The oil price stopped being about oil

**Crude ran better than six per cent in a single session to $102.48 and Brent settled at $107.63, carrying the barrel +12.60% across six settled sessions. The metals that are supposed to travel with a war went the other way — silver led them down, copper close behind, gold falling with them — and the dollar rose. The part of the curve that moved most was the near end.**

The instinct on a day like this is to reach for the war. The United States and Iran spent the session exchanging attacks on shipping around the Strait of Hormuz, crude ran more than six per cent, Brent settled above a hundred and seven dollars, and every headline wrote itself. But the rest of the tape refused to co-operate with that story. Gold fell. Silver fell four and a half times harder than gold. Copper, which has no haven property at all, fell almost as much as silver. The dollar rose. If capital had been running for shelter, the two metals it historically runs to would not have been the day's worst performers.

What did rise was the front of the yield curve, and the shape of that move is the argument. The five-year added +12 basis points and the ten-year +11, while the thirty-year managed +7. A shock to an issuer's credibility is expressed at the long end, where the term premium lives, and it steepens the curve; a shock to the expected policy rate is expressed at the front, and it flattens it. This flattened. The market was not pricing barrels going missing — it was pricing what the central bank will have to do about barrels becoming expensive. Next Wednesday's Federal Reserve decision is now carried with a hike as the consensus rather than a hold, and the European Central Bank spent yesterday raising rates while warning that its own inflation problem would last longer than it had previously said.

Two things made that conversion possible, and one of them is ours to mark. On 3 and 4 September this desk published the August payroll release as a dated catalyst and framed it as the test of whether a weak labour print still reads as dovish. The print was not weak: 162 thousand against a consensus of 56 thousand, off a prior month of 21 thousand. The question was the right one and the direction was wrong, and the error is load-bearing — an energy shock arriving on a decelerating labour market is a growth scare a committee looks through, while the same shock arriving on a labour market that has just re-accelerated is an inflation problem a committee has to answer. The second thing is that nothing has physically stopped. The engine's tanker-departure series counted 328 departures through the strait against a thirty-day average of 325.2, across 31 sessions with no gaps. The barrels are still sailing. The price of them is a premium, not a shortage.

Which brings the argument to the one desk that has not moved at all. The engine's consensus panel — the daily record of where sell-side analysts sit on every covered name — shows that across fourteen of the largest listed energy companies the rating distribution has changed by 4 single-analyst moves since 2026-08-03, with 10 of the fourteen panels rating-for-rating identical, Exxon and Hess both still carried at Hold. Over the same weeks that panel migrated for 1,115 of the 13,292 names it covers, so this is not a stalled feed; it is a choice. Ratings trail estimate revisions and estimate revisions trail the price deck an analyst is willing to underwrite, so a panel that will not move while spot runs twelve per cent is a panel telling you it does not believe the strip. That is a falsifiable position and it is the opposite of what the front end has just priced. One of the two is wrong.

The wider frame is about what a central bank can absorb. The reflex of the last three years — that a geopolitical energy shock is disinflationary in the end, because it destroys demand faster than it removes supply — was learned in a world where the policy rate had room to fall. It does not have that room now, and the cost of the answer is already visible in the place it always lands first: the thirty-year mortgage has crossed seven per cent for the first time in over a year, and August home sales are reported at their weakest in more than a year. When the cushion is gone, the same barrel produces the opposite policy reflex. That, and not the strait, is the regime change worth naming. This read is wrong if today's consumer price index prints headline at or below two tenths against a four-tenths consensus — that would say the energy move never reached the index and the near end has to give back what it took. It is also wrong if gold turns and rises with crude while the front end keeps selling, because that pairing is a credibility trade rather than a policy trade, and this desk would then be describing the wrong regime with the right numbers.

*Method: figures are drawn from the settled session of 2026-09-10 and every observed window ends there, not on the publication date. Port-congestion baselines exclude stored outage days, which otherwise inflate a thirty-day average; where a feed's outage is recorded as a low partial rather than a zero, the affected window is dropped rather than averaged. Alt-asset results are counted by distinct sale, since the underlying table carries a completed sale forward on subsequent dates.*