1Producer prices jumped 1.3% in September, and the market priced an October pause anyway
The producer price index for final demand rose to 159.78 in September, from 157.66 in August. That is a monthly gain of about 1.3%, according to FRED. It landed one day after a hot consumer print: September CPI was up 3.7% on the year, with core prices up 0.3% on the month. Apollo's chief economist, quoted by Stocktwits, called it "the definition of stagflation."
The market read it the other way. Kalshi's contract on an October quarter-point hike fell from 24¢ to 15¢, on about $274k of 24-hour volume and $1.3m of open interest. Polymarket's October Fed market moved the same way: "no change" rose 13 points to 84.5%, on $977k of volume. Long Treasuries rallied, with up 1.4% on Thursday. Small caps rose too, with up 1.3% against 0.3% for .
The mechanism was a flight to safety rather than a softer inflation read. CNBC reported Treasury yields sliding as investors fled the tech rout into bonds. On Excess Returns, Jim Paulsen gave a reason that pause could last. He counts six and a half to seven of eight economic forces as restrictive, the highest quartile since 1970. He also tracks Walmart's stock against upmarket retailers, a signal he says has led every Fed turn in this cycle. That signal peaked on 15 August and has fallen since, which he reads as a Fed that pauses before it hikes again. He expects inflation to fall even with oil stuck near $100, because a flat price adds nothing month on month.
Two caveats travel with the pause. December still prices a hike: Kalshi has a quarter-point move at 75¢, though open interest is thinner at $149k. And Mark Zandi of Moody's Analytics, on Prof G Markets, still described October as close to a coin toss. That view is now well behind the 15% the market prices. The move delays the hike rather than cancelling it. Next month's inflation print decides which side was right.
2The banks arranging AI's chip loans are selling the risk rather than holding it
Felix Nicholas Brinn, on Felix & Friends, laid out a $5 billion loan that Goldman Sachs and JPMorgan arranged for Volta, a company buying 36,000 Nvidia GPUs. He says $3.7 billion of it was offered on a "best efforts" basis. That means the banks earn the fee and look for other buyers, with no guarantee they find them. The rate he cites is 11%, against about 8% for ordinary junk debt. He also quotes a Goldman trading-desk note telling clients that investors "continued to de-risk from AI capex", sent the same week its analysts published buy ratings. His account comes from a former banker who sells training courses, and the deal terms are his reading of press reports, so treat the specifics as his.
Paul Kedrosky of SK Ventures, on Prof G Markets, applied the same logic to the most indebted AI cloud firms. CoreWeave carried about $52 billion of debt and lease obligations at the end of June, up from roughly $15 billion a year earlier. Asked whether it avoids default, he said no, and that the best case is a restructuring. His mechanism has two halves. Its 2032 bonds trade near 80 cents, so refinancing gets dearer. Meanwhile, Epoch AI estimates token prices fall about thirteenfold a year, so revenue has to double every 90 days to stand still. He put Oracle in the same group, and expects a correction by the first half of next year.
The market has not gone that far. Polymarket prices a CoreWeave bankruptcy filing by the end of 2027 at 15.5%. That book is thin, at about $41k of liquidity. bounced 5.9% on Thursday after a 16% fall the day before. Bloomberg reported its founders have sold $3.3 billion of stock. The chip side is weaker: fell 1.9% and 1.8%, even after TSMC posted a record profit. So the stress is now in the lenders' behaviour and the bond prices, not yet in the equity of the strongest balance sheets.
3This year's gains sit in AI and inflation-hit sectors; the rest of the market is barely up
Paulsen split this year's S&P 500 into three groups. AI stocks are up about 35%. Energy and materials, which inflation lifts, are up about 32%. The remaining seven sectors, which he calls the "modest seven", are up about 2.5%. That is less than a money-market fund paid. Ross Givens, in a short, made the same point another way. The index is up 14%, but the typical stock is up 4% and a quarter of the index is down 11% or more. On The Compound, Ben Carlson put the top ten stocks at 40% of US market value, up from 25% in early 2020.
The sources disagree about what that concentration means. Carlson says it can persist for ten or twenty years, because the top ten also earn about 40% of the index's profits. Paulsen sees complacency instead. Buying tech has beaten the market in 80% of rolling months over two decades, and he warns that holders who are all overweight could cut it at the same moment. Even so, he expects only a 15% correction, with tech down more than 20%. His reason is the same split: the parts of the market that never rallied have little to give back.
4Oil's $100 odds jumped 22 points in a day, even as the IEA sped up its stock release
Polymarket's contract on WTI trading above $100 in October rose 21.5 points to 59.5%. Its 24-hour volume was about $29k, which is moderate. The odds of a drop below $90 fell 24 points to 47%. US crude futures were near $95.60, down 0.9% overnight, according to Yahoo Finance. The oil fund rose 2.4% on Thursday.
The supply response is real but small. Reuters reported the IEA will speed up its emergency release, including more than 20 million barrels of diesel. On the other side, Reuters reported fresh Houthi attacks on Saudi infrastructure and new US sanctions on Iran. Polymarket prices Hormuz traffic returning to normal by 31 December at 11.5%, on $775k of liquidity, down 1 point.
This sits directly against Paulsen's case in claim 1. He needs oil to stop rising for inflation to roll over. The prediction market now gives better-than-even odds that it tests $100 within two weeks. If it does, the October pause rests on the tech sell-off alone.
What the sources recommended
Only three single-name buys came out of seven videos, all from one source, and all three are long-horizon picks. Ross Givens named GE Vernova , Eaton and Cameco as the names he would own "for the next 10 years". His case is that AI needs power the grid cannot yet supply. GE Vernova has gas-turbine agreements signed into 2031. Eaton cites a 307-gigawatt backlog of US data-centre projects. He conceded is "not cheap" at about $1,000, and it closed Thursday near $927, down 1.6%. Cameco is the one he calls mispriced. Uranium's long-term contract price hit $96.50 a pound in September, the highest Cameco has published since 1996. The stock sits in the mid-$80s, more than a third below its $135 high, and closed Thursday at $83.27. The pitch ran beside a paid membership offer, so weigh it as a sales video as well as a view.
There were no explicit sell calls. Givens, in a separate short, described buying January put options on the Nasdaq 100 fund as portfolio insurance through year-end. That is a hedge, not a call to sell. Kedrosky's view of was a forecast of restructuring, not a trade.
Next week tests whether the bond market or the oil market is right about inflation
Friday's US session opens with regional bank results before the bell, including Truist, M&T, Citizens and Regions, plus Travelers. They follow Bank of America's 5% fall on a $1.2 billion fine. The bigger test is oil: a WTI print above $100 would undercut the disinflation case behind the October pause. Kedrosky named one more dated item to watch. Anthropic ran a private investor day this week, and any leaked customer-concentration figures would test his claim that eight customers drive most of its revenue. The Fed's October decision lands in the last week of the month.
