The Ledger War
The machinery of wartime finance keeps reappearing in peacetime clothing, funded by the conscripted despair of the young, with three refinancing treadmills racing each other to the first seizure
A short note before the essay.
This is the 1st issue of the Curious Mind Research Letter. One deep dive like this every month, with a shorter letter of signals in between. It launches properly this autumn. Today’s issue is free, in full.
If you want in from the start, reply “founding” to this email, the founding list hears first, on terms that won’t be repeated.
The Letter ships as a typeset PDF, made for printing and forwarding.
Published 7 August 2026. Gold $4,377. Market levels as of the 6 August close.
Last week the new Fed chairman held his first rate decision, leaned dovish, and the long bond broke out; within two days his own Treasury was telling banks it might sell dollars. This year, gold overtook U.S. Treasuries as the world’s largest reserve asset. This piece is about the war those two facts belong to.
“Bond markets have taken out more governments than howitzers.”
— Scott Bessent, U.S. Treasury Secretary
1. The Undeclared War
A week ago Wednesday, on July 29, Kevin Warsh chaired his first full rate decision as head of the Federal Reserve. He arrived with six weeks of hawk talk behind him, traders pricing two hikes on the strength of it, Bloomberg celebrating him as the man unraveling what the desks call the debasement trade. Then he held rates, and the press conference was dovish.
The long end of the Treasury curve rose. By Friday the thirty-year had broken out of its range, and veteran technicians were publishing upside targets towards seven percent. Within forty-eight hours of the decision, the U.S. Treasury was checking rates on the yen and telling banks it might intervene alongside Tokyo, which is a polite way of saying it might sell its own dollars.
The long end rose through six weeks of the hardest talk Warsh could give, then rose through the softest delivery he could manage. Given price action since, the market has stopped grading the words and started grading the maths. The rest of this piece is my attempt to understand the maths.
Start with a renter I’ll call Maya, a composite of half the twenty-somethings I know. She’s twenty-three, does marketing for a skincare brand, pays $2,350 for the small bedroom in a Bushwick three-share, and holds $31,000 across a brokerage app and two crypto wallets against about $60 in savings. She can tell you the ticker of everything she owns and cannot tell you what a bond is, and she isn’t embarrassed about that. This spring a renter just like her told the Wall Street Journal that money feels safer in the stock market than in a house, and sixty percent of her generation, per a Harris poll, believes the only way to build wealth now is through what their grandparents would have called gambling. Crypto, meme stocks, prediction markets, sports bets.
Maya has read the terrain right. Homes at eight times her income. Entry-level postings down 35 percent. She went where the upside still lives.
She can’t see it from where she stands, but her adaptation has been noticed, measured, and built into the structure of the global financial system. Maya is now the marginal lender to three of the largest debtors in the history of money. She’s the war bond drive, minus the poster.
I keep a list of things that shouldn’t happen at the same time. This year the list got long. Bond yields rising on deflationary news. Stock markets booming in the countries with the worst demographics. Gold crashing thirty percent while central banks bought record amounts of it and stopped answering survey questions about where they keep it. An AI complex financed, when you read the contracts, like 2006 subprime. Teenagers levering the Seoul stock market to all-time-high volatility. The last week of July joined the list all at once.
For a while I treated these as separate stories. They read better as sector reports from one front. The major powers have slid into a conflict fought on balance sheets instead of beaches, and the financial machinery of total war is reassembling itself around them, piece by piece, whether or not anyone in charge decided to rebuild it.
The last century called it financial repression.
This century calls it innovation.
2. Learning From 1942
The Federal Reserve was designed, above all, not to do the thing it now does.
When Paul Warburg sold Americans on a central bank before 1913, the pitch was built around trade credit, stamps on bills for actual crates on actual ships. Lending to the government was considered a grubby old-world habit, beneath a proper central bank. The whole design was a machine that would never finance the state.
Thirty years later it was the state’s financing arm and little else. From 1942 to 1951 the Fed pegged Treasury yields and bought whatever it took to hold the peg. Inflation ran hot and quietly melted the war debt. Households were sold the other side of the trade as patriotism, with posters.
Then the 1951 Accord dismantled the machine, and the textbooks filed wartime finance under history.
The difference between an emergency and a regime is a date. An emergency ends when somebody says it’s over. A regime just is.
Nobody voted to bring the old machine back. What came back on its own were the conditions that made it useful the first time — supply shocks instead of demand cycles, inflation that won’t hold still long enough to be forecast, stocks and bonds moving together instead of against each other.
Every piece of the 1940s machine is back, wearing new clothes.
The yield cap returns as an acronym.
The captive saver returns as plumbing.
The posters came back as phrases.
And duty got rebranded as opportunity.
None of the parts that came back has an exit clause, because none was ever presented as an entrance. A war that’s never declared can never formally end.
Underneath the machinery sits a law that doesn’t change. No government in history, offered the choice between visible austerity and invisible inflation, has picked the visible one, and no electorate has ever asked it to. Default is a decision. Debasement just happens to you. Everything that follows is that one preference, compounding.
3. The Preview Is Playing In Tokyo
The endgame doesn’t need imagining. Japan is the territory the war reached first.
For years the Bank of Japan capped its ten-year bond yield near zero and defended the cap with unlimited buying. That was yield curve control, the actual thing, not the acronym under discussion. In July 2023 the cap came off. Today the ten-year JGB trades just under three percent, a level that would have been unthinkable to a Tokyo bond desk five years ago, and the thirty-year sits at generational highs.
Under the old rules, that’s a catastrophe. A country with the worst demographics in the developed world, debt beyond anything the West carries, and its borrowing costs roughly tripling. The old rule said sell that country’s stocks and hide in its bonds.
Look closer at the debt, though. Tokyo’s budget deficit is narrower than Washington’s, and its interest bill, measured against the size of its economy, is lighter than Germany’s, or France’s, or Italy’s. The country everyone assumes is the most fiscally stretched of the group is carrying the lightest debt-service burden in it. The Bank of Japan spent two years letting inflation expectations drift above target and holding policy easy through it, and bond investors are pricing that in.
Instead the Nikkei rose seventy percent in twelve months, accelerating as yields climbed. Bonds down, stocks up, and the yen quietly dying against gold. This is how a country trades when investors have stopped trusting its money rather than its government’s ability to repay it: get out of the currency, get out of the bonds, get into anything real. Tokyo’s answer to a central bank falling behind the curve was a $2.3 trillion spending plan, treating a credibility problem with the one tool guaranteed to make it worse.
Korea is marching the same route, stocks up 110 percent in a year against Chinese-grade demographics. Germany’s DAX and the FTSE have joined, each rising as their bond markets sink. Five governments with terrible balance sheets all ramping defence spending at once, all their currencies range-bound against each other while all of them fall against gold, which is either a remarkable coincidence or something closer to an agreement: if we all debase together, none of our currencies falls against the others, and we run hot until somebody’s bond market breaks. Last week the agreement stopped being tacit. Japan intervened and the yen jumped 3 percent; a Japanese currency diplomat described American assistance as “beyond psychological support”; and the U.S. Treasury told banks it might intervene in the yen itself, which means selling its own dollars. The pact now operates in daylight.
Japan matters because it’s furthest down the road. It shows you the sequence. Cap the yields, lift the cap when you must, watch the bond market reprice, and watch the stock market inflate as citizens flee their own currency in the most respectable way available. The West talks as if following Japan were still a choice.
4. Three Treadmills, One Law
A few weeks back a Substack post made the rounds on OpenAI, and in it was an interesting sentence:
A company funded by its own appreciation is solvent not in proportion to how high the mark stands, but in proportion to how fast it’s still rising.
He wrote it about one company. It’s a general law of terminal-stage credit, and it currently governs three balance sheets, each one inside the next: the AI complex, the U.S. Treasury, and China.
Call them treadmills. On each, the debt can’t be repaid, only rolled, and rolling requires the next mark, the next auction, the next quarter to price above the last. The markup is the cash flow.
The AI treadmill is the smallest and fastest. Strip the jargon and it’s potentially just a credit-driven real estate cycle in a technology costume: debt-financed construction, take-or-pay leases, supply guaranteed to land after demand turns. The $2.1 trillion backlog the hyperscalers report as contracted revenue is, read as credit, a loan book, half of it owed by two cash-burning labs with no operating income. Arguably most informed lender in the system, Microsoft, saw the books from inside for years and shortened its duration in April while keeping the equity upside. The IPO everyone is waiting for is the refinancing of last resort, and it might be delayed because the price that clears the public market doesn’t retire the burn, and the price that retires the burn doesn’t clear the market.
The tremors are on the tape. The semiconductor index doubled in six months while the companies funding the boom fell through the first half, Microsoft down 27 percent, Oracle 22, Meta 18, all now borrowing to fund capex that has blown past 100 percent of operating cash flow. Meta floated selling “excess compute” and the stock jumped nine percent, which is a market cheering the first admission of overcapacity because the alternative reading is unbearable. And the first dedicated AI fund blowup arrived at the end of July, a former OpenAI researcher’s hedge fund unwinding after losses. Whether that was 1998, a warning the cycle survives for another eighteen months, or the spring of 2007, the first domino, is precisely the question.
The bulls will argue that July earnings proved their case. The datacentres themselves are wildly profitable, cloud revenue up 43 percent on a $364 billion base, margins widening, capex guided higher again. All true. The profits are today’s. The backlog is tomorrow’s, and half of tomorrow’s is promised by frontier labs whose product is commoditising with token prices falling into rising demand.
Meanwhile the price of the roll is rising even as the belt speeds up. Bond supply from the AI complex reached $270 billion by early July, nearly double all of last year. Meta’s newest data-centre paper cleared at 7.5 percent, half a point above an identical deal a year earlier, and both Meta and Google have gone cash-flow negative for the first time in their histories. The chipmaker at the top of the chain is in talks to guarantee $250 billion of financing for its own largest customer, a sum equal to 135 percent of its retained earnings. Vendor financing at the summit of a capex cycle is one of the oldest bells in finance. And Washington, which calls the buildout a matter of national security, has just signed an executive order opening federal land and military bases to data centres. When the state starts leasing its bases to the boom, the distance to a formal backstop is short.
The Treasury treadmill is the largest and best disguised. Over eight trillion dollars of Treasuries mature inside twelve months, which works out to refinancing roughly one Marshall Plan every week, indefinitely, and hoping somebody shows up each time. The national debt has been converted into commercial paper, and the sovereign lives, like any structured vehicle, at the mercy of its next roll.
China’s treadmill is the strangest, because the operator chose to run it uphill. Total debt just crossed 300 percent of GDP on BIS figures and is still rising a point a month, a pace at which China borrows an extra year of its entire economy every eight years. Meanwhile the money data shows the tightest conditions in Chinese history, loan growth at all-time lows, the 2024 stimulus cancelled and reversed. Xi picked deflation on purpose, exchange-rate stability over growth. China’s belt is held up by state banks and the record price of its own bonds, which is to say, by the one buyer who can’t sell.
What makes this one story instead of three is the coupling. AI now drives more than a quarter of U.S. GDP growth. Foreign money is pouring into U.S. stocks at record rates, mostly unhedged, mostly chasing AI. So if the smallest treadmill stumbles, GDP loses its engine, the growth scare hits a bond market that now sends yields up on bad news, the unhedged money runs and takes the dollar with it, and eight trillion of maturities meets its auction calendar in the middle of the storm. Every government’s crisis plan quietly assumes the other treadmills fail first. That can’t be true three times.
Two things would kill this thesis: A) Real AI productivity showing up in the aggregate numbers, which would let the debt be serviced instead of rolled or B) A genuine collapse in long yields during a growth scare, the old world reasserting itself. Watch those two dials. Everything else is commentary.
5. The Conscripts: Maya Funds All Three
Every war-finance system in history has faced the same problem. The state needs to borrow more than willing lenders will provide, so unwilling lenders must be found and their unwillingness dissolved.
The 1940s solved it with patriotism and payroll deduction. The 2020s found something more elegant. It located a generation that no longer believes in the system, and turned the disbelief itself into the funding mechanism.
The raw material: sixty percent of young Americans see speculation as the only path to wealth. Forty-two percent of Gen Z investors hold crypto against eleven percent with a retirement account. Average personal debt of $94,000. A bachelor’s degree whose real median wage has moved two thousand dollars in thirty-five years.
You could call it a symptom, the sad output of a broken opportunity structure. I’d propose instead that the evidence supports that the despair is an input.
Follow the pipes. The AI complex’s terminal capital source is the pool the index funds and the retail bid sit in; the delayed IPO is a plan to hand the burn rate to the public. The Treasury’s manufactured bid runs through stablecoins, which are disproportionately the rails of the crypto-native young, and through baby accounts that convert birth itself into an equity inflow. Korea’s melt-up runs on retail leverage, famously including teenagers. American margin debt is up 54 percent in a year to 4.5 percent of GDP, half again the dot-com peak.
In 1943 the citizen-lender was recruited through belief in the flag, the cause, the certainty of victory. In 2026 the citizen-lender is recruited through disbelief. It’s precisely because the young concluded that wages, savings, and houses are dead ends that they arrive, levered and all-in, at exactly the assets the treadmills need bid. A generation that still believed in pensions would never supply this bid. The despair had to come first.
Two complications. First, none of this requires a conspiracy, just convergent opportunism: a Treasury finding stablecoins lying there, an AI complex finding the retail pool lying there, platforms finding gambling psychology lying there. Systems don’t need architects to end up with architecture.
Second, the trade could work. If the devaluation thesis is right, stocks do rise in nominal terms, and the young speculator front-running debasement is positioned correctly. The cruelty is that the young are the shock absorber either way. They win only if the treadmills never seize, and they are, by construction, the last money in, the pool the refinancing chain reaches after every informed participant has shortened duration. Microsoft saw the books and chose to rent. What Microsoft declined will be offered to Maya’s brokerage app.
Maya stores her savings in a system engineered, at the pipe level, to deliver her capital to the sovereign’s paper at negative real rates. She’ll take the haircut. She’s been told she’s being empowered.
6. The Quartermasters Are Provisioning Not Investing
Two facts from the second quarter of 2026.
Gold had its worst quarter since 2013, down nearly thirty percent from about $5,600 to under $4,000. Miner sentiment hit two percent bulls. A Wall Street Journal front page announced faith in the haven was shaken.
That same quarter, the People’s Bank of China bought gold for the eighteenth straight month, its biggest add since 2024. Chinese imports hit a two-year high, up 76 percent on the year. A record 45 percent of central banks told the World Gold Council they plan to raise their own reserves. And somewhere in this stretch, mostly unremarked, gold overtook U.S. Treasuries as the largest component of global reserve assets.
Same chart, opposite behaviour. Western fast money sold a volatile commodity. The official sector of the non-aligned world bought like bargain hunters. I think the second group is provisioning, treating gold as materiel, the monetary equivalent of a strategic petroleum reserve, rather than as a portfolio position.
The weapon teaches its targets. The more a hegemon uses its financial infrastructure as one, the less control it keeps, because every shot fired shows every watching treasury that the opt-in system was never really opt-in. The demonstration was February 2022, when the West froze Russia’s reserves. If you ran a central bank outside the alliance, the lesson wasn’t subtle. Your reserves are yours until the issuer decides they aren’t.
Everything since is that lesson at scale. Central banks custodying gold at home, up from 35 to 49 percent since 2023. The share refusing to answer the custody question at all jumped from 8 percent to 20, and respondents don’t start refusing a routine survey question unless the behaviour it asks about has turned sensitive. Swiss gold exports to Saudi Arabia, keystone of the petrodollar, are up eight-fold since 2022, while Riyadh ships four times more crude to Beijing than to America.
Why gold and not somebody’s currency? Because there’s no qualified successor and both sides know it. The renminbi can’t inherit the role while Xi refuses everything inheritance requires, an open capital account, a bond market foreigners trust, and his own money data shows him choosing the opposite. What’s being built instead is commodity trade priced in yuan with net settlement in gold, so no counterparty ever has to hold Chinese paper overnight. Gold as the bridge between two ledgers that no longer trust each other, the one collateral both sides take because it’s nobody’s liability and nobody’s weapon. Demand for a bridge is a function of the width of the chasm.
The fundamental valuations guys will say that a fair-value model built on real rates and central-bank demand will tell you gold is fully priced, and the record says treat it as a diversifier prone to violent swings, it lost 45 percent in 1980-82 and fell in the 2020 and 2022 crashes too. But models built on market-implied rates inherit the distortions of the market being managed, the same way breakevens said two percent in 2021 while inflation compounded at four and a half. A fair-value model calibrated on the old regime measures the old war.
7. Seeing The Map & The Dominos
In a bloc built on simultaneous mutual refinancing, a bond crisis anywhere becomes a crisis everywhere, shortly after. One point of failure, distributed across many addresses. So the question is which address.
Korea is the acknowledged one, the froth capital of the coupled system. KOSPI volatility is above its 2008 record, circuit breakers tripped six times in two weeks, teenage leverage converted by single-stock ETFs into structural fragility. It’s where a seizure of the AI treadmill becomes visible fastest, the war’s forward listening post.
Germany is the political one. Two straight years of contraction, Volkswagen announcing a hundred thousand layoffs, and the AfD, first in national polls, on track for its first state in September on a platform of restoring Russian energy and exiting the sanctions regime. Read coldly, that’s a proposal to defect from the western ledger, and German mid-caps rallying ten percent into it suggests markets pricing the defection as stimulus.
And then France, which almost nobody is watching, though a few desks have begun quietly shorting French bonds against gilts and Spanish paper. It tops my list. Total debt at 324 percent of GDP, above China’s, unimproved since 2019 while every peer deleveraged. A spread against Germany that widened sharply in a month when yields generally fell. A hard catalyst on a fixed date, the April 2027 election, with the Rassemblement National at the gates. The strange part is who’s missing. The countries seemingly debasing together, the US, Japan, Germany, the UK, Korea, don’t include France. It’s raising defence spending into the worst debt position of any western power without membership in whatever tacit pact the others struck. It’s inside the war and outside the alliance. The Liz Truss precedent says the break comes where debt is deepest, growth weakest, politics most combustible, and hedges least in place. That describes France better than it describes America.
The calendar, from where I sit in early August. The first dial has already reported: hyperscaler earnings landed strong and capex was guided higher, so the belt sped up rather than slowed. Under the treadmill law that defers the reckoning rather than cancels it, and it shifts the watch to the labs, where the price war has already started. Warsh has decided, for now. He held, into a Philly Fed print above 41 for only the eleventh time in sixty years, and in the only two prior instances where the Fed declined to tighten into a print that hot, gold has exploded higher. He didn’t hold unanimously — three regional presidents dissented for a hike, and two more Fed voices have already said, in public, that they’ll vote to hike in September unless core inflation cools within weeks. The long end voted within days of the decision itself, and Warsh’s own Treasury answered by preparing to sell dollars. The question from here isn’t hike or hold. It’s whether the debasement resumes in size, through the bank-rule changes and stablecoin legislation now being pushed, before the long end gets away from them, or September delivers the hike three of his colleagues already want. The France-Germany spread through autumn. China’s monthly credit prints. Saxony-Anhalt in September. Gold’s $4,000 line through the seasonal window.
When one of those dials moves, the response won’t be austerity or default, and it won’t be honest. My bet is the completion of the machine, the cap, the captive saver, the inflation tax, announced as innovation, adopted by acclamation, paid for by the conscripts.
I can even sketch the scene. A Sunday night, because these things are always announced on Sunday nights, with the Asian open looming. A programme with a reassuring name, something with Stability or Freedom in it. A cap on long yields, described as temporary. A notification on Maya’s app about new account protections, framed as being for her benefit. The biggest relief rally in years. Op-eds calling it prudent by Wednesday. Japan has already shown the sequence. After that, two ledgers, gold as the bridge between them, and Maya’s generation wondering, eventually, what exactly it was they funded.
8. What Am I Doing
For what it’s worth, my own book: some gold, some AI for momentum, some energy to hedge the mess. No long bonds. Nothing shorted, because I believe this story and I’ve also watched people who believed it get carried out on stretchers for two years running. Holding both of those thoughts at once is, I think, the honest position.
Systems only look like systems later. Monetary regimes are always built this way, messily, denied while under construction, obvious only in the archive. The bet here is that we’re standing inside one of those construction sites now, and that in twenty years these notes will read less like market commentary than like an early draft of the archive. If the bet is wrong, the dials above will say so within a couple of quarters. I’d rather hear it from them than from the crowd.
Written 7 August 2026. Market levels as of the 6 August close. Errors are mine.
Thank you for reading No. 1.
Two requests. If this is the kind of work you want arriving every month, reply “founding” and you’ll hear before anyone else, at the founding rate. And if one person came to mind while you read this — the person you’d have argued with about it — send it to them.
Research letters travel by forwarding; they always have.
Ahmed
This is not investment advice. Curious Mind Research Letter is provided for informational purposes only and does not constitute financial, investment, legal or tax advice. Do your own research and consult a qualified professional before making investment decisions.
Love your mind, didn’t like ths post: too AI-ish. Needs a ruthless editing job.