1June's disinflation killed the July hike without buying a single cut
June CPI printed at 3.46% year on year, down from roughly 4.2%, and −0.42% on the month — the largest monthly drop in more than six years. The repricing that followed was violent and it went one way only. Polymarket's July FOMC contract, the deepest market on the venue at $59.8m of volume, now prices no change at 90.5%, up 30 points in 24 hours, with a 25bp hike at 8.8%, down 28.4 points. Take away the inflation justification and the hike leg did not survive the session.
What did not move is the cut side. The same board prices a 25bp July cut at 0.5%. The separate contract on how many cuts land in 2026 has "no cuts" at 80.5% on $6.1m of volume, up 0.7 points on the day — a soft print bought nothing at all on the easing side. The full-year hike contract fell 18 points to 52.5% on $4.1m, which still leaves a 2026 hike close to a coin flip. The market's read is hold, for longer, and reading today as the start of an easing cycle reads past what the prices say.
One casualty is worth naming, because it shows how quickly the ground moved. Prof G Markets published this morning quoting Kalshi at 67% odds of a hike before year-end, with inflation described as 4.2% and rising. The episode was recorded before the release. The comparable Polymarket contract now reads 52.5%, and the inflation figure it cites has been superseded by the print itself. President Trump posted the Labor Department release, citing annual inflation of 3.5% and the biggest monthly drop in over six years, which lines up with the 3.46% computed from the source data.
The counter-argument is underneath the headline rather than against it. Eric Pachman, on Excess Returns, points out that between 42 and 45% of the roughly 180 components of the CPI basket have breached one standard deviation above their historical mean, and that the headline stays muted mainly because shelter and owners' equivalent rent — 26 to 27% of the weight — are muted. On that reading, today's number is a composition result rather than broad cooling. It does not undo the repricing. It is the reason to treat this as one soft print rather than a trend.
2Equities looked through tankers being hit in Hormuz, because the two shocks run on different clocks
Two opposing forces ran on the same session. The disinflation print pulled the discount rate down. Against it, US strikes on Iran resumed, Iranian missiles struck tankers in the Strait of Hormuz and killed a sailor, the blockade was reinstated, and a 20% toll on transiting cargo was proposed. Crude rose 2% to a one-month high.
The tape sided with the print. The S&P 500 is up 0.42% at 7,546.69, the Nasdaq up 1.02% at 26,136.46, and the VIX down 4.5% to 16.39. The mechanism is timing rather than importance: the inflation number is realised and dated, and it moved the policy path inside the session by 30 points, while the oil impulse is prospective and only becomes inflation if crude stays elevated long enough to enter the index, which takes months. The market is not judging the war to be small. It is judging it to be slow.
It is not judging it to be over, either. Polymarket's contract on Strait of Hormuz traffic returning to normal by 31 December sits at 55.5%, down 3 points on the day, on $5.1m of volume — drifting the wrong way as the hostilities resumed, and pricing a live multi-month disruption rather than a headline.
The transmission channel to watch is diesel, not gasoline. Pachman, a chemical engineer who used to model refineries at ExxonMobil, makes the point that "oil is not oil": US light sweet shale is mismatched to Gulf Coast refineries configured for heavy crude, so a closure scrambles routing in a way he likens to COVID for oil markets, and light-sweet-configured refiners are earning crack spreads near $60 a barrel, four to five times their historical margin. On the day the bombing resumed, gasoline drew about 2m barrels and diesel about 3m, and diesel jumped 13%. Diesel feeds freight surcharges, then groceries, then services — the slow, sticky path into the index. That is why a benign June print and a fourth-quarter re-acceleration would be sequential rather than contradictory. He names , and only as refinery archetypes and expresses no view on them, and his anecdote about , and cutting shelf prices into rising diesel surcharges comes with his own statement that he has not verified it.
The administration's framing sat oddly against the tape. President Trump posted that "Oil is flowing like never before, thanks to the awesome Power of the United States Military", and separately touted low oil prices as an economic win, on a day crude rose 2%.
3The day's tape paid for compute and took it out of software
fell 22 to 24%, its worst day in 39 years, on a second-quarter miss accompanied by a rare revenue warning. The explanation offered across the coverage is a crowd-out — AI hardware budgets eating the software and IT-services line — which, unlike a sentiment story, comes with a profit-and-loss consequence attached. went the other way, with Goldman doubling down on the name.
is up 15% on the week and 6% on Friday after launching Muse Spock 1.1, claimed to run about 25% cheaper than frontier models. Alex Kantrowitz, of Big Technology, on Prof G Markets, thinks the rally is not about the model at all: his read is that it followed Mark Zuckerberg confirming Meta may license out spare compute at high margins, because the models themselves are commoditising. Same conclusion from the opposite end of the stack — the scarce, paid-for thing is capacity.
The day's other large move had no fundamentals in it. SK Hynix fell more than 15%, its biggest one-day drop on record, dragging the KOSPI down 9% and triggering an index-wide halt; its new Nasdaq line, listed on Friday, fell 9%, and , and sold off in sympathy. Luke Kawa of Sherwood News, whose account Prof G Markets carried, finds no fundamental catalyst: 16 single-stock leveraged ETFs on Samsung and SK Hynix launched in late May have tripled to more than $9bn, those ETFs plus the two underlying stocks now account for more than 70% of all Korean trading volume, and retail bought the leveraged products with margin loans — leverage on top of leverage. Pro-cyclical daily rebalancing did the rest. Kawa's own view is that memory fundamentals are strong, with rising earnings revisions, intact pricing power and optically cheap forward multiples, making this a positioning flush rather than a cycle top. is rebounding today, which is consistent with him. Only one of the day's two big moves is the reversible kind: a flush unwinds, a revenue warning is a new number.
The same contest over hardware is now running in court. Apple filed a federal trade-secrets suit against OpenAI on Friday 10 July, alleging its hardware effort is "rotten to its core" and that it mined job candidates for confidential information. Kantrowitz thinks is dead to rights on the facts, since the alleged espionage happened on Apple-owned machines, and expects no settlement and a painful discovery process. OpenAI's hardware unit has no revenue, so there is no near-term earnings effect either way — but an injunction could force a rebuild from scratch, which is a problem for an IPO story built partly on roughly 400 ex-Apple hires, Jony Ive among them.
4US job growth has been the care economy, and the cut to it is already legislated for 2027
"You will not see the recession coming with the unemployment rate." That is Eric Pachman, formerly of Bancreek, on Excess Returns, and his case is built from the composition of the jobs data rather than its headline.
Start with what the 2.5 million missing workers are not. Pachman is explicit that the number is a model-revision artefact: the Census Bureau and the Bureau of Labor Statistics re-aged the population, reclassifying working-age people into the 75-plus bucket, so no humans disappeared — the projections simply carry fewer future taxpayers. What is real is participation. The labour-force participation rate fell 0.4 points to 61.5%, buried under an "unemployment unchanged" headline. For the over-45s it has gone from 55.9% in 2009 to 49.6%, with men over 45 at 54.7%, the lowest on record.
Then the quality of what was added. Healthcare accounted for 60 to 80% of payroll growth, and by the end of 2025 he puts it at two, three, even four times all job growth — which is only possible if everything else was contracting. Not doctors: home health and personal care aides are now the single largest occupation in the country at about 4.34 million people, median pay $17.21 an hour, roughly $35,800 a year. Wages at that level leave close to no disposable income, so the jobs that were created feed consumption weakly.
The dated part is what turns this from description into a catalyst. The One Big Beautiful Bill cuts Medicaid by about $1trn over ten years, with most of the cuts starting in the first quarter of 2027. On the Bureau of Labor Statistics' own county-level employment data, 32 or 33 states would have shown negative job growth through the end of 2025 without the Medicaid-funded care economy.
Three things sit against it, two of them his own. He says outright, "I can't tell you which way the market's going to go," and frames the whole thing as a quantified tail risk that could even resolve upward — no job growth, then rate cuts, then speculation. The county-level data runs six to nine months in arrears. And the 2027 cut is legislated but not executed, which makes it politically soft going into a midterm. Polymarket's US recession contract sits at 10.5%, but on about $21,000 of liquidity and with an end-2026 horizon — so it is not evidence against a 2027 event, it is a market that does not reach that far.
5The case that AI's credit window is closing rests on three months of issuance data
Felix, on Felix & Friends, argues that the AI build-out is debt-funded rather than cash-funded, and that credit cracks before equity does. The spine of the evidence is a sequence: AI-related borrowing of roughly $75bn in February, $50bn in June and $25bn in July, which he reads as lenders withdrawing rather than borrowers restraining themselves. Around it he assembles circular-financing examples — investing in startups that buy chips, 's $13bn into OpenAI returning as Azure revenue, and funding Anthropic which buys their cloud, and as the most debt-dependent of the six large AI issuers — plus Financial Times reporting that JPMorgan and Morgan Stanley have quietly been tranching and offloading AI debt, Apollo's observation that non-tech companies show no evidence of earning more from AI, and SpaceX's post-IPO $25bn bond trading about 10% below issue with no operational change. His contagion chain runs from lenders refusing, to borrowing costs rising, to the $5.8trn build-out being cut back, to the circular revenue evaporating, to the Mag 7 — about a third of — falling, and finally to passive and retirement-account holders. He is explicit that it is not a call: "I'm not saying sell everything tomorrow."
Three things weaken it, and they belong in the same breath as the thesis. The headline claim that Goldman said the same thing before 2008 rests on an internal note from one derivatives trader reporting the word "carnage" on the credit desk; the 2008 equivalence is the presenter's own framing, not a quoted position. Three monthly issuance figures describe front-loaded funding or seasonality just as well as a closing window, and the thesis carries no trigger level, no timing and no condition that would falsify it. And the circular-financing frame flattens a real distinction: , and generate very large genuine external operating cash flow and are investment-grade issuers with real earnings, which is structurally unlike a borrower with no income.
The packaging deserves its own line, because it bears on how much weight the argument can carry. The video is built as an advertisement: two lead-capture offers, a paid academy behind them, escalating urgency, and a closing segment pitching a co-host's past picks as turning $10,000 into $115,000 — an odd note to end a crash warning on. The only dated catalyst in it is the presenter's own free webinar on Saturday. The underlying citations are worth chasing independently; the framing around them is selling something. Which leaves the credit question open, and the loudest version of it the least usable.
The hawkish case now has to be re-argued in front of a 3.46% print
Warsh's congressional hearing is running today, and it landed on the same day as the release — so the hawkish framing that carried into this week has to contend with a 3.46% annual print and a 90.5% probability of no change at the July meeting. Second-quarter bank earnings opened before the bell, with , , , , and reporting; soft inflation plus bank results are what carried the tape. , , , , , and follow tomorrow morning. Underneath all of it, US strikes and Iranian attacks on shipping continue, with the blockade and the proposed 20% transit toll still live — the largest source of tail risk into next week, and the one development that could put the inflation question back where it stood this morning.
