The Strait of Hormuz: reading toward an endgame. The signal to watch is narrow and specific: not the President's words, which will claim victory in either direction, but the scheduling of a US-Iran round and the posture of the US Navy in the southern corridor. A stand-down there is the tell that the win is real. Until then, this publication treats the Strait as reopening in rhetoric more readily than in fact — and the toll, in whatever form survives the reopening, as the feature of the arrangement rather than the casualty of it.
The week gave that toll a respectable co-owner. On the twenty-sixth of August the Revolutionary Guard confirmed a revenue-sharing agreement with Oman covering each country's share of the Strait's waters and revenues; Brent eased on the news. A fee collected by the IRGC reads as extortion; a revenue-share administered jointly with Oman, a trusted neutral, reads as two coastal states managing their shared waterway — harder for the IMO or Washington to contest, and advancing Tehran's actual war aim from the outset: not the toll, but its legitimisation.
For the investor seeking an end goal, the structure is a sequence, not an event. The Iran-Oman precondition has now cleared; mediators have said the push to revive US-Iran talks follows it, so the leg that matters for oil begins now. A framework already exists in the June memorandum, and prior rounds have compressed from sixty-two days to eight. A political stage sits roughly four weeks out in the Xi visit of the twenty-fourth of September. Against this runs one countervailing force: the President has declined to schedule talks, insisting he will wait until Tehran is "ready" on his terms. The duration is therefore open — but bounded, pointing to weeks rather than months, and gated on his price being met rather than on Iranian delay.
That price is transactional, and it is the reason nothing has moved. The impasse is a compensation dispute in both directions: Iran wants reparations out, the President wants compensation in and refuses to be the payer. Iran learned after the June memorandum that the President does not pay; the reparations-out demand is the true blockage, fused to the question of who won. The likeliest end state is one in which it is dropped, deferred, or laundered through a Gulf mechanism so he never visibly pays — while he banks a face-saving sum and claims a reopening.
The end goal to hold in view, then, is not resolution but a claimed reopening: a declaration, staged for effect, delivering an immediate drop in crude and at the pump — gasoline fell over ten per cent in a day when Iran agreed to reopen in April. Within a week the President was accusing Tehran of "doing a very poor job," and the arrangement frayed. Relief is not resolution. The reopening will move prices before it moves a single tanker; whether it stands a fortnight later is the separate question, and the one that decides whether this is a turning point or another false dawn.
Which gives the Strait its clock, and the clock a face: the twenty-fourth of September, when the President hosts the Chinese leader and would want cheap oil for a backdrop. If the reopening is coming, it is likelier to come before that date than after it — the interval in which it can still be staged for maximum effect is measured now in weeks. It is the first of several such dates this issue keeps in view. The summer has been an interval, a held breath; it is running out, and the diary that replaces it is unusually crowded. Each section that follows carries its own clock, and each clock runs to the same autumn.
The first of those clocks has already struck. Warsh, in his first Jackson Hole address as Fed Chairman — and one criticised beforehand for ambiguity — struck a hawkish note, warning that inflation had not meaningfully slowed and that the Fed "may have work to do," effectively taking September rate cuts off the table and putting a hike back in play. Markets repriced the front end accordingly: the two-year yield, at 4.36 per cent, jumped toward a ten-year at 4.73, flattening the curve without inverting it, as money-market odds of a September hike moved to roughly even from about one-in-three the week before. From a Chairman faulted for reticence, an unusually plain instruction.
The Budget is weeks away, but its first tea-leaves spell deferral. When John Healey resigned as Defence Secretary this summer, it was on a matter of substance: the Prime Minister, he wrote, had been "unable" and the Treasury "unwilling" to fund the nation's defence, and three per cent of output by 2030 was the vital test. He returned within the month as Chancellor — and his first Budget, it now emerges, will decline that same target, the hard roadmap pushed to a 2027 spending review. Not a decision taken but one postponed beyond the current fiscal pressure — and defence is unlikely to be the only file marked 2027. And the shortfall now carries a price tag from an unexpected quarter: The Telegraph reports that Pentagon officials, pressing NATO allies toward the five-per-cent pledge, have floated withdrawing US support for British sovereignty over the Falklands unless defence spending rises — a single-sourced leak of internal discussion rather than a stated policy, and first aired in a memo Rubio played down in April, but a pointed reminder that the transactional alliance now prices the guarantees it once gave freely. Here is the Chancellor, ahead of his own Budget, answering the question the long end has asked all summer: when principle meets the cash available, principle yields, and the reckoning is booked for later. Postponement is what a fiscal premium is the price of. The 30Y closed the week yielding 5.78 per cent, the 10Y at 5.16, and the spread over US Treasuries a fatter 43bp — moving in its own existential orbit.
And the calendar sharpens it. This is the last week of summer: after the bank-holiday Monday the desks fill and the serious decision-makers — the ones who set risk rather than mind it through August — are back. The overseas third of the register, which extended the summer its patience, does not extend it past the holiday. That clock has weeks on it, not months, and runs to the same autumn as the Strait's.
The most instructive position in the equity market this summer is not a trade but a refusal to make one. Steve Eisman, who earned the right to be heard on concentrated risk when he called subprime, has reduced his own exposure — and frames the danger exactly as he framed the last one: it is all one trade. Some seventy per cent of the artificial-intelligence revenue reaching the large cloud providers flows, on his estimate, from just two private laboratories, OpenAI and Anthropic, and the futures of companies worth trillions rest on a bet that those two succeed. He calls it the trade's weak point. And yet the man who called subprime is not shorting it. He holds cash and says plainly why: not until the trouble at those two firms begins to spread will the story stop running. The waiting is the position.
What he is waiting for is a document. The concentration he describes cannot be confirmed or dismissed from the outside, because the two firms it rests on are private, and the figures that would settle the question — the quality of the revenue, the margin behind it, the true cost of the compute, the concentration of the customers — are precisely the ones a public offering forces into daylight. This is the signpost worth marking above all others in this issue, because it has, this quarter, acquired a date. Anthropic filed confidentially with the Securities and Exchange Commission on the first of June and is expected to list on NASDAQ in October off a private valuation near $965bn — though rumours circulate of a $2 trillion ambition. The prediction markets put the offering before November at roughly seventy per cent. OpenAI, the larger of the two, is the laggard — reported to be leaning toward a listing in 2027, which means the first of the two accounts to open will be Anthropic's, and it opens within weeks. When it does, the most closely-read prospectus in years will settle a question no amount of commentary has: whether the demand beneath the most crowded trade in the market is real and merely narrow, or narrower than it is real.
That is the disclosure the bear is waiting on, and the reason his patience is the shrewdest available reading. He is not predicting the direction of the reveal; he is declining to guess ahead of it.
Nvidia reported for its second quarter on the twenty-sixth of August: revenue of $96.2bn, up 106 per cent on the year, data-centre revenue of $89.0bn, gross margins at seventy-five per cent, earnings of $2.46 a share — a beat on every visible line. The headline is a triumph, and an honest reading starts by conceding it. The composition beneath it is where the conundrum lies. Single-customer concentration eased: the largest direct customer fell to sixteen per cent of revenue, from twenty-one in the prior quarter, and the top three no longer cluster as tightly as they did. But the deeper concentration held: the hyperscalers together supplied some $48.7bn, more than half of everything sold, and the receivables remain owed by the same short cohort — a recent quarter had three customers standing behind close to two-thirds of the sum. The single name matters less than it did; the small circle of names matters as much as ever.
It is the concentration Eisman describes in revenue, reappearing one layer down as credit exposure — a handful of buyers on whom both the growth and the money owed depend. The same names recur up and down the stack: as customer, as counterparty, as the equity on the other side of the vendor's own investments. The bull reads that as the natural shape of an infrastructure build; the bear remembers Cisco, whose own vendor-financed boom read exactly the same way in 1999.
The publication holds the same discipline as Eisman, and for the same reason: the case for the bubble condition is already strong without the guess. The cyclically-adjusted earnings multiple sits near 42 — the reading crossed above that level in July for the first time since July 2000, at the start of the last great halving. The market-capitalisation-to-output ratio Buffett once favoured reads near 236 per cent, a record on the daily trackers. Both carry the same honest caveat: buybacks and changed accounting inflate the one against its own history, foreign revenue and mega-cap composition the other, and both drift structurally higher over time — so they are best read not as a timing signal but for what they plainly are, two independent gauges at or near record highs. Valuation states the condition. It does not start the clock.
What starts the clock is liquidity, and here the week supplied its news. Warsh warned that inflation has not meaningfully slowed and that the Fed may yet have work to do — and gave the front end its orders while the long end kept its counsel. The old dictum applies in the direction most who quote it forget: do not fight the Fed means, in a tightening season, do not stay long into the tightening. This publication's view — and it is a view, not a forecast — is that the risk and reward are poorly matched here: little room above an already-stretched market, real room below if a hawkish Fed meets a seasonally unkind autumn. That is an argument about asymmetry. It is not a call on the turn, which cannot be made, and which Eisman himself is conspicuously declining to make.
So the clock that governs the market's most crowded trade, which for a year has kept no calendar at all, has at last been given a face — and the face reads October. Eisman will not call it until he sees the numbers; the numbers arrive when Anthropic files. Until then his cash is not caution but method: a refusal to convict ahead of the disclosure, from the man who last time read the disclosure better than anyone. A clock with no hands is still a clock. This one has just grown its first.
| Gold | 4,460 | dramatic Friday correction; MACD reversal now bearish |
| Copper | 14,280 | neutral, low vol; MACD neutral, price sideways |
| WTI | 82 | lower, less uncertainty — Hormuz hope returns |
| Brent | 88 | eased on the revenue-share news |
| Carbon | 82.85 | low volumes, low volatility; support at 50 DMA |
| UST 2Y | 4.36% | Warsh — front end takes its orders |
| UST 10Y | 4.73% | curve narrowed; Sept hike back above even |
| UK Gilts 10Y | 5.16% | new order, same old dither and defer |
| UK Gilts 30Y | 5.78% | +43bp over UST — its own existential orbit |
| German Bunds 10Y | 3.29% | edging higher; ECB bearish, rates to rise |
| JGB 10Y | 2.92% | BoJ normalisation; carry under pressure, FX intervention pending |