1A September rate hike is now the priced outcome, and oil put it there
Polymarket's contract on the 16 September Federal Reserve decision now prices a 25 basis point increase at 64.5%, against 35.5% for no change. That is an eleven-point move in twenty-four hours, on $1.8m of volume in that leg alone and $114m across the event — deep enough to take seriously. A cut is priced at 0.45%. Twelve hours earlier the same question sat at 57% on Kalshi, a number Ed Elson reported on Prof G Markets before the latest repricing.
The chain that got it there is short. Brent settled above $100 and then ran another 6% after a fresh wave of tanker attacks. Rising energy costs feed straight into headline inflation, and central banks are already responding: the European Central Bank hiked on Thursday, with Reuters framing the decision as the Iran war adding to inflation angst. The US ten-year Treasury yield closed at 4.83% on Tuesday, its highest since 2023, and kept climbing — even though the Treasury was buying back up to $6bn of its own paper on Thursday, three times its usual size. Buybacks that fail to hold the long end down tell you the selling is coming from a repricing of inflation, not a lack of a bid.
Equities did what a higher discount rate implies. The S&P 500 fell 0.58%, the Nasdaq 0.65%, and the volatility index rose 8.4% to 17.84.
The consumer price index prints on Friday 11 September. It is the last inflation reading the committee sees before it decides.
2The oil move is a shipping-route problem, and the prediction markets are not reading it as escalation
What moved crude was not a lost oilfield. Shipping traffic through the Strait of Hormuz has been running in single digits, well below its ten-day average. Houthi forces advanced along the Yemeni coast and are closing on the Bab el-Mandeb Strait, the southern gate of the Red Sea, threatening Saudi exports on that route. The US military says it destroyed five Iranian oil carriers after attempted missile attacks. Reuters reporting puts roughly a third of Gulf crude unaccounted for despite so-called dark crossings, and the Energy Information Administration raised its price forecasts on the grounds that the war is draining global stockpiles. OPEC's own August output fell.
Barrels that exist but cannot sail price like barrels that do not exist. That distinction matters, because it means the premium unwinds on a route reopening rather than on new supply.
Here the market disagrees with the tape. Polymarket's contract on a US–Iran effective ceasefire beginning by 18 September rose 8.5 points to 70.5%, and the 11 September leg rose nine points to 60.5% — a de-escalation bid, placed on the same day the shooting intensified. Yet the far-dated legs of the separate Israel–Iran ceasefire contract went the other way: holding through 30 September stayed firm at 84%, while 30 November fell 6.5 points to 63% and 31 December fell six to 58.5%. Near-term calm, deteriorating tail. Both cannot be the whole story, and nothing in the day's news settles which one is right.
The contract that would actually reopen the route — an Iran–Oman agreement to manage Hormuz — sits at 26% by 30 September and 45.5% by 31 October, on thin volume of about $36,000 a day. Treat those odds as weak. A ceasefire stops the risk premium building; only a route agreement brings the tankers back.
3The standard bond index is built to lose money if rates rise
John Kerschner, global head of securitised products at Janus Henderson, and Michael Contopoulos, its head of multi-asset macro investing, spent an hour on Excess Returns taking apart the Bloomberg US Aggregate index. Their arithmetic is the whole argument: the index carries about six years of duration against a yield of roughly 5%. A 100 basis point rise in rates — a move that has already happened this year — costs about 6% in price and leaves the holder negative. Nearly half the index is Treasuries or Treasury-like paper, up from about 30% coming out of the financial crisis, so an investor is paying 30 to 60 basis points in fees for a lot of interest rate risk and not much yield. Over the past five years the index has been flat to slightly negative.
Contopoulos's case for higher rates is structural before it is cyclical. He dates the break to 2015: deglobalisation stopped importing disinflation and started importing the opposite, home-country manufacturing costs more, and productive capacity was never built. On top of that he puts a rerun of the 1960s guns-and-butter problem — record defence spending alongside stimulus still flowing from earlier legislation, with money supply and velocity accelerating. His conclusion is that with credit spreads at all-time tights, "the Fed is way too easy at the moment."
What they own instead is floating rate and short: AAA-rated collateralised loan obligations, which reset higher when the Fed hikes, and actively managed agency mortgages for the duration they do want. This is the part to read with the conflict in view — Kerschner manages both of the funds he describes, and the CLO product he calls "almost 31 billion" is his firm's own. The one idea with no such attachment is long-dated municipals, which Contopoulos flagged as a tactical opportunity and noted plainly that Janus Henderson has no product in. Their dislike is aimed squarely at investment grade corporates: not on credit quality, which they call fine, but because spreads leave no compensation for downgrade, default or illiquidity risk.
Contopoulos also offered a theme he called out of consensus: artificial intelligence is inflationary, not disinflationary, to the tune of 50 to 60 basis points. His four reasons are capital spending, electricity demand, labour scarcity — he cites a shortage of 300,000 electricians — and a wealth effect from rising AI-linked share prices. Measured productivity, he notes, has gone down rather than up.
The same view that makes them refuse the bond index makes them constructive on shares: profit cycles expanding globally, a Fed they do not consider tight, and what they call one of the better environments for international equity investing in some time.
4The capex money lands on the suppliers, and the spending base behind it is narrower than it looks
Two of the day's three videos arrived at the supplier side of a capex boom from opposite directions. Ross Givens made the case for , Regal Rexnord, on the argument that actuators — the motor, gearbox, screw and magnet package inside a robot joint — are 40% to 60% of the cost of building a humanoid robot, and that Regal Rexnord owns brands covering the whole joint. He puts booked humanoid-related business at more than $30m a year against a $200m tracked pipeline, and reads a new chief executive's unnamed reference on the 5 August call to "one of the largest US robotics companies" in Austin as a customer his suppliers are barred from naming. His point about the share price is that it fell for unrelated reasons: an earnings beat flattered by a $32m tariff refund, a soft quarter in the residential heating and cooling half, and index selling that took the whole industrial group down with GEV, ETN and CAT. The stock closed at $156.62 on Thursday, roughly a third off its high and below the $163 he quoted on tape. His caveat, in his own words, is that margins are still a problem and the housing-linked business needs help.
Contopoulos framed the same idea without a ticker: not the sprinkler company, he said, but the company making the ball bearings in the sprinkler systems, and he extended it to small and mid-cap value industrials over technology growth in both the US and Europe.
Against that sits the sharpest counter-fact of the day. Ara Karazian, lead economist at Ramp, told Prof G Markets that about 80% of enterprise revenue at OpenAI and Anthropic comes from 1% of businesses — a level of customer concentration, he said, unseen in any other software category Ramp tracks — and that those top spenders cut spending roughly 10% month over month, from $8,000 to $7,200 per employee. Token volume rose over the same period, which rules out a simple demand explanation; his read is a price war plus a migration from frontier models to cheaper standard and light tiers. He was careful to place himself on neither side, and offered the bullish reading too: a larger share of a growing pie, at lower margin and higher volume. The same concentration reaches further than the model vendors, because AI revenue at the large cloud businesses — MSFT, AMZN and GOOGL — leans heavily on those same two customers.
Kerschner put a number on the debt side of it: roughly $400bn of this year's $2tn in investment grade corporate issuance is financing data centres, a category that barely existed three or four years ago. Contopoulos calls that an equity valuation question rather than a credit calamity, with the caveat that downgrades would concentrate risk in the high yield market next year.
What the sources recommended
Three single-name buys came out of three videos, and every one of them carries a qualification from the person making it. There were no sell calls on a named security at all — the day's avoid-arguments were aimed at asset classes rather than tickers.
Ross Givens was the only outright stock pick: as a picks-and-shovels route into humanoid robotics, with a two-to-three year horizon, an analyst consensus target near $250 and his own $300 to $400 case. He is explicit that margins and the residential business are unresolved, and the pitch is interrupted twice by promotion of his own $5 subscription service.
The two fixed income buys, and , are Janus Henderson's own AAA CLO and agency mortgage funds, described by the manager who runs them. Neither ticker was said aloud on the recording; both are identified here from the descriptions given, and the conflict is the reason the caveat travels with them. Contopoulos's long-municipal idea and the multi-sector income fund Kerschner manages are named in the conversation but not by symbol, so neither is logged as a call.
The inflation print tests all of this on Friday
Consumer prices are released on Friday 11 September, five days before the Federal Reserve decides. A hot print confirms the 64.5% and validates both the oil chain and the case against duration. A soft one leaves a market that has moved eleven points in a day looking early.
Oracle and Adobe both reported after Thursday's close. ORCL had already fallen 5.3% to $153.08 in the regular session before the release; a thin after-hours print near $163 on barely a thousand shares hints at a positive reaction, but the volume is too light to lean on. ADBE closed at $248.73 and traded near $242 after hours on similarly light volume. Friday's earnings calendar is otherwise empty of large names; the next ones are General Mills on 15 September and FedEx and Lennar on 16 September, the day of the decision.
