Lily Caruso on the tanker war that escalated while the tape was closed, a Brent number this desk printed backwards, the firmest look at September we can build out of history and the calendar — with a probability attached, graded on the 30th — the price of memory arriving at Apple on Wednesday, and the week ahead, redrawn by the hour.* **By Lily Caruso · Sunday, September 6, 2026 — The Morning Brief, weekend edition, filed 11:30 AM ET, fact-checked and revised at 11:50** --- Good morning. It's Sunday. The stock market is closed, the bond market is closed, and the only price that moved overnight was bitcoin's — $79,832 at 1:14 this morning, up a quarter of a percent, per CoinDesk.
That's a stamp, not a settlement, and it's the last one you'll get from me today. Everything else waits for 6 PM Eastern, when futures reopen and the oil market puts the first number on a weekend in which the United States struck three Iranian tankers, destroying one and disabling two, after Iran fired ballistic missiles at an aircraft carrier. That's the first room.
But the house rule is corrections first, and I owe one at full size. **The correction.** Yesterday's episode said Brent settled Friday at $92.68. It didn't. Reuters' settlement sentence reads: *"Brent crude futures settled at $96.28 a barrel, up 76 cents, or 0.8%."* I transposed the digits, 92.68 for 96.28, and the mistake survived the fact-check because the percentage next to it was right, the weekly gain next to it was right, and nobody on this desk ran the one check that would have caught it, which is the spread.
At $92.68, Brent would have been $1.20 over WTI's $91.48. That is not a spread you see in a tanker war. At $96.28 it's $4.80, which is.
The spread *is* the war premium on seaborne crude, and we printed a number that erased it and didn't notice. So, for the record: Brent **$96.28**, up 0.8% Friday and 7.6% on the week per Reuters, which Trading Economics rounds to 8% for the week and +21% for the month, and which Al Jazeera puts at "roughly $70 before the war." WTI **$91.48**, up 0.20% Friday and 10% on the week by Reuters' rounding, a little more than the "roughly 9.7%" we carried. And a new check for the tracker, in the spirit of Friday's H.15 rule: any oil level gets printed beside its companion benchmark, and somebody looks at the gap between them.
A spread that's wrong is a level that's wrong. Now the weekend. **The first room: three tankers, two ships, one carrier.** The sequence, with the timestamps we can stand behind. On Friday, according to Euronews, Iran fired anti-ship ballistic missiles from Isfahan and Kerman at a U.S. aircraft carrier and a destroyer in the Strait of Hormuz — reportedly the first time it has used that class of weapon in this war.
Both ships evaded. No Americans were hurt. By mid-morning Saturday, Eastern time, U.S.
Central Command had confirmed the answer: three Iranian crude carriers struck. The *Downy*, hit by four missiles about six miles off Kharg Island, "permanently disabled." The *Stark 1*, near Jask, east of the Strait, also "permanently disabled." And the *Kylo*, an unladen tanker in the Gulf of Oman, "completely destroyed." Admiral Brad Cooper's statement is the whole policy in two sentences: *"If you shoot at two of our ships, we will impose an even higher economic cost — taking out three of yours. We will not hesitate to defend American forces, and if necessary, destroy Iran's limited and exposed oil fleet."* Saturday evening the Revolutionary Guard said it had targeted *"three oil tankers on the unauthorised route of the Strait of Hormuz"* and three vessels it described as affiliated with America, in language I'm not going to reprint in full.
Central Command has confirmed none of those hits. At 2:53 this morning the Guard added a claim on an American unmanned surface vessel entering the Strait. And on Sunday, on Telegram, Iran's parliament speaker, Mohammad Bagher Ghalibaf, wrote, per Reuters: *"From now on, any attack against Iran's interests and security will receive a faster, heavier and more painful response."* The rules of the game, he said, have changed.
Three facts to set beside that sequence. The month of relative calm ended Tuesday, not Saturday — NPR dates the resumption to "nearly a week" before the tanker strikes, and Reuters calls it the most serious escalation in weeks and the first significant fighting since late July. Saturday was day 190 of a war that began with U.S. and Israeli attacks on February 28, whose negotiations collapsed after a memorandum of understanding in mid-June.
Fortune counts eighteen U.S. service members dead in it. And the two men at the top of the government spent the two days before the missiles flew minimizing it. Thursday, at a White House briefing, the Vice President: *"I wouldn't call it a war.
Right now, there is no active shooting."* Friday, in the Oval Office, the President: *"It's small potatoes,"* and *"It's not a big thing,"* per Al Jazeera — CBS has the fuller sentence as *"I call it a military conflict because it's small potatoes for us."* By the weekend the shooting was active. Defense analyst Wolfgang Pusztai, to Al Jazeera, on ballistic missiles aimed at a carrier: *"This is a very, very significant development, and it doesn't come as a surprise that the Americans retaliated."* I'll put those assessments side by side and let you choose. What the Strait looked like while this happened.
Lloyd's List Intelligence, via CBS: 102 transits last week against 126 the week before, in a waterway that saw 130 or more *a day* before the war. Kpler, via Al Jazeera: five vessels Monday, eleven Tuesday, six Wednesday, a ten-day average of thirteen. Reuters had four commodity vessels on Thursday.
The "unauthorised route" in the Guard's statement reads, to this desk, as the lane the U.S. Navy has been guiding ships along — NPR notes the guidance and notes that traffic stayed low anyway — and Tehran's position, as written, is that the lane it doesn't control is the lane it will shoot at. That's the new fact for Sunday night. **What's priced: nothing yet.** Oil hasn't traded since Friday's settlement.
Brent's $96.28 was already its highest close since July 24, per Al Jazeera. The price that moves first at 6 PM is Brent. The price that moves second is the 10-year, and you don't have to take my word for that, because the Treasury Secretary said it Friday on Steve Bannon's program: *"If you go and look, the interest rates — this is the highest correlation they've ever had to the oil price."* Scott Bessent said it while predicting oil would eventually fall to *"$50, $40 crude maybe"* once the conflict ends, "just because there's so much coming online." Both halves can be true.
The half that matters tonight is the first one. Sunday's oil print is Monday's bond print, and Monday's cash bond market is closed, so Tuesday's open will carry two nights of it. On the ground the price is already in.
AAA's national average reached $4.14 on Friday morning — the first Labor Day above four dollars, ever; the old record was $3.82, in 2012 — and diesel set a record at $5.85, past the $5.81 of June 2022. Brown University's tracker, cited by WFSB, puts the added fuel cost since the war began at more than $741 per household. My Saturday row, *65% that WTI settles Friday, September 11, above $90*, is likelier this morning than it was yesterday.
It stays exactly where it was written. We don't re-price rows on weekends. We grade them on Fridays. **The second room: September, firmly.** You asked for a firm look.
Here it is in four parts: where the index actually stands, from the official closes on Yahoo Finance's S&P 500 history page; what history says; what's different this time; and a number. *Where we are.* The S&P 500 closed Friday at **7,718.60**. That is 12.75% above the December 31 close of 6,845.50. It's 1.03%, about 80 points, below the record close of 7,798.99 set on August 13.
It's 21.7% above the year's low close, 6,343.72 on March 30. August finished at 7,686.14, up 2.62% for the month; September is up 0.42% through four sessions, and the lowest of those closes, Tuesday's 7,631.47, was only 2.15% off the record. Friday opened at 7,750.19, which was also the high of the day; the index never saw that price again and closed 32 points lower.
So this is a market that has spent three weeks somewhere between 13 and 168 points under a record, refusing to make a new one and refusing to give back more than two percent of the old one. *What history says.* Citadel Securities' market-intelligence desk ran 1928 through 2025 in its August 31 note: September averages −1.1%, closes lower in 55% of years, carries an average intra-month selloff of 4.7%, and its second half, averaging −0.91%, is "the weakest two-week period of the year." In midterm years the average is −1.5% and the average selloff 6.2%. Opening Bell Daily's series since 1975 says −0.8% and a 45% win rate. Jay Woods of Freedom Capital Markets gives the last twenty years: eleven of twenty Septembers up, a 55% winning percentage, but the ups averaged +3.0% and the downs −4.8%.
That is the shape of the thing. Modest when it's kind, violent when it isn't. Over the last fifteen years, per the Motley Fool, the average is −1.3% and 53% were down.
Over the last five, −2.7%. Now the other side, which is real. Ryan Detrick of Carson Group, whose job this past week has been to say "get ready to hear that a lot" about September, has the conditional that fits this year: since World War II, when August was positive and the year came into September up between 10% and 17.5% — which is us, exactly — September has averaged **+1.0%**, and the final four months closed higher ten times out of eleven, averaging +5.6%.
His other line is the one I'd underline: *"the four best September returns ever all took place in a midterm year, and five of the top seven."* Midterm Septembers are worse on average and better at the extremes. That's another way of saying the variance is what changes, not the mean. The Fool's midterm statistic runs the same direction: since 1950 the index has fallen an average of 18% at some point during a midterm year, usually bottoming in October, and from that bottom it has "never been lower a year later," averaging +32% over the following twelve months.
So history's message about September isn't *sell*. It's *the tail is fat, and the quarter after it has the best base rate on the calendar.* *What's different this time, in order of weight.* Rates, first, because the fear ranking says Fed over missiles and I'm not going to argue with a ranking gold has confirmed twice. The Fed's H.15 table puts the 10-year at 4.77% and the 30-year at 5.25% as of Friday; Bloomberg had the 10-year at its highest since 2023 this week.
CME FedWatch, at Friday's close, prices a hike on September 16 at **57.6%**, rates higher than today by October at 68.8%, and by December at 89.1%. The July decision was a 9–3 vote, with the three dissents *for* a hike, and the Chair's sentence from that day — *"There is no soft implicit target. Not on this committee's watch.
There's only a target and it's 2%"* — is the reason the market won't price him away. A hike on the 16th would be the Fed's first increase since 2023. Meanwhile the index trades at 21.4 times this year's estimated earnings: FactSet's bottom-up 2026 number is $361.38, and 21.4 is just Friday's close divided by it.
That's an earnings yield of 4.68%. The 10-year pays 4.77%. Read it once more.
At Friday's prices the index yields less than the bond it's supposed to beat. That isn't a sell signal. It's a statement about how much has to keep going right.
Earnings, second, because a great deal *is* going right, and this paragraph is the bull case. FactSet's John Butters wrote Friday that analysts raised the third-quarter EPS estimate by 1.2% during July and August, to $89.69 — the second straight quarter of upward revisions in the first two months, in a season when the 5-, 10-, 15- and 20-year averages are *cuts* of 1.7%, 2.1%, 2.6% and 3.1%. The calendar-year 2026 estimate rose 6.1% over the same two months.
The catch is breadth: only four of eleven sectors were revised up, and the leader was Energy at +11.8%, which is the war again in a different hat. The quarter just reported was extraordinary. The 493 companies outside the Magnificent Seven grew earnings 31.8%, per FactSet's August 28 note; the Seven grew 118.5%, or 43.2% once you strip out Alphabet's $98 billion of unrealized gains on equity securities and Amazon's $53.4 billion gain, "primarily from investments in Anthropic." So the market is expensive against earnings that are growing very fast, some of which are marks.
Both halves, again. Oil and the war, third; see the first room. The September-specific mechanism is that Brent at $96 is already inside the August CPI that prints Friday, and whatever prints at 6 tonight goes into September's.
The prediction market on Robinhood, checked this morning, has August headline CPI above 3.2% at 89 cents, above 3.3% at 63, above 3.4% at 20, above 3.5% at 7. July was 3.4% headline and 2.5% core, with gasoline up 24.6% from a year earlier and airline fares up 25.5%. The crowd's imagination is 3.3 to 3.4.
The hike case lives in the headline; Waller's hold case lives in the core. A 3.5 headline over a 0.2 core is the split decision this whole month is built around, and it arrives at 8:30 Friday. Flows, fourth, and these are the ones nobody prices until they show up.
Citadel again: the buyback blackout accelerates September 12, out of a year in which more than $1.1 trillion of authorizations were opened; equity mutual funds' September purchases, a median 1.79% of assets, are the weakest of any month since 1984; the top 100 pension plans are 112% funded, the highest since 2001, which is a polite way of saying *sellers of equities on a schedule*; and the September 18 expiration carries $6.2 trillion of notional, on pace to pass June's $7.7 trillion record. That same Friday is the S&P rebalance. At 7:15 Friday evening, S&P Dow Jones announced that Bloom Energy, Illumina and Everpure join the S&P 500 before the open on September 21, replacing Molson Coors, The Trade Desk and Builders FirstSource — all three of which drop not to the MidCap 400 but straight to the SmallCap 600.
In the S&P 100, Dell, Palo Alto, Arista and SanDisk go in; Honeywell Aerospace, Nike, Simon Property and Colgate-Palmolive come out. Nike closed Friday at $38.40. An index committee just swapped sneakers and toothpaste for servers and memory, and the trade executes into a triple-witching close.
This desk's own ten-session model, built Friday, says Dell and SanDisk are the crowded rooms to leave alone through the 21st. The index says a mandatory buyer arrives on the 18th. I'm noting the disagreement rather than settling it, because the 21st is the model's grading day and the tape will settle it for us.
Positioning, fifth, and this is the one category that argues *for* the month. The VIX closed Friday at 14.13. Citadel's readings from August 28: one-month 25-delta S&P puts the cheapest since December 2024, Russell one-month volatility at five-year lows, VVIX in its first percentile since January, and, the part that matters most, a systematic crowd that already reset in July — leveraged-ETF assets down $70 billion, or 31% from the June peak, with semiconductor leveraged funds down 50%.
AAII's survey for the week ending September 2 is 39.7% bullish, 37.6% bearish and only 22.7% neutral, against a 31% average. Everyone has an opinion and nobody's in the middle. That's polarized, not euphoric.
Cheap hedges and a de-levered systematic book don't prevent a selloff; they shorten it. Citadel's own conclusion, in its author's words: *"I view September as a tactical downside window, not the beginning of a broader bearish turn,"* because *"the upside catalysts are becoming less obvious just as the downside catalysts are becoming more numerous."* *The number.* Here's the ladder, every rung sourced above, so you can argue with a rung instead of the answer. I start at 48%.
The long history says 45 and the last twenty years say 55, and I weight the long one. Momentum and earnings, meaning Detrick's conditional and FactSet's revisions: plus seven, to 55. Rates, meaning a 58% hike into a market that yields less than the 10-year: minus eight, to 47.
Oil and the weekend, with the first price still hours away: minus five, to 42. The September flows, the blackout and the pensions and the mutual-fund calendar: minus two, to 40. Positioning already reset, hedges already cheap, retail neither all-in nor all-out: plus five, to 45. **The desk's bullish likelihood for September is 45%.** Specifically: a 45% chance the S&P 500 closes Wednesday, September 30 above August's close of 7,686.14.
A coin flip that leans down. And because the shape matters more than the sign, take Woods' twenty-year averages, +3.0% when September is up and −4.8% when it isn't, and a 45% coin has an expected value of about −1.3% for the month. The path is the other number.
At a VIX of 14, seventeen sessions of pure noise put a 3% closing drawdown at roughly 41%; add four scheduled binaries and one unscheduled war, and I'll say **55% that the index prints at least one close at or below 7,487.04**, three percent under Friday, between Tuesday and the 30th. Both rows are below. What would make me wrong on the first is simple, and I'd welcome it: a core CPI at 0.2% or less on Friday, and a hold on the 16th.
That's the outcome where the month finishes green, the row loses, and I write the loss in the same font. **The third room: Wednesday, and what memory costs now.** Apple's event is Wednesday at 10 AM Pacific. TrendForce's price estimates, reported by MacRumors on Thursday: iPhone 18 Pro at $1,249 to $1,299 for 256 gigabytes, up $150 to $200; Pro Max $1,349 to $1,399; the foldable "Ultra" at $2,099 to $2,299, and above $3,000 at the top. The reason, in the research firm's words: *"the cost of a 256GB memory chip has risen nearly 400% year-on-year,"* with the Pro models' bill of materials up roughly 38%, because *"the insatiable AI infrastructure buildout has made memory costs even more expensive."* MacRumors' own guide adds the tell: "Apple is paying more for memory and storage, and prices for Macs and iPads already went up." Read that next to this week's tape and the circle closes.
SanDisk closed Friday at $1,738.84, its fifth straight close above the line. Micron closed at $1,015.48. Zscaler told its shareholders Thursday it's paying "higher prices and tighter availability for memory, storage, and processors." Foxconn's August filing, in yesterday's episode, said its consumer-electronics clients rebuilt inventory for "new products." Now the bill arrives at the customer — a customer paying $4.14 for gasoline, whose wages are growing 3.1% against a 3.4% CPI, who fired Lululemon on Friday and who has taken Nike to $38.
The memory cycle has had one price all year, the one the hyperscalers pay. On Wednesday it gets a second one, and the second one is the one that shows up in the CPI. So I'll put a row on the announcement itself, because it's the cleanest imagination-versus-arithmetic test on the calendar: **60% that the iPhone 18 Pro's U.S. starting price, as announced Wednesday, is $1,199 or higher** — at least $100 above the 17 Pro's $1,099, the figure in Apple's own September 2025 release.
The crowd, via TrendForce, expects more than that. If Apple holds the line at $1,099, the memory cost lands on Apple's gross margin instead of the consumer's, and that's a different September for a stock that fell 2.54% on Friday before anyone had said a word. Either answer is a story.
Only one of them is inflation. **The fourth room, briefly: the supply nobody's watching.** Three Treasury auctions this week, per the department's tentative schedule, and they land on the three loudest days. A 3-year Tuesday. A 10-year reopening Wednesday, at the usual 1 PM close — the same hour Apple starts talking.
A 30-year reopening Thursday, the day of PPI, Oracle and Adobe, with the long bond already at 5.25% and a letter from Oslo still on Jens Stoltenberg's desk asking to cut Norway's Treasury weighting by more than a third. A 20-year follows on the 15th, the first day of the FOMC. If Bessent is right that rates are oil now, Thursday's 30-year auction is the first place a $100 barrel would show up in the government's own borrowing cost.
And overseas, on the days we're closed: China reports foreign reserves Monday, trade Tuesday, and CPI and PPI Wednesday, where Lundgreen's expects producer prices up 3.2% from a year earlier. A Chinese factory-gate price rising 3% is a sentence the U.S. inflation debate hasn't had to include for years. It does now. **Housekeeping, at full size.** The Brent correction is above, where it belongs.
Two decoys for the chronicle, neither of which reached the page: a Yahoo Finance "what to know this week" that turned out to date from September 2024, and Colgate's Friday close, which one vendor's page showed as a 2:58 PM snapshot and another with a previous close that matched nobody — so Colgate appears in this column as a deletion from an index and nowhere as a price. The chronicle stands at 63. One note on method: two outlets counted the S&P's monthly winning streak through August as four and as five.
By the month-end closes on Yahoo's table, July finished about ten points *lower* than June, so this column prints August's +2.62% and no streak. When sources disagree on a count, we print the closes and skip the adjective. **The fact-check notes, printed in full, because that's the deal.** The first draft of this episode, filed at 11:30, had the *Stark 1* "destroyed"; Central Command's word for it, via Euronews, is "permanently disabled," the same as the *Downy*. Only the *Kylo* was destroyed.
The draft timed Central Command's confirmation to a 4:11 PM CBS entry that I couldn't find on a second read; NPR had the strikes at 11:05 AM Eastern, so the confirmation is now "by mid-morning Saturday." The draft ran the President's remark as a composite; it's now the two fragments Al Jazeera reported and the fuller sentence CBS reported, each attributed. The Vice President's "wouldn't call it a war" is now dated to Thursday's briefing and quoted whole, including "there is no active shooting," which is the part that aged. "First time" on the ballistic missiles is now "reportedly the first time," which is Euronews' word.
Ghalibaf's warning is now Reuters' full sentence, with the platform and the day. The transit counts now carry their originators, Lloyd's List Intelligence and Kpler, not just the outlets that relayed them. The index announcement now carries its time, 7:15 PM, and the detail that the three deleted companies go straight to the SmallCap 600.
The iPhone 17 Pro's $1,099 is now sourced to Apple's own release instead of inferred from TrendForce's arithmetic. And "Freedom Capital" is Freedom Capital Markets. None of it changes a row.
All of it changes how much you should trust the rows, which is the point. **The rows, in the same font as the loss.** Three new ones from this episode. *45% the S&P 500 closes September 30 above 7,686.14* — settlement: S&P Dow Jones Indices' official close as shown on Yahoo Finance's history page; falsifier: core CPI at or under 0.2% Friday plus a hold on the 16th. *55% at least one S&P 500 close at or below 7,487.04 between September 8 and September 30, inclusive* — same source; falsifier: a soft CPI and a quiet Strait, together. *60% the iPhone 18 Pro's announced U.S. starting price is $1,199 or higher* — settlement: Apple's newsroom release Wednesday; falsifier: Apple absorbing the memory cost. Standing, unchanged: my three from Friday and Saturday — *60% the Russell trails the S&P through Friday* (IWM $296.01, SPY $770.24), *55% FedWatch above 60% Friday* (from 58%), and *65% WTI above $90 Friday* — and my colleague's *60% Micron above $1,000* and *60% Tesla below $376.37*, all graded on Friday's settlements. Episode 28's XLF-versus-XLU grades at Wednesday's close; the ten-session shortlist at the September 21 close; the longer rows on their dates.
Settlement rules, as adopted: yields on the Fed's H.15, index levels on the official close, oil on the exchange settlement as reported by Reuters — and, from today, every oil level printed beside its companion benchmark. **Next week, redrawn, by the hour.** *Tonight, 6 PM:* futures and crude reopen — the first price on three tankers. *Monday:* closed here; Taipei trades Hon Hai's record August; Beijing reports reserves. *Tuesday:* reopen with two nights of oil in it; China's trade data overnight; the 3-year at 1 PM. *Wednesday:*