1A September rate cut is off the board, and a hike is a coin flip
Polymarket's contract on the 15–16 September Federal Open Market Committee decision prices no change at 51.5% and a 25bp increase at 48.5%, with a 25bp cut at 0.35% and a larger cut at 0.15%. That book is the deepest thing checked today: $6.09m of resting liquidity, $100.7m of lifetime volume, and $1.72m traded in the last 24 hours. Read against the same contract 24 hours earlier — hike and hold both at 49.5% — the hold leg gained two points over the weekend.
The cash market is priced the same way. The 2-year Treasury closed at 4.34% on 3 September against an effective federal funds rate of 3.63%, so the front end sits 71bp above the policy rate. A cut is not a small probability in that pricing; it is absent from it.
The August employment report, released 4 September, is what moved this. Non-farm payrolls came in at 159,075,000 against 158,913,000 in July — a gain of 162,000, against July's own gain of 21,000. The unemployment rate held at 4.1%, unchanged from July and down from 4.2% in June. President Trump described the number as "nearly TRIPLE the predictions of the Bloomberg survey's so-called 'Economists'", and in a separate post argued the market should have risen on it rather than fallen.
The inflation market drifted the other way over the same 24 hours. Polymarket's contract on August core CPI, which the Bureau of Labor Statistics releases at 8:30am on 11 September, prices year-on-year at 2.4% with 42.5%, 2.3% with 32.0%, and 2.5% with 15.5%. Twenty-four hours earlier the same three legs read 47.5%, 28.0% and 21.0% — the whole distribution shifted down, reversing the previous day's move up. The month-on-month book has 0.2% at 55.0% and 0.3% at 31.0%. These are thin books — $1.4k on the modal leg and $2.2k of 24-hour volume across the event — so the direction is the signal and the level is not.
Energy pushes back against that cooling. Reuters reported record Labor Day weekend gasoline prices on 5 September and a record diesel price the same day, with a ship-fuel shortage forming as war-strained refiners reallocate output. The last released retail gasoline figure is $4.071/gal for the week ending 31 August, against $4.085 the week before and $4.006 on 10 August — the record-price weekend is not in that series yet. Core CPI excludes energy directly, but freight cost travels into it with a lag, and the print lands four days before the Fed meets.
One more thing is priced, and it is not about direction at all. On the dissent contract, zero dissents carries only 10.5%, three dissents 30.5% and four or more 31.5% — a 62% chance of three or more, on a $15.6k book that is too thin to lean on. The argument inside the committee is no longer about how fast to move. It is about which way, and whichever way it goes it is priced to split.
2The oil market prices a ceasefire and a still-shut Strait at the same time
US forces struck three Iranian crude oil carriers on 5 September, and both sides exchanged attacks on shipping into 7 September. Hormuz traffic fell to its lowest since May. Trump said on 4 September that the US may hit Iran's Pickaxe Mountain site soon.
Against that, Polymarket's US–Iran effective-ceasefire ladder rose at every horizon on the day: by 7 September 52.5% (+8.5pp), by 11 September 58.5% (+9.0pp), by 18 September 71.5% (+5.0pp), by 30 September 76.5% (+2.0pp). That is a full reversal of the previous day's fall, when the same legs dropped as much as 12.5 points on the tanker strikes. Reuters supplies the mechanism: Iran's Hormuz leverage is described as waning as the US economic squeeze bites, Iran said it would tackle economic problems, and its government raised fuel prices for heavy users.
The shipping book does not follow the ceasefire book. Normal Hormuz traffic by 15 September prices at 0.45% on $258k of liquidity, by 31 October at 9.0%, by 30 November at 17.5%, and by 31 December at 25.5%, down a point on the day, on $442k of liquidity. A US–Iran nuclear deal by 31 December sits at 10.5% on $308k. Iran full airspace closure by 30 September is 9.5%.
Crude has moved with the shipping book, not the ceasefire book. WTI printed $91.48 on 1 September, from $87.03 on 31 August and $84.57 on 28 August. September's $95 leg has already resolved yes; $100 now prices at 35.0%, up 3.5 points on the day, and a fall back to $80 at 29.5%, down four. OPEC+ left October output policy unchanged on 6 September, so no offsetting barrels are scheduled.
A ceasefire stops the escalation premium from building. It does not put the barrels back through the Strait, and these two books are priced consistently rather than in conflict: a pause in the shooting with the supply disruption intact.
3The largest capex cycle on record now runs on bond investors, not cash flow
Tom Wickenden, investment strategist at BetaShares, told Equity Mates that hyperscaler capital spending is heading for "trillions next year, the largest investment cycle in history" — and that the funding source has changed underneath it. "In years past they spent their own money, their own cash flow. Now we've moved to more money coming out of the bond market." He adds that spreads on that debt are "widening a little, but there are no problems", and that bond investors remain willing to fund it. The condition on the cycle continuing is now credit appetite rather than operating cash flow, which is a different failure mode than the one this cycle was underwritten against.
He is more confident about compute than about the layer above it. Cloud revenue and margins rose last reporting season, but capex is amortised over years, so in his words earnings quality is "still somewhat hazy". The combined order book across the three largest hyperscalers he puts at about $1.5 trillion. Open-source and Chinese models could commoditise the model layer — he names Anthropic at $65bn of annualised revenue, still short of profitability and private, so the revenue quality cannot be checked — but he notes that commoditisation still consumes compute. Bullish the infrastructure, cautious the models.
The margin evidence points at memory. disclosed a supply-and-capacity commitment of $279bn, against $119bn a quarter earlier, described in the filing as for data centre infrastructure systems "primarily memory and manufacturing facilities". Brian at Business With Brian screened 24 companies across the chain on one test — gross margin over the last four quarters against the four before — and the two that passed hardest are both memory. went from 37.7% to 84.6% gross margin in the same quarter a year apart, and is up 35.4 points on the four-quarter test; Micron's own CEO said some key customers could be filled to only "50% to two-thirds of their demand", with 16 supply agreements running to 2030 locking in about a quarter of revenue. kept 85 cents of gross profit per dollar against 26 cents a year earlier, up 41.4 points with revenue nearly tripled — and it sells nothing to Nvidia, which is his actual point: the scarcity belongs to the category, not to one contract.
The names that failed are as informative. grew revenue 46% while gross margin fell 2.2 points — his read is that bus bars and cooling loops are things a competent factory can build, so the margin flows to the buyer. passed, but by only 4.3 points on 26% revenue growth, despite customer prepayments doubling from $1.8bn to $3.6bn in six months. Ross Givens put $50,000 into Vertiv on camera a week ago, and Brian names that call and takes the other side of it — Givens read the market's one-hour post-earnings interest in Vertiv as distraction, Brian reads the return to arguing about memory as the market getting it right. scored the largest pass outside memory at +24.5 points, on the argument that a licensed running reactor is the one scarce electricity asset a factory cannot reproduce — but he says explicitly he has not run his own valuation work on it and that it is not a recommendation from him. passed by 2.2 points on a 20% gross margin, which he reads as a long order book with little pricing power.
He is unusually direct about what cuts against his own two picks. Every name in the screen sits near the top of its own valuation history — Vertiv at the 94th percentile, nVent the 96th, Micron the 98th — and on Micron he says: "At six times earnings, the market is not calling Micron cheap. Instead, it seems to be calling at the very top." Memory margins last peaked at 59% in 2018, and five years later Micron was losing money on every dollar it sold. The named breaker is a state-backed Chinese memory maker that listed in Shanghai in July, raising about $8.6bn, and which analysts he cites expect to be starting roughly nine wafers for every ten Micron starts by year end. fell 7% the day its lobbying fight with became public and 9% then 10% after the Shanghai listing, so the market has already rehearsed the trade. And on he flags a missing instrument rather than a green light: 18 months of listing history means his valuation-percentile guardrail does not exist for it.
4Copper and gold now out-earn iron ore at Australia's two largest miners
Wickenden's second point is compositional, and it has already happened rather than being a forecast. Copper and gold now account for more than 60% of Australian materials-sector revenue, ahead of iron ore, and at both and copper profit now exceeds iron ore profit. The materials sector rose roughly 47% over the past year while the ASX 200 barely moved — so the index-level drag is financials.
His diagnosis there is concentration. Australia's large banks lost diversification after the financial crisis and concentrated on domestic mortgage lending, which is now under pressure from three rate rises, budget changes and falling house prices; global banks are earning record profits from diversified capital-markets and investment books. He names 's valuation as the specific problem at index level. The ASX 300 is up about 4% this year against 10–20% for the US, while MSCI All World is up about 14% — roughly 60–65% of which is the US.
This is a sponsored segment and the disclosure matters more than usual. Wickenden works for BetaShares, which sponsors both the podcast series and the festival plugged in the same episode, and the outperformance figure he cites — the firm's Australian value ETF beating the ASX 200 by more than 12% last year — is his own firm's product. His own co-host attributed that outperformance to the fund not holding , which Wickenden agreed helped. Treat it as a product claim, not an independent one.
What the sources recommended
Three single-name buys and one avoid came out of four videos, and every one of them carries a caveat from the person making it. Two of the buys are the same trade.
and are Brian's two hardest passes on a gross-margin screen, and he is explicit that his ladder will let him start a position in either. He is equally explicit about what is wrong with each: Micron at the 98th percentile of its own valuation history and, in his words, priced as if the market is calling the top; SanDisk with no valuation history at all, which removes a guardrail he normally relies on rather than clearing one. comes from a BetaShares strategist naming a BetaShares fund on a BetaShares-sponsored show, which is a disclosure, not a disqualification, but it is the whole context of the call.
The one avoid is , on the specific finding that gross margin fell 2.2 points while revenue grew 46% — demand without pricing power. , and all came up favourably in the same screen but none was put forward as a call; on Constellation he said outright that he had not done the valuation work and was not recommending it.
Two names were re-examined rather than newly called. Robert Hagstrom, on Excess Returns, said his firm sold its software holdings in 2024 on the view that they could not compete against AI, and that on and others "the news of your death was premature. Maybe there's something there." That is a stated reconsideration, not a purchase.
The inflation print lands four days before the Fed meets
August core CPI publishes at 8:30am on 11 September, and the FOMC decides on 16 September. That sequence is the test of the first claim: a hike and a hold are priced within three points of each other, and one inflation number sits between now and the decision.
US markets are shut on 7 September for Labor Day. The earnings that follow are consumer and enterprise reads rather than macro ones: , , , , and on 9 September, then and on 10 September — the two that will be read for whether enterprise AI spending shows up in someone's revenue rather than in someone else's capex line.
