Four European electorates voted in the eight days ending Sunday. By Monday morning, none of them had a government.

This is not a metaphor. In each case the count is finished or nearly finished, the winner is known, and the machinery that is supposed to convert votes into executive power has either stalled, been refused, or been rendered irrelevant by the fact that the incumbent was never at risk in the first place.

Russia. United Russia took 57.83 percent with more than 95 percent of ballots counted, the Central Election Commission said Monday — up from 49.8 percent in 2021, in the first Duma election since the invasion of Ukraine. Every genuine opponent of the president was barred from standing. The Organisation for Security and Co-operation in Europe was not invited to observe; the foreign ministry accused it of double standards in advance. A vote in which the outcome cannot change the government is not a transfer of power. It is a census of consent, and the state wrote the questionnaire.

It also could not be protected. On the closing day of polling, Russia's defence ministry said it had shot down 1,110 Ukrainian drones across Russia, Crimea and the Black Sea — the largest attack of the war. Several got through to the Moscow Refinery in the south of the capital, one of the country's largest, and set it alight. Two people were killed and twenty injured, including three children, according to Moscow region governor Andrei Vorobyov. A power station, a warehouse and a market were hit. The Kremlin promised to intensify strikes on Kyiv; overnight Russia fired 138 drones at Ukraine, half of them jet-powered, killing a mother and two small children in the capital and four more people in Sumy.

Mecklenburg-Western Pomerania. The AfD won 38.2 percent, more than doubling its 2021 result of 16.7 percent and becoming the largest faction in the Schwerin parliament. It will not govern. Every other party except the BSW has ruled out cooperation, and the BSW took 4.8 percent and did not make the chamber. The SPD, on 35.5 percent under Manuela Schwesig, will almost certainly continue in office having lost the election.

The number that will outlive all of these is 4.9. That is the Christian Democratic Union's share — down 8.4 points from 13.3 percent — and it puts Chancellor Friedrich Merz's party below the five percent threshold and out of a German state parliament for the first time in the party's history. Merz came out of the chancellery on a state election night, which chancellors do not do, and called it a disaster. He said he would remain chancellor and party leader and continue the reform course. In Berlin, voting the same day, Die Linke took 25.7 percent and the CDU 18.8 percent, pointing to a far-left-led coalition in the capital. It was the AfD's second first-place finish in a month, after 43.8 percent in Saxony-Anhalt against the CDU's 17.2.

Sweden. The centre-left bloc won 176 seats to 173 — a margin of 50,359 votes in a country of ten million. Prime Minister Ulf Kristersson resigned on 17 September, by post on X, as soon as the count was certified. Under Swedish law he and his entire cabinet remain in their jobs as a caretaker government, forbidden from calling an extraordinary election and confined to ongoing and urgent business. The Riksdag session opens on 29 September. A new prime minister can be elected no earlier than the week that follows. Sweden currently has a head of government who has been discharged from office and cannot leave it.

What connects these is not disorder. It is suspension. The vote happens on schedule; the government does not. And the interval between them has become long enough, and predictable enough, to price.

Which brings us to the fourth electorate, the one that has not voted yet. France holds a presidential election in April and May 2027. On Friday, the premium France pays over Germany to borrow for ten years hit 104 basis points, the first time above a full percentage point since the euro zone debt crisis in 2012. Nobody has cast a ballot. Investors have simply read the polls — Marine Le Pen leading, Jean-Luc Mélenchon proposing that the central bank cancel the government debt it holds — and concluded that the next French government will be no more capable of closing a deficit than this one.

European bonds rallied hard on Monday, the French ten-year down ten basis points to 4.469 percent and the German down five to 3.472. That was not a verdict on any of this. It was oil: Brent fell about two percent to just above $101 on evidence that Saudi crude is moving again, and the entire continent's borrowing costs moved with it.

That is the real result of the weekend. Four elections changed nothing that a tanker could not change more.

There is a useful test for whether an institution believes its own statements: ask whether it has to post a price.

Parliaments do not. Coalition negotiators do not. Rating committees post a price roughly twice a year. Sovereign bond desks post one continuously, in size, against their own capital, and they have spent six consecutive weeks selling.

The headline number is the ten-year yield averaged across the Group of Seven, which sits around 4.2 percent. That is the highest since 2008. It is being driven by two things at once — an oil-led inflation scare and a slower, uglier repricing of what developed-world governments are actually good for — and the second is the one that will still be here in a year.

## France is the tell

On Friday the spread between French and German ten-year debt reached 104 basis points, the first reading above a full percentage point since 2012. French ten-year borrowing costs have risen faster than any other developed economy's in this selloff. Since January the yield has gone from roughly 3.6 percent to about 4.5, a rise of around a quarter in nine months.

The cause is not mysterious and it is not really about oil. France's deficit is among the worst in the euro zone, its debt-to-GDP ratio reached 115.6 percent in 2025 against 98.2 percent in 2019, and the IMF projects 121.1 percent by 2028 even assuming continued consolidation — under which interest costs alone climb to 2.7 percent of output, from 1.3 percent in 2020. The spread has doubled since the 2024 snap election produced a parliament incapable of passing a budget.

The rating agencies are now catching up rather than leading. Scope cut France to A+ with a stable outlook, citing repeated fiscal slippage and the absence of credible measures to stabilise the debt path. Morningstar DBRS confirmed AA but moved the trend to negative, noting explicitly that the April–May 2027 presidential election is likely to prevent the government from securing support for consolidation in the 2027 budget, pushing adjustment to 2028. Fitch and S&P already have France at A+. Moody's is at Aa3 with a negative outlook and reviews on 23 October.

What makes France instructive is that the market is pricing a political event with no ballots cast. Le Pen leads the polls and proposes lowering the retirement age for some workers. Mélenchon has floated having the Banque de France cancel the government debt it holds. Neither is a forecast of policy. Both are inputs into a spread.

## Where it stops

The comparison everyone reaches for is 2011–13, when the French spread peaked just under 190 basis points and it took Mario Draghi's "whatever it takes" to break the trend. That precedent is precisely why a repeat is unlikely: the European Central Bank has a facility now and the market knows it. ING's read is a 100 to 125 basis point range in the coming months absent good news from the Middle East or from French politics.

That is the honest version. France is fully financed, its market is liquid, and nobody is describing a funding crisis. What is being repriced is something less dramatic and more durable — the cost of a government that cannot legislate, charged daily, forever.

One oddity worth sitting with. French ten-year paper now yields more than ten basis points above Italy's, a country with worse debt metrics and lower ratings, on the grounds that Italian politics is currently more predictable and its investor base more domestic. Monday's rally took the French spread back under 100 on falling crude.

It will be back. The oil trade resolves in weeks. The arithmetic does not.

On 16 September the Federal Reserve raised its benchmark rate to a 3.75 to 4.00 percent range — the first increase in three years — and signalled more to come. The decision was unanimous, which matters, because it was the first meeting chaired by Kevin Warsh and the first formal acknowledgment by the central bank that the administration has not brought inflation down.

The projections put the policy rate at 4.00 to 4.25 percent by the end of 2026 and leave it there through the end of 2027. In other words: one more hike, then a long hold. The CME's FedWatch tool has markets pricing roughly a 53 percent chance the hike lands next month. A coin flip is not a signal. It is an admission that nobody, including the people setting the rate, can see past the oil price.

The bond market got there first. The benchmark ten-year Treasury yield hit a nineteen-year high above 5 percent in the week before the meeting, and was trading around 4.958 percent when the decision landed — barely moved, because the hike was already in the price. The thirty-year dipped to 5.312 percent.

On Monday the ten-year slipped back below 5 percent as Brent fell about two percent, and equities took it as permission. Dow futures were up 311 points, S&P futures 0.63 percent and Nasdaq futures 1.09 percent, led by chipmakers: Intel up 5.4 percent premarket, Micron and AMD around 2 percent each. Energy-sensitive airlines gained about one percent.

Hold the shape of that in mind. The Federal Reserve raised rates into an oil shock and told the market to expect another increase. The market then rallied because a tanker tracking firm reported that Saudi exports had recovered. The most consequential input to American monetary policy this autumn is not being generated in Washington. It is being generated by the repair schedule on a pipeline across the Arabian Peninsula.

Oil fell about two percent on Monday, to just above $101 for Brent, and the stated reason was a recovery in Saudi exports. That recovery is real. It is also one of the least reassuring things in the energy market right now, and the reason is entirely a matter of geography.

On 10 September, Houthi forces hit pumping stations on the East-West pipeline across two Saudi regions. This was materially worse than an April strike that disabled a single station and was repaired in three days. Multiple stations went down. The pipeline itself is largely intact, but no crude has loaded at Yanbu on the Red Sea since 11 September, and terminal inventories there cover only three to five days of loadings at normal rates.

Aramco is building a bypass around the damaged sections. Full repairs are estimated at four to six weeks, with an internal aim of restoring roughly half of Yanbu's throughput — around two to two and a half million barrels a day — within about a month. Kpler assigns roughly even odds to that partial recovery, with the remainder split between a faster fix and a scenario in which renewed attacks on the bypass stretch the outage from weeks into months.

So where did the barrels go?

Into the Strait of Hormuz. Satellite data cited by JPMorgan shows Saudi crude moving through Hormuz averaging 2.9 million barrels a day over six days, against roughly 700,000 in August. That single pivot is most of the story. Kpler's provisional figures put total Saudi exports at just over four million barrels a day so far in September, after August's 2.4 million — the lowest reading since at least 2013. The rest has come from ship-to-ship transfers off Oman.

Which is to say: the world's largest crude exporter responded to attacks on its western coast by concentrating its export book in the waterway that has been under a US-Iran shooting war for seven months and remains subject to a naval blockade. The East-West pipeline exists specifically so that Saudi oil can bypass Hormuz. Its whole purpose is redundancy. That redundancy is currently running in reverse.

Then came the weekend. On Saturday the Houthis fired ballistic and cruise missiles and drones at what their spokesman Yahya Saree called sensitive sites in Riyadh, and at Aramco facilities in Yanbu. The Saudi-led coalition said the missile aimed at the capital was intercepted and destroyed. An AFP journalist watched firefighters put out a blaze on an Aramco-branded fuel tank at a depot near Riyadh's international airport. It was the first time the capital had been targeted since the fighting resumed. The United States warned that hostilities could escalate rapidly. The Security Council had condemned the attacks on Friday.

The market read all of this and sold crude, because the export number went up.

It is a defensible trade over a week. Over a quarter it is a bet on two things at once: that the bypass holds, and that a strait already closed to Iranian ports stays open to Saudi ones. Both bets have the same counterparty, and it is not the market.

Iran's Khatam al-Anbiya Central Headquarters, which oversees the country's armed forces, said on Sunday that it had received information the United States is preparing to resume strikes. It warned of retaliation against American bases and interests across the Middle East, and told regional governments that assisting Washington would make them parties to the conflict.

Speaking to Fox News the same day, President Trump said he was in "deciding mode" on Iran, listing further military action, economic pressure and a deal as his options, and threatening to destroy the Iranian economy or remove its leadership absent an agreement. Tehran promised a "painful" response.

The war is seven months old. It began in late February with joint US-Israeli strikes. A sixty-day memorandum agreed in June reduced full-scale fighting and expired last month; attacks around the Strait of Hormuz have increased since.

Underneath the threats, an actual negotiation is being assembled by intermediaries. Qatar and Pakistan are carrying proposals between capitals. Iran's stated conditions for restarting talks are an end to the war and to US attacks, the release of frozen Iranian funds, and the lifting of the naval blockade on Iranian ports. Qatari foreign ministry spokesman Majed al-Ansari says Doha has received assurances from US officials that Washington wants an agreement.

The timing is not accidental. The General Assembly's general debate opens Tuesday and runs through 26 September and again on the 28th, with roughly 130 heads of state and government expected. Secretary-General António Guterres delivers his state of the world address first — his last as chief. Trump and Emmanuel Macron speak Tuesday. Volodymyr Zelenskyy and Iranian President Masoud Pezeshkian both speak Wednesday. Benjamin Netanyahu speaks Thursday. Mahmoud Abbas, denied a US visa for the second consecutive year, will address the chamber by video the same day.

The Security Council has added a high-level meeting on artificial intelligence to Wednesday afternoon, after a high-level session on Ukraine that morning. That is a striking pair of agenda items for a body designed around territory and armies, and it tells you what the building thinks is coming.

The session's theme, chosen months ago by General Assembly president Khalilur Rahman of Bangladesh, is "Restoring trust, managing transformation: A United Nations that delivers for all." It will be read out to a hall containing the presidents of two countries at war with each other, one of whom will be in New York while his military command says it expects to be bombed.

The useful thing about a speaking schedule is that it is a schedule. On Wednesday, Pezeshkian has to say something specific about those four conditions, in public, on the record, with the American delegation in the room. It is the first moment in months where the Iranian position will be stated by the head of state rather than relayed through Doha. Watch what he drops.

The choreography was supposed to run one way. Russia scheduled its first parliamentary election since the 2022 invasion as a demonstration of support for the war, barred every genuine opponent from the ballot, declined to invite OSCE observers and pre-emptively accused the organisation of double standards. Three days of voting, a controlled result, a legitimising headline.

Kyiv scheduled something over the top of it.

Russia's defence ministry said it shot down 1,110 Ukrainian drones across Russia, annexed Crimea and the Black Sea — by its own account the largest single attack of the war. Several reached the Moscow Refinery in the south of the capital, one of the country's largest. Video verified by the Financial Times showed a drone exploding into the plant, black smoke across the sunrise, and multiple sections alight. A nearby power station, a warehouse and a market were struck. Drones also hit residential buildings near the refinery and at least eleven towns in the surrounding region. Two people died and twenty were injured, three of them children, according to regional governor Andrei Vorobyov.

Turnout stood at 48 percent at noon local time on the final day. United Russia finished with 57.83 percent against 49.8 in 2021.

The Kremlin accused Ukraine of trying to disrupt the vote and vowed to intensify strikes on Kyiv. Overnight, Ukraine's air force recorded 138 incoming Russian drones, half of them jet-powered models that are considerably harder for air defences to engage. A mother and two small children were killed in Kyiv; another mother and child were injured in a separate strike. In the north-eastern Sumy region, a drone destroyed a vehicle and killed four people.

On Saturday, President Zelenskyy said the SBU had brought down jet-powered drones using a new interceptor drone of its own — the first claimed operational success against that class of weapon. If it holds, it is more consequential than anything on the ballot.

Kyiv has urged other countries not to recognise the vote. That request will be tested this week in New York, where the Security Council holds a high-level session on Ukraine on Wednesday morning with Zelenskyy present.

Russia will report a constitutional majority and a mandate for the war. It will do so from a capital where the largest refinery took direct hits during polling, where eleven satellite towns were struck, and where the state's central claim — that it controls the space in which its citizens vote — was falsified on camera at dawn on the last day of voting.

Chip stocks led a global rally on Monday, with Intel up 5.4 percent premarket and Nasdaq futures over one percent, on what was described as evidence of booming AI demand. The evidence was a supply failure.

Speaking at Splunk's conference in Denver, Intel chief executive Lip-Bu Tan said the company can currently supply about 50 percent of what its customers are asking for. "CPU demand is so high that we can only supply 50% of customers," he said, adding that chief executives have called him directly and that he has had to apologise. Intel's second-quarter revenue rose 25 percent year on year to $16.1 billion, its strongest quarterly growth in fifteen years, with the foundry business at $5.77 billion.

That is a story about general-purpose processors, not accelerators, and that is the point. The first phase of the AI boom was a GPU story: enormous compute concentrated in training runs. The second phase is inference and agents, which run continuously, every time a user touches a service, and which need orchestration and scheduling — work that lands on CPUs. Tan expects agents to number in the trillions within a decade. Intel has raised client CPU prices three times since late 2025 and is reported to be raising PC processor prices around 10 percent from 5 October.

But the binding constraint Tan keeps returning to is not logic. It is memory. Prices have climbed five to sevenfold, he said, and limited production is forcing customers to delay projects outright. He has been repeating a line from the memory makers all year: "Lip-Bu, there's no relief until 2028."

The mechanism is specific. High-bandwidth memory, the stacked DRAM that sits beside Nvidia and AMD accelerators, is made almost entirely by Samsung, SK hynix and Micron. Demand has outrun all three. To keep up, they are converting existing lines to HBM, which removes capacity from conventional DRAM for everyone else. So a data centre can secure its GPUs and still stall, because the memory to pair with them will not arrive in time.

The downstream evidence showed up in Seoul on Monday. SemiAnalysis data puts the average global hourly rental for an H100 at $2.80 at the end of August, up 43.6 percent from $1.95 — and that is the previous generation, which should be getting cheaper as newer parts ship. Korean AI startups are abandoning projects outright because they cannot secure capacity, and cloud providers now prefer large long-term contracts, which prices smaller customers out of even bidding. One Korean cloud executive noted the second bottleneck behind the first: there are not enough data centres with the power and cooling to run the chips, so buyers in a hurry pay premium rates for the few that exist.

Hold the chain together. Memory makers cannot expand fast enough, so DRAM triples in price, so Intel cannot ship CPUs, so Intel raises prices on consumer processors on 5 October, so PC makers pass it on. The most-owned trade in global equities and the cost of a mid-range laptop this Christmas are the same supply curve, viewed from opposite ends.

There is a version of this in which shortage is bullish: constrained supply, pricing power, fat margins. Monday traded that version. The other version is that a boom whose limiting input has a two-year lead time and three suppliers is not a boom in computing. It is a boom in memory rents, and the people collecting them are not the companies whose names are on the data centres.

Here is the weekend in two numbers. Resident Evil, Zach Cregger's reboot of a 1990s zombie video game, opened to $60 million in 3,684 theatres. Practical Magic 2, a legacy sequel to a beloved 1998 romance, took $12.2 million in its second weekend from 4,163 theatres — more rooms, fifth the money, down 59 percent.

Resident Evil was tracking at $40 to $50 million. It is the largest opening in the franchise's history by a wide margin; the previous record was Afterlife in 2010 at $26 million, unadjusted. It crossed $100 million worldwide in three days. Cregger's last picture, Weapons, also beat its projections, opening to $43 million and finishing at $270 million globally on a $38 million budget, with an Oscar for Amy Madigan.

So: a director whose first feature was Barbarian has now, twice in thirteen months, outperformed a studio's own tracking by a third or more. That is not a genre trend. That is a person the audience has decided to follow, which is the rarest asset in this business and the one the industry has spent twenty years insisting no longer exists.

Practical Magic 2 is the control group. Warner Bros. spent a reported $75 million on a sequel to a film people genuinely love, opened it in more theatres than the number-one movie in the country, and has $50 million domestic and $76.8 million worldwide to show for it. Exhibitors keep roughly half of every ticket. That picture does not get out of the red theatrically. It is not a disaster of taste; it is a disaster of arithmetic, the kind that happens when a studio confuses affection for a title with demand for a title.

The rest of the top ten is stranger and more instructive. Spider-Man: Brand New Day is in its fifty-second day and has $944.5 million domestic. The Odyssey is on day sixty-six with $609.8 million. Those are not this weekend's story; they are last summer's, still paying out, still occupying 3,017 and 1,546 screens respectively. Meanwhile the 40th anniversary re-release of The Transformers: The Movie, a 1986 animated film, placed eighth with $3.1 million — ahead of two new wide releases.

The entire reporting universe of 32 films generated $110.3 million. More than half of that was one horror movie.

What I take from this is not that horror is hot, which everyone says every September. It is that the theatrical business has bifurcated into two things that no longer explain each other. There is a very small number of events — a Spider-Man, an Odyssey, a Cregger — that behave like utilities, printing money for months. And then there is everything else, which now opens into a market with no floor whatsoever, where a well-liked IP and four thousand screens buy you nothing.

The studios understand the first category and keep trying to manufacture it. They cannot. What they can do is notice that the man who made Barbarian on a modest budget is the only director in America whose name currently moves tracking numbers in the wrong direction on purpose, and stop asking him to explain himself.

Let me state the awkward arithmetic of Mecklenburg-Western Pomerania plainly, because the German commentariat will spend the week not stating it.

The AfD took 38.2 percent, more than doubling its 2021 result. It will not form a government, because every party except the BSW has ruled out cooperating with it, and the BSW missed the threshold by two-tenths of a point. The SPD took 35.5 percent, lost the election, and will in all likelihood continue to run the state. The CDU took 4.9 percent — the worst result in its history in any state — and its national leader remains Chancellor of the Federal Republic.

Every one of those outcomes is legal and procedurally impeccable. That is the problem I want to argue about.

The *Brandmauer*, the firewall, was constructed as a defensive instrument. Its premise was temporal: isolate the far right, deny it the legitimacy of office, and it will decay, as pariah movements historically do when they cannot deliver anything to anyone. That was a reasonable bet in 2017 at 12 percent. At 38.2 percent in one state and 43.8 in Saxony-Anhalt a month ago, it has stopped being a quarantine and become a permanent constitutional arrangement — one in which roughly four voters in ten are represented in a chamber but structurally excluded from any executive outcome, forever, by agreement among the other six.

I do not think that arrangement is illegitimate. I think it is a policy, and policies have effects, and the effect of this one is now measurable and running in the wrong direction.

Here is the mechanism. A protest vote that can never become a governing vote never has to survive contact with governing. The AfD in Schwerin will never have to publish a budget, never have to explain why the promised savings did not materialise, never have to disappoint anyone about anything, because it will never hold a ministry. The firewall has generously relieved it of the single force that destroys populist parties everywhere: office. Meanwhile every excluded voter receives, on schedule, a confirmation that the system is rigged against them, delivered by the system, in public, with the arithmetic attached.

The usual answer is that the alternative is worse — that governing with the AfD normalises it and the historical stakes in Germany are not hypothetical. That answer has real force and I will not pretend otherwise. But it is an argument for the firewall's necessity, not its sufficiency, and the necessity argument has been doing all the work for nine years while the number went from 12 to 38.

What the firewall was never supposed to be is a substitute for the CDU being a functional party. And that is what it has become. Merz's Christian Democrats did not lose to the AfD in Mecklenburg-Western Pomerania. They lost to everyone. They finished sixth, behind the Greens, out of the chamber entirely, and they did it while holding the chancellery and running a reform agenda that members of his own coalition spent Monday attacking in public. You cannot outsource your electoral proposition to a moral cordon. The cordon does not campaign. It does not fix a pension system. It does not explain to a voter in Schwerin why their party of forty years should get their vote back.

Merz says he will stay and continue the course. Fine. But somebody in Berlin should say out loud what the weekend actually demonstrated: the firewall is now holding up a governing coalition that the electorate has twice declined to endorse, and a defensive wall that is bearing load is not a wall. It is a structure. Structures need maintenance, and someone has to be able to say what happens when the number reaches fifty.

Nobody in German politics currently has an answer to that question. They have eighteen months to find one.

The dominant failure mode of the news is not inaccuracy. It is the announcement that never gets audited.

A government declares a policy. A company announces a capacity. A central bank pencils in a path. Each of these generates a headline on the day, and each contains, buried in it, a date on which it either happened or did not. Almost nobody goes back to that date. The announcement is free, instantaneous and permanent in memory. The mechanism is slow, contingent and usually smaller than advertised — and by the time it fails to arrive, the newsroom has moved on and the reader has no way to score anyone.

So starting with this issue, we are publishing The Wall. It is a short list, at the bottom of this column, of the specific dates on which the claims in today's paper become checkable. We will keep it. We will return to each item on or after the date. When we were wrong, we will say so in the corrections block, with a severity rating, where you can find it.

This is not a gimmick and it is not humility. It is the opposite. It is the most aggressive thing a small paper can do, because it means every forecast we publish carries a bill that comes due in public.

Two notes on method while we are here.

First, we will keep telling you which participants have money on the outcome and which do not. Today's front page describes four elections that produced no government. That is an interesting observation. It is not a tradeable one. The tradeable one is that the French sovereign spread hit 104 basis points over Germany on Friday for an election that happens in 2027 — because bond desks are obliged to settle a position every afternoon and parliaments are not. When a politician's statement and a trader's position disagree, we will show you both and tell you which one costs the speaker something.

Second, we will not pretend a number is a fact when it is an estimate. Kpler's provisional figure for Saudi September exports is just over four million barrels a day. Provisional means provisional. We have said so in the piece, and if it is revised we will publish the revision at the same length.

Write to us. Tell us where the reasoning is lazy, which of these dates we will quietly skip, and what you think we have got wrong. We publish objections. We would rather print a good letter than a mediocre column, and we will make that trade whenever it is offered.

---

## The Wall

29 September — The Riksdag session opens in Stockholm. Sweden can elect a prime minister no earlier than the following week. Kristersson's caretaker government has run the country since 17 September.

5 October — Intel is reported to raise PC processor prices by around 10 percent, its third increase since late 2025. We will check what actually shipped, and at what price.

Mid-October — Aramco's bypass around the damaged East-West pipeline pumping stations. The internal target is roughly half of Yanbu's throughput, two to two and a half million barrels a day, within about a month of the 10 September strike. Kpler puts that at even odds.

Late October — The Federal Reserve's next meeting. Futures put a second hike at 53 percent. The dot plot says 4.00 to 4.25 percent by year end.

23 October — Moody's reviews France, currently Aa3 with a negative outlook. Fitch and S&P are already at A+. Scope cut to A+ last week; DBRS moved its AA trend to negative.

2028 — The year every memory executive has given Intel's chief executive as the earliest date for relief on DRAM. We will not wait until then to check it. We will check the quarterly capacity announcements against it.