1A war in the Gulf has taken a July rate cut off the table
The Iran war ran through the whole of the day's news, across Reuters, Bloomberg and CNBC coverage: Brent above $100, physical cargoes changing hands nearer $110 than the paper price, and disruption through both the Strait of Hormuz and the Red Sea. The clearest read on what that does to policy is not in the commentary but in the pricing. Polymarket's contract on the 29 July Fed decision — roughly $95m of volume, the deepest on the board — puts no change at 82.2% and a 25bp hike at 17.0%, against 0.4% for a cut. The argument in front of the Fed this month is hold-versus-tighten. Easing is not in it.
The mechanism is the uncomfortable kind. A supply-side oil shock lifts near-term inflation without lifting demand, so it pushes a central bank away from cutting at exactly the moment growth risk rises. The bond market has taken the same side: the 10-year sits near its highs at about 4.71%, with the week's coverage framing it as oil battering bonds.
The disruption is priced to last. Polymarket's contract on Hormuz traffic returning to normal by 31 August trades at 15.5% yes against 84.5% no, on about $2.87m — far thinner than the Fed contract, but one-directional. The physical market says the same thing with money: tankers are paying roughly $2.5m and an extra month to route around the Strait, and CNBC's reporting raises the possibility that traffic does not normalise for a year or more. The political posture is hardening rather than converging — Truth Social posts on 22 and 23 July threatened to destroy an Iranian asset for every ship fired on in the Strait, and demanded Iran pay for damage to ships and cargo.
One dissent is worth holding on to, because it points the other way. On Excess Returns, a clip of Andy Constant floated a Fed under Warsh running a bull steepener — cutting short rates while shrinking the balance sheet — with Constant adding that he would not mind stocks down 10-15% to rebalance asset holders against Main Street. That is a preference about a future Fed, not a forecast for Wednesday, and it is the opposite of what the July contract prices. The same segment made a second long-horizon argument: that recessions are simply rarer now, because a services-and-tech economy is less cyclical than a manufacturing one, household and government balance sheets absorb shocks, and policy intervenes sooner. Both are decade-scale claims. Neither touches Wednesday, and an oil shock a central bank cannot ease against is what sits between here and there.
2The carry-trade story for the Nasdaq's worst July in 22 years needs a dollar reversal that hasn't begun
Felix, on Felix & Friends, brought the day's most complete single thesis, and it starts from a real number: the yen is at a 40-year low against the dollar while the Bank of Japan has taken its policy rate to 1%, a 30-year high. That pairing, he argues, unwinds the yen carry trade — money borrowed cheaply in Japan and parked in US assets — forcing sales of US tech and Treasuries, which is his explanation for the Nasdaq posting its worst July in 22 years. The sharpest part is his warning about the shape of it: a 60/40 book does not protect you here, because stocks and bonds fall together. He sizes the carry trade at anywhere from $1trn to $20trn — a range wide enough to tell you the number is not known.
He is careful about what is not driving it. A viral claim that a provision of the Japanese civil code is forcing repatriation gets dismissed as noise; the structural case is what he leans on, that Japan now has inflation, loan growth and a trade surplus, so its capital wants to come home.
What follows is a three-phase book. First protect: trim high-valuation US tech funded by cheap debt, cut leverage and margin, hold cash at 4-5%. Then hold value: gold, citing Goldman Sachs's $4,900 and JPMorgan's $4,500 year-end targets and China's central bank adding 15 tonnes in June, alongside healthcare and utilities. Then rotate into the weak-dollar winners: emerging markets, US multinationals, currency-hedged Japanese banks, commodities.
And that is where the thesis argues with itself. Phase three needs a weaker dollar. The dollar is at a 40-year high against the yen — the very fact the first phase is built on. The unwind and the rotation cannot both be the near-term trade, and only the first half is consistent with the current tape. Gold is the second oddity: it is bid while the dollar is strong, so the two are rising together rather than inversely, which makes it a safe-haven and central-bank-demand story rather than early confirmation of a dollar decline.
Two things to weigh it with. The presentation is a funnel to a paid live seminar, promising the actual names he will buy, wrapped in urgency about the next four weeks — the analysis and the sales pitch arrive together. And the call carries no dated catalyst. What would settle it is simply whether the dollar turns.
3AI spending still has no measured return at the level that would justify it
Excess Returns gave its weekly segment to the question the AI capex debate keeps deferring: is anyone actually making more money? The test is Bob Elliott's, and the segment's answer is that it stays unanswered — token prices are venture-subsidised, so what a user pays for AI is not what AI costs, and the real-economy return has not shown up in the numbers.
Azeem Azhar, in a clip on the same show, supplies the reason it may stay unanswerable for a while. You do not get an AI-native company by buying more co-pilot licences, any more than factories got the assembly line by adding light bulbs; the reorganisation is the point, and most firms are one or two steps into it. That produces a J-curve, with an awkward consequence: a good adopter and a bad adopter look identical from the outside right now, because both are spending and neither has earned it back.
The consequence is a warning about method rather than a call. For this year at least, the income statement cannot separate the companies getting a return on AI from the companies merely spending on it. Both of the day's videos are frameworks rather than dated calls — there is nothing in either that a print next week can check, which is its own kind of caveat.
4The AI trade has stopped moving as one thing
The semiconductor complex moved in both directions at once. was up about 4% pre-market on an earnings beat, while fell 21.5% on its own weakness — same industry, same session, opposite tape. carried a Goldman Sachs upgrade that repeated from the previous day rather than a fresh one. slipped 0.92% and was under discussion precisely over how discerning the AI trade has become. cut both ways inside the same coverage: the AI trade staying on reads as good for the semis it buys, while its own capex and free-cash-flow overhang persists.
Above the single names, CNBC's sell-chips-buy-software rotation appeared for a second day running — a reiterate, and worth marking as one rather than counting it twice.
The mega-cap headwinds were name-specific too. A 24 July Truth Social post named EU fines of $15bn on , $3bn on and $2.5bn on , framing Brussels as taking aim at American companies. The tone is friendly to those three; the fines are still a cost carried by those three and not by the rest of the index.
That ties back to the previous section. With the return on AI spending unmeasurable from an income statement, the differentiation is happening on the things that can be measured — a quarter's beat, a fine, a capex line.
What the sources recommended
One buy came out of the day's two videos, and the person making it hedged it three ways.
— Felix's long-term Nasdaq dip-buy after what he calls the worst July in 22 years, framed on roughly a ten-year horizon. It sits inside a near-term book that is explicitly defensive: trim high-valuation US tech funded by cheap debt, cut leverage, raise cash. So it is a decade-horizon accumulation call embedded in a de-risking posture, not an entry for this week. It also arrives inside the promotion for the paid seminar noted above.
Nothing was named as a sell. The exit-adjacent material was thematic — the same instruction to trim high-multiple US tech funded by cheap debt, and the chips-to-software rotation running through the news flow — neither of which is a single-name call.
Felix's other winners are asset classes rather than tickers: gold, emerging markets, commodities, currency-hedged Japan, healthcare and utilities. and came up only as illustrations of the weak-dollar multinational — a business whose earnings translate better when the currency moves, without the business itself improving. Examples, not picks. Excess Returns named no ticker at all; its material is the framework in the third section.
The hold-versus-hike argument gets settled on Wednesday
The July Fed decision lands on 29 July with no change priced at 82.2% and a hike at 17.0% — the direct test of the first section's claim, that a supply-side oil shock takes the easing option off the table. reports around 28 July having fallen 14.9% into the print, and bank earnings season continues with expectations of strong profits from the large lenders (). The longer-dated test is the Strait itself: the Hormuz normalisation contract runs to 31 August and prices a return to normal at 15.5%, so if traffic recovers earlier than that, the energy premium unwinds faster than anything on the Fed's calendar can do.
