1October hike odds fell from about 70% to near even in one day, while long yields kept climbing
Kalshi's contract on a 25bp hike at the 28 October Fed meeting last traded at 49 cents, down from 70 cents a day earlier. The bid sat at 44 cents. The hold contract jumped from 30 to 55 cents. This was not a thin move: the two contracts traded about 357,000 and 449,000 contracts in 24 hours, on more than 800,000 each of open interest. Polymarket's October market agrees, with no change at 56.5% (up 26 points) and a hike at 43.5%, on about $2.0m of 24-hour volume.
The trigger was a speech. Barron's and Seeking Alpha reported that New York Fed President John Williams said there is no rush and signalled he is open to a pause. Governor Michael Barr was reported the same evening as saying more hikes are likely needed. That report came from crypto-news outlets only, so treat it as unconfirmed. December still leans to a hike: Kalshi prices it at 73 cents, barely changed.
Long bonds did not rally with the hike odds. FRED shows the 10-year yield at 5.24% on 28 September, up from 5.17% on the 25th, and the 30-year at 5.56%. Yahoo Finance reported the 30-year at its highest level since 2002. The gap between 10-year and 2-year yields widened to 0.37 points on 29 September, from 0.32 the day before. Short yields fell relative to long ones, which is what fading hike odds should do. The Wall Street Journal reported the dollar at an eight-week high.
Two explanations for high long yields compete. The news framing is oil and inflation: Brent near $105–108 keeps price pressure alive. On Prof G Markets, Michael Green of Tier 1 Alpha rejected both the fiscal and the inflation story. He said US credit default swaps are tightening, inflation swaps are falling and the dollar is rising. None of that fits a market fleeing US debt. He called the sell-off a mechanical effect of passive bond indices, which steer new money away from bonds trading below par. He sees deep-discount long bonds as neglected, but warned the selling "can go quite a bit further." He called hiking into a war-driven oil shock "counterproductive."
The labour data softened at the margin. FRED shows August job openings at 7.08m, down from 7.34m in July. Kalshi prices September payrolls above 100,000 at 48 cents, up from 45, on about 17,400 contracts of open interest. It prices unemployment above 4.0% at 74 cents and above 4.1% at 40 cents. August printed 4.1%.
So what: the policy leg and the bond leg have split. Hike odds are now a coin flip, yet the long end did not fall. If Green is right, a pause will not rescue long bonds quickly. If the oil-inflation read is right, a pause would be a mistake the long end punishes. Friday's payrolls and the October inflation print are what settle the October meeting. Neither settles the long end.
2Oil markets price a Strait of Hormuz that stays shut even if the ceasefire holds
Ed Elson, on Prof G Markets, said Brent reached $108 on Monday before settling near $105. That followed President Trump's rejection of an Iranian proposal to reopen the Strait. Reuters reported Trump denying he had offered Iran sanctions relief, and oil gained on the denial. Reuters also reported the US will lend up to 40 million barrels from the Strategic Petroleum Reserve, the last batch of a global deal.
The prediction markets separate the fighting from the shipping. Polymarket prices the US-Iran ceasefire continuing through 31 October at 60.5%, up 2.5 points, on about $57,000 of 24-hour volume. It prices Hormuz traffic returning to normal by 31 December at only 19.5%, down 2 points, on $131,000. The market on average daily transits at the end of September puts 96% on zero to five ships, up 27 points in a day. Polymarket prices a senior US-Iran diplomatic meeting by 31 October at 30%.
So what: a ceasefire that holds and a strait that stays shut can both be true. The first stops the escalation premium building. Only the second returns barrels. The market is currently pricing the ceasefire without the barrels.
3Memory chip stocks are holding up while the average stock falls, and Micron reports tonight
Ross Givens, on his channel, pointed out that the S&P 500 sits about 2% below its high. The equal-weight version has fallen for five to six weeks. He blamed rising yields, which make stocks less attractive against a risk-free 5%. His own breadth gauges have turned weak, and he said he has cut his position sizes.
The one group he said is still strong is memory and storage, driven by AI demand. He bought Micron at about $1,045 and SanDisk at about $1,716. SanDisk carries a stop at $1,540, about 10% below entry, and a target near its prior high. He attached a caveat himself: if leaders with clean chart patterns fail here, "there's almost nothing worth buying right now." His Micron trade also runs straight into tonight's earnings.
Gil Luria of D.A. Davidson, also on Prof G Markets, read the same group the other way. He said the market is pricing Nvidia , Broadcom and Micron as if their cycle is almost over. So the two sources agree memory is where the money went, but disagree on whether the market believes it lasts. Micron's results settle part of that.
Finnhub's consensus has Micron reporting after the close today, with earnings of about $32.32 a share on $52.2bn of revenue. The sizing point came from a different conversation. On Excess Returns, Kris Abdelmessih cited a 2016 study in which 61 mostly finance-trained players got a coin that won 60% of the time. Some 28% went bust anyway, by betting too much of their bankroll.
So what: memory has become the market's test of whether AI spending still converts into earnings. A strong Micron print supports the breakout read. A weak one would support Luria's late-cycle read, while the average stock is already falling.
4Nvidia is buying back its own stock at its lowest earnings multiple in a decade
Nvidia added $150bn to its buyback, taking the total authorisation to $235bn. Elson said Nvidia expects to spend it within 16 months, and that it beats Apple 's $110bn record from 2024. The stock closed up 1.5%, he said, and is up about 21% this year against more than 75% for the Philadelphia semiconductor index. It trades at under 17 times forward earnings.
Luria said the buyback is simple arithmetic: at 17 times earnings, repurchases give shareholders a good return. He contrasted that with AMD at about 40 times and Intel at about 60. He called the market's two valuations "internally inconsistent." He named frontier model labs and cloud-leasing firms as the excess. He said CoreWeave borrows at 9% to earn about 1%. Separately, the hosts said Meta hired MongoDB's chief executive for a new enterprise AI unit, and MongoDB fell more than 20%.
Yahoo Finance reported chip stocks fell on AI-safety fears after an agent-security incident, while Nvidia rose. Nvidia launched software to safety-test AI agents, and OpenAI paused training of its most advanced models.
So what: Nvidia is betting its own cash that the market's growth fears are wrong. The buyback is a floor on demand for the shares, not proof of the earnings. Micron tonight is the nearest outside test of the same AI earnings stream.
What the sources recommended
Only two single-name buy calls came from three videos, both from Ross Givens, and both carry his own warning that conditions are poor. He bought Micron and SanDisk as memory-sector breakouts, with a stop about 10% below entry on SanDisk. He said he has cut his usual position size, and his Micron position is held through tonight's earnings. Michael Green called deep-discount long-dated bonds a buying opportunity, but named no specific bond or fund, and warned the sell-off can extend. No source recommended selling a named stock.
Tonight's Micron print and Friday's payrolls test both halves of the week's story
Micron reports after today's close, with Jabil, FactSet and Cal-Maine reporting before the open. Accenture reports before the open on 1 October, and Nike after the close. September payrolls land on Friday 2 October, with the market pricing a figure near 100,000. A strong payrolls number would push October hike odds back up from a coin flip. A Micron miss would test the one group still carrying the AI trade.
