1The Treasury is buying its own long bonds into a market that prices a rise, not a cut
The 30-year Treasury yield closed at 5.23% on 24 August, against 5.27% three sessions earlier and its highest sustained level since 2007. Treasury Secretary Scott Bessent's answer is to buy the paper himself. As Michael Batnick and Ben Carlson lay out on Animal Spirits, the Treasury has doubled the size of its own debt-buyback operations from $2bn to $4bn, starting 9 September, after a first attempt pushed yields down for a day and gave it all back the next. Felix Prinn, on Felix & Friends, puts the firepower behind it at roughly $1tn sitting in the Treasury General Account — about 250 times one operation.
The rate market is moving the other way. Polymarket's September FOMC contract, the deepest book checked today at $53.8m of volume and $2.7m of liquidity, prices no change at 64.5%, down two points on the day, and a 25bp increase at 34.5%, up one. A cut of any size sits at 1.5% combined. The separate "Fed rate hike in 2026" contract prices 57.5%, up a point, though on a much thinner $97k book. The European Central Bank is reportedly set for a September hike of its own, per Reuters sources.
Batnick's objection is that it will not work: "I don't think there should be government treasury political interference with the bond market." Carlson's is that the price is not broken in the first place. Take the 30-year and subtract the effective fed funds rate — 5.23% against 3.63% — and the 1.6-point spread is close to its median going back to the late 1970s. Rates were abnormal for fifteen years; they are not abnormal now. His prescription is unwelcome and specific: end the war in Iran, drop the tariffs, and the yield falls without anyone buying anything.
Carlson makes the same argument for the debt itself, with numbers. Federal interest payments as a share of total federal spending sit at their long-run average of about 15%. Morgan Housel's figure has total debt, government plus private, flat as a share of GDP over twenty years — the government borrowed so households could deleverage. Colin Roche puts public debt at 9% of total financial assets, against 5.5% in 2005. Batnick's one asterisk is Charlie Munger's: the liabilities are always fine, it is the assets you have to worry about. A 30-40% bear market re-prices that denominator in months.
The two dates are close enough to watch together. The buyback scales up on 9 September; the FOMC decides on 16 September. A programme designed to lower the long end starts a week before a meeting the market thinks may raise the short one.
2Long duration has the best payoff table on the board, and only a recession collects it
The most useful thing on the tape today is a table, not an argument. Batnick ran F/M Investments' rate-shock tool across the Treasury curve for a twelve-month total return. On the 10-year, a 100bp fall in yields returns about 12%; a 100bp rise costs about 2%. On the 30-year — the volatile end — a 100bp fall returns over 21% and a 100bp rise costs 8%. "I feel like bonds can't hurt you anymore," Batnick says, and Carlson's version is that at some point people will kick themselves for not owning any at these yields. CNBC reported the day before that large bets on a bond rally now dominate the options market, which is the flow saying the same thing.
Keep the caveat attached, because the same episode supplies it. Batnick's own read is that rates are structurally higher, and Carlson quotes Conor Sen on fifteen years of an aging-demographics case for low rates that turned out wrong. AI capex is a two-sided force here: it feeds inflation, and it floods the bond market with corporate supply after a decade when demand exceeded it. So the pair are recommending duration into a rate regime they think has moved up.
The tension resolves into a question rather than an answer, and they ask it plainly: under what scenario do long rates fall outside a recession? The scenario they name is Mark Zuckerberg or Alphabet tapping out on capex — inflation falls, yields drop hard, and stocks fall 25% on the way. That is what makes the payoff table honest. It is not a free option on lower yields; it is a hedge that pays when the equity book does not.
Ben Carlson's own framing puts the size of the thing in perspective. Jesse Livermore's figure: the entire market value of Treasuries maturing in ten years or more is $4.3tn. Apple alone is worth $4.6tn. One stock outweighs the long end that the whole argument is about.
Note also that the same channel argued the opposite two days ago. On 24 August, Nick Colas on The Compound and Friends decomposed the long yield into flat inflation expectations and a real-rate breakout and told listeners to keep duration under five years until the economy weakens. Same house, two shows, forty-eight hours, opposite conclusions. Nothing settles it but the rate path.
3The ceasefire market is fading the sanctions escalation the metals trade is pricing
Prinn's case is that the dollar is being turned into a weapon and hard assets are the trade. The facts underneath it are real. Bessent launched what he called Operation Economic Outcast, an "economic D-Day": roughly 60 companies, people and ships sanctioned over Iran, any country continuing to help Iran removed from the dollar system, and five named target sectors — digital assets, technology, shipping, aviation and gold. Prinn's point about that last one is fair: a government naming gold as a pressure point is telling you gold matters.
The flows corroborate him. Prinn cites roughly $22bn of gold futures bought in three weeks, the most in over a decade, gold trading near $4,600 and up about 15% off its low, Goldman Sachs telling clients $4,900 by year-end on what it calls a mechanical price amplifier — dealers hedging a flood of call options into the price — and one desk buying three-month calls on $90 silver. He cites a World Gold Council report saying every central-bank reserve manager surveyed is buying gold except Russia, which is selling. Animal Spirits saw the same week from the other side: bitcoin up 20%, gold up 5%, silver up 7%, and spot crypto ETF flows waking up after months of apathy — all of it, Batnick says, after Bessent spoke.
Batnick's caution is the one to keep. Why did bitcoin not care about the 30-year at 5.0% and care at 5.2%? Why is $40tn of government debt the number and not $39tn? "Sometimes the stories the market believes are the only thing that matters" — which is Keynes's beauty contest, and it works right up until the story changes.
Two markets are already disagreeing. Polymarket's US-Iran ceasefire contract, on $3.96m of volume, firmed across every horizon in twenty-four hours: through 31 August 96.9%, up 2.4 points; through 15 September 91.5%, up 8; through 30 September 83.5%, up 12; through 31 October 74.5%, up 9. The escalation is being priced as de-escalation. Reuters Breakingviews reached the same verdict in a headline — the US "D-Day" on Iran barely made it off the beach — and European shares gained on 25 August on softer sanctions and falling oil. Meanwhile gold's own contract cut the odds of a $4,700 print in August from 68% to 52.6% in a day, though on only $14.6k of 24-hour volume, which is thin enough to treat as a whisper rather than a verdict.
The Reuters file is a reminder that the war itself has not moved. Six months in it has settled into what Reuters calls energy trench warfare: Gulf traffic through the Strait of Hormuz still running below its ten-day average, Qatari gas stranded for six months, British energy price caps up 4%, Dubai airport traffic down more than 30% in the first half. Iran and Oman have reportedly reached an agreement on Hormuz revenue shares. A stalemate that neither escalates nor ends is the outcome the ceasefire prices point at, and it is the one that does least for the debasement thesis.
4Four of the five most-discussed stocks are the same trade, and the largest reports tonight
Brian, on Business with Brian, did an unusual piece of work: he watched 141 stock videos across 21 channels over eleven days and wrote every named ticker into a spreadsheet. Five companies came up more than any others — Alphabet , Nvidia , Micron , CoreWeave and Uber . His finding is the count underneath it. Of the 141 videos, 76 touched chips, the clouds they run in, or the models on top. More than half of eleven days of commentary, from channels with the whole market to choose from, came back to the same handful of names.
Four of his five are that one trade. The fifth, Uber, sits outside it until you ask what the overhang is — driverless cars, which run on the same silicon. His conclusion is the line of the day: a watch list built from videos like these is not five companies, it is "one bet wearing five different names."
Steve Eisman, quoted in the same video, had already acted on it: "I sold my Google. I wanted to reduce my exposure to AI. What scares me is that it's all one trade, so it better succeed."
The disagreements inside the five are worth more than the agreement. On Alphabet, Brian counted five creators splitting three ways — two buying, one passing because it is the most expensive hyperscaler at a $354 fair value, Eisman selling, and one who simply had a pre-set exit level and refused to move it when Berkshire Hathaway nearly doubled its stake. Alphabet trades at $341 today, which has quietly put it below that pass-side fair value since the video was recorded at $362. On Nvidia, the split is between a $300 equity fair value and the credit market: Prinn's observation that the cost of insuring Nvidia's debt has roughly doubled since late May, against customer borrowing he sizes at half a trillion dollars — nearly twice Nvidia's annual revenue — to buy chips that may be obsolete in five years, with the lending tracing back to pension funds. Brian's own fair value is about $290 against $210 today, and his floor at $148 exists specifically because of the credit point.
On CoreWeave the argument is whether the demand is real. One side points to an 8-K under which Nvidia agrees to rent back capacity CoreWeave cannot let to anyone else through 2032 — a seller underwriting its own demand, which is exactly what Lucent did to its customers in 1999 before the borrower went bankrupt and the revenue turned out to be its own money. The other side points at $130bn already under contract and a cost advantage over the hyperscalers of more than 40%. Brian refused to price it at all: seventeen months of public trading history is not enough to measure a company against itself. CoreWeave trades at $88, below the $113 fair value the bull side quoted at $104.
Carlson's guess at how this ends is not a dot-com crash. It is 1968-70 — the end of the Go-Go Years, a 36% index bear market that rarely gets its own chapter, underneath which the thirty most-crowded glamour names fell about 80%. He points out 2022 was already that shape: the Nasdaq 100 down 35%, Meta down 70%, Nvidia down 66%, the whole disruption complex down 80% or more. A run-of-the-mill index decline with a wipeout inside it.
What the sources recommended
Ten single-name calls came out of five videos, and the split between them is unusually clean: everything fresh and unhedged was in copper, and everything in the crowded trade came with a caveat from the person making it.
Ross Givens made the day's only unqualified new buys, both in copper, both after his prior week's calls worked — Freeport-McMoRan from $66 to $79 in five days, Southern Copper up 15-18%, on which he now suggests peeling back some profit while keeping a $250-300 target. The new names are Ivanhoe Mines , a $12.9bn Canadian producer trading over the counter around $9 with an initial target of $12-13, where he names the risk himself as African geopolitical exposure; and Ivanhoe Electric , a $1.9bn pre-production developer around the Santa Cruz project in Arizona, which he calls the more speculative of the two and which moves 5.5% on an average day. He bought both on camera with stops at swing lows. His second video follows Nancy Pelosi's disclosed late-July purchases into Bloom Energy — 15,000 shares plus 200 deep in-the-money June 2027 calls, roughly $7m leveraged — and Intel . He hedges it explicitly — "I'm not saying run out, buy Bloom Energy right now just because Pelosi did" — while noting both trade near or below her entry, which is the pitch. Only half of that holds today: Intel at $88 is genuinely below the roughly $92 he quotes, but Bloom at $216 sits at the top of the $185-215 range he gives for her first purchase and well above the roughly $175 he cites for her second.
Brian's four calls all carry their own brakes. Uber is the deepest band on his ladder, at $80 against his $109 valuation and an independent $120 — the only name where two people who never saw each other's work landed within $10 — with the driverless overhang named as the thing that breaks it. Nvidia sits in his buy band on valuation, but he is not adding because it is already over 12% of his portfolio. Micron he keeps buying on schedule and will not add a dollar to until it climbs back over its 50-day average at $961; it trades at $942 today. Alphabet he leans positive on as a half-size dollar-cost average only. He gave CoreWeave no rung at all and said standing aside is not the same as holding.
Two things sources did not recommend, despite spending most of their airtime on them. Prinn made no ticker call: gold and silver are the whole thesis and he named no vehicle, adding that a pure metals portfolio "is a lunatic's portfolio". And Animal Spirits' bond argument names no bond fund. Neither is a call, and neither is logged as one.
The one-trade thesis gets a number after the close
Nvidia reports after today's close against a consensus of about $93.6bn in revenue and $2.13 in adjusted earnings per share. Two Polymarket contracts price the components, both far too thin to lean on — data-centre revenue above $85bn at 87%, above $90bn at just 29%, on $438 of 24-hour volume, and adjusted gross margin in the 74-76% band at 92.5% on $260. Salesforce , CrowdStrike , Synopsys , Veeva , HP , Okta and Nutanix report alongside it, into a session where software fell broadly and Intuit dropped around 11% despite beating on adjusted profit guidance. Memory reversed hard as well — SanDisk , Micron and SK Hynix all giving back the prior day's surge, which is the same tape that has Micron under its 50-day line.
Tomorrow brings Marvell , Dollar General , Best Buy , Ulta , Workday , Autodesk and Affirm . The macro calendar has already printed: core PCE rose 0.25% in July to 3.3% year on year, against core CPI at 2.5% — the two inflation gauges are 0.9 points apart and pointing different ways, which is precisely the ambiguity a September hike is being priced into. The Fed's own next decision lands on 16 September, a week after the buyback scales up.
If four of the five most-discussed names really are one trade, tonight is when a large part of it is marked.
