1A September rate hike became the market's base case, priced off crude rather than inflation
Polymarket's September FOMC contract, which resolves on 16 September, now has a 25bp increase as the most likely outcome at 59.5%, up three points on the day, against no change at 38.5%, down two. The easing side is barely there — 1.9% for a 25bp cut, 1.0% for 50bp or more. At $2.83m of liquidity and $710k of 24-hour volume it is the deepest contract checked. The longer-dated market says the same thing more slowly: a hike landing somewhere in 2026 prices at 67.5%, up a point.
What makes that pricing awkward is the inflation data underneath it. June CPI fell 0.42% on the month, the first monthly decline after four hot prints (March +0.86%, April +0.64%, May +0.47%), and core PCE rose 0.13%, roughly 1.6% annualised, with the effective funds rate at 3.63%. One month is not a trend, and this one predates the oil move entirely. But on the prints alone there is no hike case at all.
The pressure is in the curve rather than in the data. The 10-year/2-year spread printed 0.47 on 31 July against 0.34 to 0.36 as recently as 27-28 July — eleven to thirteen basis points of steepening in three sessions, with the Fed having just held while divided and the Treasury sell-off continuing afterwards. That is a bear steepener: the long end selling while the front end stays anchored.
One part of the story does not survive contact with the series, though. The 10-year itself is not surging. It printed 4.68 on 30 July, inside a 4.61-4.71 range over the previous fortnight. The move is in the shape of the curve, not the level of the long yield, which is a far narrower claim than "yields are up".
Joseph Wang, on Fed Guy, adds a term-premium mechanism that has nothing to do with oil. New York Times reporting has Warsh wanting fewer FOMC meetings — possibly the statutory minimum of four a year — alongside no dot plot next year and lighter communication. Wang's chain is short: fewer meetings mean fewer chances to move, so the moves get bigger, 50s and 75s rather than 25s, which raises front-end volatility, which raises the risk premium, which lifts the whole curve. He rejects the stated rationale outright, calling the idea that "the market should decide interest rates" incoherent, because the market prices Treasuries off the expected path of a policy rate the Fed itself sets.
The consequence is that September pricing is not currently an inflation call. It is a term-premium and oil call, so it can unwind on a crude reversal without a single new inflation print. The Bank of England, which also held this week, said as much by explicitly waiting for a clearer read on war inflation before moving.
2The war is repricing energy through refineries and shipping lanes, not crude supply
The whole August WTI distribution moved up in 24 hours. The odds of touching $100 rose 10.5 points to 37.0%, touching $90 rose ten points to 86.5%, and touching $95 rose four to 61.0%; the tails at $110 and $120 sit at 17.5% and 6.4%. The downside faded in step — touching $80 fell 6.5 points to 69.5%, and touching $70 fell ten points to 15.5%. That is $665k of liquidity and $347k of 24-hour volume repricing in one direction.
The conflict contracts moved with it. An effective US-Iran ceasefire by 14 August fell four points to 34.5%, and by 31 August fell a point to 48.5%; the 31 July version has effectively resolved no at 9%. A US invasion of Iran before 2027 rose two points to 25.5%, on $1.29m of liquidity. Both are pricing escalation over the next fortnight rather than de-escalation.
Wang's mechanism explains why crude at roughly $90 understates the problem. The number that matters, he argues, is gasoline futures, because refining is the bottleneck rather than production — Ukraine is destroying Russian refineries while Houthi and Iranian strikes hit Gulf refineries. Crude can look contained while refined products do the inflationary damage.
The week's shipping news says the same thing from the infrastructure side. A drone caused a fire on two gas vessels at Egypt's Damietta port, near Suez. QatarEnergy bought 33 US LNG cargoes to offset Hormuz disruption, and moved its first LNG tanker out of Hormuz in nearly three weeks. Canadian oil sailed to Japan for the first time in over a year, and Tokyo Gas warned that a prolonged war keeps spot LNG elevated. The US strikes of 29-30 July ran two hours across dozens of targets; oil jumped 7% and then settled back on a proposed Saudi-led maritime defence coalition.
Two caveats travel with this, and Wang supplies both. The first is that his escalation call is a prediction about a weekend that had not happened when he recorded. The second is his own war-economics scepticism, which cuts against a clean escalation read: he argues the administration has bound itself to two constraints, no US casualties and no equity drawdown, that are incompatible with prevailing against an opponent willing to absorb pain, and he names Iran's theory of victory as inflicting enough economic damage to cost Trump the midterms and let a new Congress defund the war.
The consequence is that energy and rates are one story this week rather than two. The transmission runs from refining capacity to gasoline to the inflation print and out into the long end, which is the same loop the September pricing above is sitting on.
3AI spending accelerated into the drawdown, so the de-rating is hitting shareholders rather than budgets
Wang calls the week's rebound a dead-cat bounce, and his case is about positioning rather than fundamentals. The semiconductor index fell roughly 40% in a month, and memory names bounced around 20% in a single day — which he reads as a momentum-and-leverage move that a fundamentals-are-still-fine argument cannot rescue. He puts over a million retail speculators in the trade on leverage, many of them now wiped out and structurally unable to buy back, with the cohort that bought 40% higher selling into any strength. His evidence is Leopold Aschenbrenner's Situational Awareness fund, liquidated after running a reported 400% leverage into semis and AI — the visible tip, in his framing, of comparable damage sitting in smaller funds and US retail accounts.
The spending went the other way in the same week. beat, Azure passed $100bn of revenue with growth at a four-year high, and Microsoft committed more than $130bn to new datacentre leases. Amazon's Jassy defended the capex line. Samsung's Q2 operating profit beat on AI chip demand, and Apple's constraint is reported as a memory crunch. Memory and storage led pre-market all week, with , SK Hynix, SanDisk and recurring as the barometers, and Asian stocks posted their biggest gain in four months on AI optimism.
It was not uniform. Meta's mixed Q2 dragged peers lower, and Vertiv fell 17% on a weak quarter. But the two halves point in opposite directions: cloud revenue and lease commitments accelerated while the equities that supply them de-rated. That is a de-rating of the shareholders rather than of the budgets, and it is the strongest available answer to the positioning argument, because forced sellers do not change what a hyperscaler has already contracted to spend.
Wang flags his own risk twice, which is worth carrying. He concedes that the liquidation-event reading — forced selling exhausted, market free to run back to the highs — could be right, and says outright that he could be totally wrong. This is tape-reading and psychology, not a model. How he argues matters too: the whole twenty minutes runs at index and asset-class level, the semiconductor index, the memory complex, the curve, crude and the yen, with no symbol attached to any of it. That makes it a positioning read rather than a stock-picking one.
Two single-name moves in the week were idiosyncratic rather than sector reads. fell hard pre-market on AI-licensing friction with Google, per Wells Fargo, and Hims & Hers fell 10% as the FTC sued over its data and billing practices.
4Washington joined the yen intervention, and the rate gap behind the slide is untouched
The recent yen intervention was not Tokyo's alone. Wang says it was joint — the Bank of Japan and the Ministry of Finance, plus the US Treasury's Exchange Stabilization Fund, with euros sold for yen, reportedly visible in photographed notes from Bessent. His argument is that it buys time and nothing else, because the cause is the rate differential: Japanese inflation is comfortably above 2% while policy sits near 1%, and an intervention moves neither number. He also dismisses the "protect the Treasury market" rationale for it, on the grounds that foreign central banks hold their reserves in bills and the belly of the curve rather than in 10s and 30s, so their selling cannot dislocate the long end in the first place.
The consequence is a reset level rather than a changed direction. Until that differential narrows, the yen's path is decided at the Bank of Japan's policy meetings rather than by the size of the next operation.
The weekend marks these claims to market before any scheduled data does
Wang recorded his escalation call ahead of a weekend in which further US or Israeli strikes on Iranian economic targets were the expected next step, so the first read on the whole energy chain arrives before the calendar does. The ceasefire-by-14-August contract at 34.5% is the running scoreboard on it, and the August WTI ladder is where a strike shows up in minutes rather than weeks.
The scheduled tests come after. Monday 3 August has no earnings inside the window; Thursday 6 August carries the cluster — , , , , , , and . Then come the two prints that decide the rate question. July payrolls land on Friday 7 August, against June's +57,000 and an unemployment rate that fell to 4.2% from 4.3%. July CPI lands on Wednesday 12 August, and it is the first inflation reading that can contain the oil move rather than predate it. If it runs hot, the September pricing stops being a term-premium story and becomes an inflation one. If it does not, a 59.5% hike is resting on crude alone.
