ3’26, Aug 26: EPS $3.91 vs. ~$3.47 expected; revenue $2.477B; FY26 guidance raised to ~$9.715B revenue / ~$15.07 EPS at the midpoints; Design IP $474M, ~+11% y/y (company release). The competitive-risk question per the Aug 17 outlook, §5. ↩
S&P Dow Jones Indices announcement after Friday's close (Sep 4): Bloom Energy joins the S&P 500 effective Sep 21 (Motley Fool, Sep 8; 24/7 Wall St., Sep 8). YTD figure per the same coverage. ↩
Vistra buy-zone and authorization history per the Aug 14 note and the Aug 17 outlook (§3 and its risk card); the early-September retest and Constellation above $300 per market data at this writing (quote feed, levels only); Constellation ceiling question per the Aug 17 outlook, §5. ↩
Week-ahead calendar: 3-year auction Tue Sep 8, 10-year Wed Sep 9, 30-year Thu Sep 10, with Treasury buybacks up to $14.5B Wed–Thu; PPI Thu Sep 10; August CPI Fri Sep 11, 8:30 a.m. ET, consensus ~3.4% headline / ~2.4% core; FOMC blackout from Sat Sep 5; decision Wed Sep 16, 2:00 p.m. ET, hike priced near 60% (Charles Schwab, Sep 8; Kiplinger calendar; TheStreet week-ahead). August's 5.216% 30-year stop, highest since 2001, per the Aug 17 outlook (fn1 there). ↩
Operation Earnest Will (Jul 1987–Sep 1988), the reflagging of Kuwaiti tankers, the Bridgeton mine strike on the first convoy (Jul 24, 1987), the Samuel B. Roberts mining (Apr 14, 1988), and Operation Praying Mantis (Apr 18, 1988) per the Naval History and Heritage Command. Crude prices: annual averages roughly $19/bbl (1987) and $15–16/bbl (1988), per EIA annual averages. ↩

Methodology: tests are graded against their written wording. Quotes are verbatim where primary documents or official transcripts are available; other factual claims are sourced in the notes. Fills and positions are disclosed only when real. Intraday levels marked “quote feed, levels only” are indicative, not official closes. Nothing here is a price target.

Nothing on this page is investment advice. We work in scenarios and trip-wires, not predictions. See disclaimer.

Disclosure: I/we have beneficial long positions in NVIDIA, Bloom Energy, Alphabet, Synopsys, Micron, ASML, and a memory-sector ETF (ticker DRAM) through stock ownership. I/we have no position in Dell, Vistra, Constellation Energy, or Walmart. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. This commentary is for informational and educational purposes only and does not constitute investment, tax, or legal advice. Past performance is not indicative of future results. See disclaimer.