The Nasdaq, the stock index heaviest in technology, fell 4.5% this week, its worst week since June. CoreWeave, which borrows to build data centers and rents out the chips inside them, fell 12% on Friday and 25% on the week. The VIX, the market's gauge of how much turbulence traders expect, closed at 19.99, up 35% over five sessions (under 20 counts as calm). And the interest rate the US government pays to borrow for ten years fell by 0.146 percentage points to 5.097%, its biggest weekly fall since mid-February, before the Iran war began (WSJ).

Those last two facts belong together. For three weeks the story was a jump in interest rates pushing the government's borrowing cost to its highest in 26 years. This week the fear changed shape: investors worried that those rates would choke the AI investment that has been carrying the economy, and money ran back into government bonds, which is where it goes when people are scared. The protection worked. The thing it was protecting against got worse.

What changed since Thursday is how far the selling spread. The New York Times puts CoreWeave's debts at $72bn across 51 data centers, with the market pricing a better-than-60% chance it fails to pay them, and the damage is no longer confined to one company: Oracle's bonds have fallen, the cost of insuring SoftBank's debt against default is rising, and Barclays says the extra interest investors demand to hold risky AI debt has risen by more than 50% in six months. Oracle is also struggling to get state permits for the power its New Mexico data center, Project Jupiter, needs (NYT). It is global, too. China's Star50 index of technology shares is down nearly 40% this half-year, and the chipmakers MetaX and Moore Threads have roughly halved, in a market where shares are 'the default source' of AI funding (Bloomberg).

The mechanism is the one Ben Inker of the fund manager GMO describes: 'You've got to bribe the lenders that much more to take more data center risk.' The newer AI cloud companies, the so-called neoclouds, borrow against their chips and customer contracts, and once their bonds pay 15% the next round of building has to be paid for by selling shares, just as Anthropic and OpenAI prepare to sell shares by the bucketload. Look at the split among the companies we track: Applied Digital and IREN are down 36% and 42% over 20 days, while Microsoft is 2.7% off its high. The market is not yet pricing a bust. It is pricing who has to borrow to stay in the game. That is more tinder on the pile, not a spark, but it is the kind of pile that turns a small fire into a large one.

Anthropic is offering Blackstone an interest rate of 10% to anchor the debt on its latest chip purchase. The package, called Project North Star, is $60bn and would pay for about 1.5 gigawatts of Google's tensor processing units, the chips Google designs for AI. Blackstone's funds would put in $9bn, about half of the $18bn slice that ranks last for repayment, with the rest to be sold to other investors over the coming months (FT).

The price is the news. Four months ago Anthropic sold $4.4bn of similar debt, Project Big Sky, at 8.5%. A company with a comparable risk profile borrowing in the public bond market pays about 7% (the ICE BofA BB index). So the same borrower, using the same kind of structure, is paying 1.5 percentage points more than in June and about three points over the public market, and more again once the discounts and fees on a new deal are counted.

The structure is a lease. Anthropic does not own the chips outright; it commits to five years of lease payments, and the lenders are secured on those payments. The $42bn that is not junior is not described in the report. What has to stay true is that Anthropic's revenue grows fast enough to cover the leases, and the company told investors this week that third-quarter revenue of $16.8bn grew less than 50% from June, a slowdown the market noticed.

Blackstone's pitch to its own investors says a lot: a rare chance to earn a high interest rate on debt expected to be graded safe enough for the most cautious investors, a grade called investment grade. That is the bet. Buy at the price of shaky debt, hold a bond that gets upgraded. Anthropic's current grade, as the FT reported yesterday, is BB+ from KBRA, which is still short of that line. If the upgrade comes, Blackstone's funds, and the insurance companies whose money sits behind many of them, have done a fine trade. If it does not, a private fund manager whose own shares are 20% off their high is holding $9bn of five-year AI lease paper at a price the public market would not pay. Nothing has blown up; a price has been set, and it is higher. That stacks the pile rather than lighting it.

Sound Inpatient Physicians is close to a $1bn refinancing with Monarch Alternative Capital and PGIM at 6 percentage points over the benchmark rate (S+600), after failing in June to find enough buyers in the syndicated loan market, where banks arrange a loan and sell it on to funds. Barclays, which led that attempt, had planned to try again with revised terms; it dropped the plan after bond yields surged. The private deal pushes the debt's due date out by five years and is expected to close next week (Bloomberg).

This is the migration we keep writing about, in one transaction. The syndicated market is made up of loan funds, including the bundles of loans sold in slices known as CLOs, which price their holdings daily and can say no. When they said no, the borrower turned to lenders who already owned it. Monarch bought the company's debt at distressed prices in a 2024 restructuring, when it and PGIM put in more than $100m of new money and cut about $200m of debt. Now they are refinancing their own position. And PGIM manages money for an insurer, which means a loan the loan market rejected ends up on an insurer's balance sheet, the list of what it owns and owes, with no daily price attached.

As a loan, this may be fine. Profits before interest, tax and depreciation have reportedly doubled since 2024 to more than $250m, and six points over the benchmark is a real price, not a favor. The lenders who bought cheaply are being paid to extend. The question is not whether Sound can pay. It is what the syndicated market's refusal tells you about the queue behind it: borrowers who cannot raise money in a public market at today's rates do not disappear, they get absorbed by lenders with a reason to keep them going, and the price that would have told you so is the one that was never set. That makes the pile harder to see, which is a different thing from making it bigger, and the harder it is to see the less warning anyone gets.

The US government paid $1.05tn in net interest in the fiscal year to September, up more than 8% from $970bn, according to Treasury Department data released Friday. The Peterson Foundation reckons that is 3.2% of everything the economy produces, above the record set in 1991. The average rate paid on the government's tradable debt was 3.52% at the end of September, more than double the 1.47% of five years ago, and the going rate on that debt today is around 5% (the Bloomberg Treasury index), so every bond that comes due is replaced by one that costs more (Bloomberg).

That is the snowball. What stops it rolling is growth. Gennadiy Goldberg of TD points out that the economy grew 6.3% year on year in the second quarter, before adjusting for inflation, which is faster than the government's borrowing rate, so the debt still shrinks relative to the economy's ability to carry it. 'A fiscal apocalypse is not upon us just yet.' Louise Sheiner of Brookings calls it 'a slow eating away'.

The place to watch is the shortest debt, where the $1.99tn deficit is actually being financed. The Treasury has sold nearly $100bn of extra bills, IOUs that come due within a year, in the first half of October, and Barclays expects more than $320bn over the next six weeks. This week's auction of eight-week bills drew the weakest demand on record, at 4.05%, the highest rate in over a year, and bills have become cheaper relative to what traders expect overnight cash to cost, because money-market funds are taking in less new money just as the supply of bills speeds up (Bloomberg).

So far the strain has not reached the short-term cash that keeps trading desks running day to day: SOFR, the rate on cash borrowed overnight against bonds put up as security, has sat at or below what the Federal Reserve pays banks on their reserves all month, which means cash is still plentiful. A bill market that cheapens while overnight borrowing stays calm is a market absorbing supply, not a market breaking. But both draw on the same pool of cash, and Barclays' own strategist describes money markets as 'caught between two very different narratives'. If overnight rates start to follow bills, the Federal Reserve's decision on its balance sheet in two weeks stops being academic. For now this is the slow kind of risk, the pile growing rather than a spark.

Strategy, the company whose business is selling its own shares to buy bitcoin, closed Friday down 9.6%. Bitcoin was down 0.1%. A gap like that means the market is pricing something other than the coin, and the something is MSCI, the firm whose indexes decide what the world's tracking funds must hold.

On Thursday MSCI adopted rules making 'non-operating companies', meaning investment funds, the funds that lend to mid-sized private companies, and anything whose finances make it look like one, ineligible for its main global indexes from the close on November 30. Strategy and Japan's Metaplanet are on the list of likely deletions; SharpLink and several other companies built to hold crypto are on a watchlist. The final list comes on November 11 (Globe and Mail).

The mechanism is simple and brutal. Funds that track MSCI's US and world indexes are required to hold Strategy's shares, and will be required to sell them on a known date. Strategy itself warned last year that being thrown out would mean $2.8bn of its stock being sold. Against a market value of $64.3bn that is not large. But the company's whole model depends on selling shares for more than the bitcoin behind them is worth and using the proceeds to buy more bitcoin. That premium exists because there are buyers who have no choice. Take them away and the premium shrinks, which is what a 9.6% fall against a flat coin looks like. The shares are already down about 50% over the past year.

Two things we cannot tell from the report. Whether the watchlist companies (SharpLink fell 7.5% and BitMine 7.3% on Friday) get caught in November or later. And whether Strategy, if its premium disappears, becomes a seller of bitcoin to pay the debt and preferred stock it issued to buy it. The source is a single wire report, so treat the detail with care; the index decision itself is well established. This is a spark with a date written on it, which is rarer than it sounds, and it moves ignition rather than fragility.

La Trobe Financial has sold its $31.2m first-ranking loan on a Bathla Group apartment project in Marsden Park, Sydney, to a private credit fund run by 360 Capital, and got back every dollar of principal, interest and late-payment charges. Bathla, a large developer, collapsed earlier this year; work has stopped on its sites. La Trobe had disclosed $38.1m of exposure in August (ABC).

On its face this is a price tested and held: a first-mortgage loan on a failed developer changed hands at full value. Look at who bought it. 360 Capital's listed mortgage trust already held an $8.9m second mortgage on the same building, ranking behind La Trobe. It has put in a further $5.2m and now controls both the first and second mortgages. The trust said the step 'was not taken lightly, but is very important to protect the value of the trust's subordinated position'.

So the senior lender got out whole because the junior lender could not afford to let the senior one foreclose. That is not a market price. It is the price at which the lender with the most to lose from a bad outcome chose to buy control of the outcome. The building helps: 72 apartments nearing completion, half of them pre-sold for $25.8m. Winky Tan of Morningstar notes that first-mortgage holders are generally in a strong position while the lenders ranked behind them 'risk losing most, if not all'. Which is the point. The junior lender here has avoided that fate by doubling down, and whether it has saved its $8.9m or thrown $36m after it depends on selling the other 36 apartments.

Australian private credit, lending by funds rather than banks, lent heavily to housing developers at valuations nobody tested. This week it entered the clean-up phase, and the first clean-up tells you how the valuations get defended. Nothing here makes a crash more likely on its own; we are finding out how bad it already was.

Investors who insure CoreWeave's debt against the company failing to pay pushed the price of that insurance to a record on Thursday, according to ICE Data Services figures reported by Bloomberg (the five-year contract rose as much as 74.5 basis points to 1,194.5). At that price, protecting $10m of bonds costs about $1.2m a year, and the market is saying there is a better-than-60% chance the company fails to pay by 2031. CoreWeave is the company that borrowed most, and most visibly, to build the AI boom's data centers, and the open market now rates it a coin flip. Its bonds that pay 9% and come due in 2031 slipped to 82.75 cents on the dollar, which works out to a 14.5% interest rate for anyone buying today; its 9.75% bonds of the same year rose 2.25 cents to 87.75. The shares went the other way: after Wednesday's 15.5% fall they closed up about 6% after Jensen Huang, who runs Nvidia, said Nvidia 'didn't need to backstop' the company and that its cost of capital 'is actually decreasing'.

Two things are worth slowing down on. First, what set the bonds off was not bad numbers but a plan: the company's talk of raising money from life insurers against debt secured on its chips and graded safe by the rating firms. Existing bondholders heard that as new lenders arriving ahead of them in the queue, with first claim on the chips if things go wrong. That is our inference, but it is the only reading that explains bonds falling on news of more money coming in. Second, Huang's remark removes a comfort that skeptics of the circular financing had assumed was there: the supplier says it will not stand behind the customer's debt. Bill Zox of Franklin Templeton put the loop neatly: 'Debt investors are freaked out about CoreWeave's cost of equity and equity investors are freaked out about its cost of debt.'

The losses have already landed somewhere. Goldman's prime brokerage, the desk that lends to hedge funds, says global long-short funds, which bet on some shares rising and others falling, lost money every day this week, 2.9% to Wednesday. Asia-focused stock pickers are down 6.8% in the first half of October and have given back more than half of the gains that had them up over 40% at the late-June peak. For scale: from Covid to the end of 2025, Goldman's three broad fund strategies never all fell more than 0.8 percentage points (80bp) on the same day. It has now happened three times this year, the latest on Wednesday. Thursday's bounce was in the stock. The insurance price did not bounce.

Money-market funds have pulled $7bn from a set of lending vehicles run by Guggenheim Treasury Services, which borrow short-term from those funds and lend the cash on to global investment banks to finance their trading. Nothing about the banks or the loans has changed; what changed is that the sponsor's name turned up in a federal investigation. That is how a run begins, and the money here is bank funding that happens to flow through something that is not a bank.

The vehicles sell commercial paper, short-term IOUs that come due in days or months, to money-market funds. At the end of September they had $45.8bn of it outstanding, down from $53.1bn at the end of June, a 14% fall, and data from DTCC, the firm that clears the trades, show the number kept falling through last week. The biggest competitor, Nearwater Capital, slipped about 1% to $86.1bn over the same quarter. One shrinks by a seventh; the other barely moves.

The Wall Street Journal's reporting is that customers are backing away because of the federal probe into Mark Walter, the Dodgers owner whose business group includes Guggenheim, and whose insurers the Justice Department is investigating for possibly financing other parts of the group without proper disclosure. In theory that should not matter. The people buying the paper are exposed to the banks that borrow the money, not to Guggenheim, which arranges the deals and takes a small cut. Guggenheim's spokesman says the business 'remains strong' and that third-quarter moves in interest rates hit several participants. JPMorgan's own data on issuers show a 12% decline for Guggenheim.

Why we care about $7bn. This kind of funding lasts only as long as investors are willing to lend it again when it comes due, and here it is being withdrawn not because the security or the borrowers changed but because the sponsor is in a headline. That is what a run looks like in its first inch: money funds, whose job is never having to explain a holding, quietly let the paper come due and do not buy more. The article does not say how the banks replaced the money; at this size they can. The mechanism is the point. Bank trading desks now lean on vehicles outside the banks, whose stability depends on reputations the banks do not control. Nothing was added to the pile this week; this was a test of a connection we already knew was there, and it did not break.

Jakarta's main stock index fell nearly 12% on Friday, its biggest one-day drop since January 1998, and trading had to be halted for 30 minutes. The trigger was not a bank failing or a bad number. It was MSCI, a private company in New York whose indexes decide which countries global funds have to hold, saying it is consulting on whether Indonesia should keep its emerging-market status. The rupiah neared a record low and bonds fell. Indonesian shares were already the world's worst-performing market this year.

What MSCI is proposing is unusual: not the usual one-step demotion to frontier status but a double downgrade to 'standalone', the bucket that holds Palestine and Lebanon, on grounds of murky ownership structures, suspected coordinated trading and weak enforcement. The decision is due in June unless there is 'clear evidence of material improvements'. Indonesia has announced reforms, disclosure of who owns what and higher minimums for the share of a company that must be freely tradable, and MSCI's response is that big investors want to see them enforced.

The mechanism is money that does not think. A fund that tracks the emerging-markets index holds Indonesia because the index says so. A standalone market is in no index, so every such fund becomes a forced seller on a known date. The market did not wait for June. 'The damage to foreign confidence starts right now,' as BCA Sekuritas's Felix Darmawan put it. The same afternoon MSCI froze index changes for Turkey 'immediately and until further notice' over coordinated trading in thinly traded stocks, after funds that had concentrated bets there collapsed into selling to meet withdrawals, and it adopted a rule that could push Michael Saylor's Strategy out of its indexes from November 30. Three triggers for capital to leave, in one day, all pulled by a private company in New York.

Put it next to the week: Korea down 8% on Thursday, enough to call it a bear market, India throttling its own currency market to hold the rupee, and now Indonesia. The dollar index, which tracks the dollar against a basket of other currencies, is down on the week and US government bonds have rallied. The stress is not in the dollar; it is in everyone who needs dollars. That is ignition, not fragility: nothing new has been borrowed, but the sparks are landing closer together.

Argentina has not sold a Bonar, a dollar bond sold in its own market, for two months, the longest pause since it started the program in February, because the interest rate buyers demanded rose by more than 5 percentage points (500 basis points) to above 12%. Across the world, India has made it so expensive for its companies to protect themselves against the rupee falling that most have stopped. These are two faces of the same thing: dollars are drying up at the edge of the system, and the substitutes each country has lined up are not markets.

A first sale of Argentine bonds abroad, once hoped for, is off the table at current global rates. The plan had been almost $1bn more of local sales this year and $5bn in 2027. Instead Economy Minister Luis Caputo said on Wednesday that the government could draw on its currency swap with China, an arrangement under which China lends Argentina money, if the local market does not reopen. About $25bn of payments in foreign currency fall due next year, the front of a schedule that gets heavier through 2030. 'When authorities had the chance to smooth next year's financing schedule, they chose not to,' said AllianceBernstein's Armando Armenta.

India's defense of the rupee is producing a different kind of breakage. The rupee has had its best week in four years, up 1.9%, but the rule that produced it, which forces lenders to set aside 20% of the rupee value of any currency derivative trade over $2m, has made protection so expensive that the three-month contract to buy dollars with rupees closed at 6.54% on Thursday, its highest since May 2016, and trading in it has collapsed. It is now cheaper for an Indian company to borrow rupees and swap them into dollars than to borrow dollars outright. Local bond interest rates are at a three-year high. 'We will have to swallow poison for a long-term gain,' said RBL Bank's head of treasury.

These are two ends of the same wire. With the interest rate the US government pays to borrow for ten years at 5.12%, the price of dollars is set in Washington, and the borrowers at the edge pay it first: Argentina by not borrowing, India by rationing the tools its companies use to protect themselves. The China swap is Argentina's lender of last resort. India's is its own reserves. Neither is a market. This is the slow kind of risk, tinder rather than a spark, but it is growing rather than merely being discovered.

Charles Schwab, the largest broker for ordinary American investors, said on Thursday that loans to its customers against their holdings reached roughly $188bn in the third quarter, nearly double a year earlier. About $61bn of that is tied to one product, 'tax-aware long-short' strategies, which chief executive Rick Wurster said had grown 'exceptionally quickly' this year. Brokers lend against portfolios all the time; doubling in a year, with a third of it coming from one strategy that cannot exist without borrowing, is how debt piles up with nobody choosing to take more risk. Morgan Stanley, Bank of America and Goldman Sachs all reported double-digit growth in the same kind of loans in the year to September.

The product works like this. A manager (AQR and Quantinno are the big names) builds a portfolio that tracks an index, then borrows against it to add two 'extensions', one betting on a set of stocks rising and one betting on another set falling. The falling bets throw off losses the client can set against gains elsewhere on their tax return, which is the selling point; the rising bets are meant to add return. Money in these strategies has gone from just over $1bn five years ago to nearly $190bn, on Brent Sullivan's estimate. The borrowing is not incidental to the trade. It is the trade: without borrowed money there are no extensions and no losses to harvest.

That makes this a different animal from a retail investor borrowing to buy Nvidia. The books are closer to balanced, with the rising and falling bets roughly offsetting (we are inferring that from the structure, since Schwab's disclosure does not break down which way the money leans), and the clients are wealthy people solving a tax problem. But $61bn of brokerage loans now depends on a strategy whose pitch assumes markets keep producing gains to shelter, and whose falling bets can be forced into losses by exactly the kind of sudden reversal Asia's hedge funds just lived through. Schwab's own shares fell 2.7% on the day and are 10% down over the month, the worst of the brokers we track. Borrowing against shares has sat near record levels all year. This is one reason why, and it is not the one anyone expected. It is tinder, not a spark: the pile got taller and nobody noticed a tax product was doing the stacking.

Partners Group, a Swiss firm that buys companies using borrowed money, is refusing to let hedge funds onto the list of approved lenders for Emeria, the French property manager it owns, and is applying similar limits at Ammega and Breitling. Between them the three companies owe about €6bn due in 2028 that Partners is trying to refinance. A loan that can only be sold to buyers the owner approves is a loan whose price nobody is allowed to test, and that is how the prices on private loans stay unchallenged until the moment they cannot. Hedge funds and specialists in troubled debt can still buy the companies' bonds, which the lists do not cover, but the loan terms are among the most restrictive in the market, according to people who have seen the documents.

The mechanism. A loan to a company that already carries heavy debt can only be sold on to buyers its owner has approved, which in normal times is a formality. In a restructuring it is a weapon. Distressed funds make money by buying loans cheaply, forming a bloc and either forcing a deal that favors lenders or taking the keys to the company. Shut them out and the owner negotiates only with the original lenders, who are often less willing to fight. The cost is stated in Bloomberg's first sentence: the pool of buyers shrinks 'as their loan prices fall'. A loan that cannot be sold to the people who want to buy it is carried at a price nobody can check.

This sits beside a second number from Thursday. Funds that lend to companies in place of banks raised a record $129bn in the first nine months of the year, up 35%, while the amount they actually lent in the third quarter fell 28% to just under $50bn, and $16.4bn of their loans were refinanced away into the market where banks arrange loans, against $2.6bn coming the other way. Record money chasing fewer deals pushes lenders to 'concede on spreads, leverage or documentation', as Raymond James's Sunaina Sinha Haldea put it. In other words, to accept less interest, allow more borrowing, or write weaker contracts. Documentation is exactly what Partners is leaning on now. The lists being written into this year's loans, in a buyer's market, will decide who gets to show up at the restructurings of 2028. This does not make a crash more likely this week. It makes a risk we already count harder to measure, which is its own kind of tinder.

CoreWeave sold a set of bonds due in 2032 four months ago at full price, 100 cents on the dollar. On Wednesday they traded below 78. At that price the bonds pay anyone who buys them an interest rate above 15%, and the cost of insuring the company's debt against it failing to pay now implies a better-than-even chance that happens within five years. The shares fell as much as 17%, the most since February, and at least seven bonds backing data centers tied to CoreWeave have traded lower this week (Bloomberg). A bond losing more than a fifth of its value in four months is not a bad quarter. It is the market saying it doubts the money comes back.

The money financing CoreWeave sits in layers. At the top are loans backed by its chips and by contracts with named customers; the $8.5bn loan tied to Meta won a top safety grade from the rating firms in March, the grade that lets insurers and pension funds hold it. Below that are bonds backed by nothing in particular, like the 2032s, and below those the shareholders. The plan, as co-founder Brannin McBee described it to the FT this week, was to keep winning high safety grades for the chip-backed debt and eventually for the company itself, which would unlock "all the money in the world that sits in the insurance tranche." In other words, move the financing to life insurers, who hold a loan until it is repaid and never have to show a daily price for it. That is this site's whole thesis, stated by someone on the inside.

The public market got there first. The bonds backed by nothing in particular are the one layer where a price is set every day, and that price now says the shareholder cushion protecting the chip-backed lenders is in doubt. Gil Luria of D.A. Davidson, who has called CoreWeave "the subprime tranche" of AI, meaning the riskiest slice of the boom's borrowing, put the arithmetic plainly: "They've run out of the ability to borrow even at 9%," and a company borrowing at 9% to earn 5% does not fix that by getting bigger.

What we cannot see: how the insurers and funds holding the chip-backed loans value them. Those loans do not trade. No borrower has been reported to break a promise in its loan contract and no lender has been reported to withdraw a credit line. But 15% is not a rate at which a company renting out AI computing power can borrow again to pay off what is coming due, and the rest of the sector closed Wednesday as if it knew: Nebius down 10%, IREN 12%, Core Scientific 13%, Applied Digital 9%, Oracle 9%. This is ignition, not new tinder. The debt was already there; what moved is how close the match is.

South Korea's main stock index, the Kospi, fell nearly 8% on Thursday, taking it 21% below its September high, past the line traders call a bear market. The exchange stopped trading for 20 minutes as the selling deepened. SK Hynix and Samsung Electronics, the two memory-chip giants, each lost more than 9%; Japan's Nikkei 225 dropped almost 5%, with Ibiden and Kioxia among the worst hit; and the main US index of chipmakers has now shed $1.6 trillion of value (Bloomberg). This is the third day this week Korea has been hit, and the hardest. Trading halts in a major market are rare; they exist for days like this one.

Why Korea, when the borrower in trouble is in New Jersey? Because Hynix and Samsung are the two most easily sold stocks on earth whose profits depend directly on how much the giant cloud companies spend building things. When a global fund decides at 9am to cut its AI exposure, it cannot sell a loan backed by CoreWeave's chips or a stake in a private data center; nobody is standing there to buy. It can sell memory makers in size, in an hour, with buyers waiting on the other side. "Korea is becoming the shock absorber for the global AI selloff," said Vantage Global Prime's Hebe Chen. The IMF, which lends to countries in trouble, made the same point in advance in its stability report this week, naming the stock markets of rich Asian countries as the most sensitive to whether AI pays off (Enterprise).

The chain runs like this: CoreWeave's bonds break, US inflation comes in hot, bets on the Federal Reserve raising rates rise, and the bet that gets closed is the one that can be closed. South Korea's won and Taiwan's dollar are leading the currency losses for the same reason. Nothing we can see suggests Hynix or Samsung have a money problem of their own; they are being sold because they are sellable.

The US figures in our tables are too narrow to show the scale. The Nasdaq 100 closed Wednesday down 1.9% and 3.3% off its record; the S&P 500 is 1.7% off. Korea has just lost a fifth of its value in a month. That lopsidedness is the point: the losses are landing on the suppliers, in the market where selling is easiest, before they land on the lenders. For the odds of a crash this is ignition moving closer, not fragility growing. Forced sellers have appeared; the question is whether they run out of easy things to sell.

On the same Wednesday that CoreWeave's bonds fell into distress, Anthropic told prospective investors at its San Francisco headquarters that its revenue in the third quarter reached $16.8bn, that it made a profit in the quarter by its own adjusted measure, that more cash is coming in than going out, and that after the direct cost of serving customers it keeps 85% of its revenue, with more than 11% left after all running costs. It plans to start formally selling shares to the public as soon as the week of November 9 and to be trading before Thanksgiving (Bloomberg).

Take the numbers at face value for a moment and they answer the question the whole building boom rests on: is anyone actually paying for this? One model company, at least, is being paid handsomely. Last week OpenAI told investors its yearly run rate of revenue was $20bn below what it had signaled; this week its rival says it is generating cash.

But look at where the money is kept. Keeping 85% of revenue after direct costs means the computing power that serves the model is a small fraction of what customers pay. That computing power is what CoreWeave and its peers borrowed to build. The model company keeps the margin; the infrastructure company keeps the debt, and at Wednesday's prices earns, by Gil Luria's estimate, 5% on money it borrows at 9% or more. If Anthropic's figures are right, the AI boom is not without economics. They are concentrated at the top of the chain, and the bottom of the chain is financed as if it owned them.

The caveats are real. These are adjusted figures, told to prospective buyers of the stock by people familiar with the presentation, not audited accounts. The company does not say what "adjusted" leaves out, and its commitments to buy computing power are the obvious candidate. Nor does it say how fast the $16.8bn is growing. The formal document it must file before selling shares will. Until then the day's most interesting disagreement stands: the customer reports a profit, and the supplier's bonds trade at 78. On its own this nudges the odds of a crash neither way. It tells us where the money is, which is something we did not know this morning.

The interest rate the US government pays to borrow for ten years fell by 0.1 percentage points to 5.09% in Asian trading on Thursday, its lowest in three weeks, and the rate on every other length of government debt fell at least 0.07 points. The Bloomberg dollar index, which tracks the dollar against a basket of other currencies, dropped with it, and government bonds in Japan, Australia and New Zealand rose. A gauge of what US government bonds have returned is now up almost 0.7% in October after a 2.2% loss in September, which if it holds would be the best October for those bonds since 2014 (Bloomberg). Our own tables have the ten-year rate at 5.05% this morning, from 5.32% on Tuesday.

This happened on the day US consumer prices were reported to have risen 0.6% in September, the fastest since April. Austan Goolsbee of the Federal Reserve's Chicago branch, the central bank that sets American interest rates, called the figure "discouraging" on air, pointing to inflation in services above 5%, the core rate above 4% for the month, and "very little that we're hearing that sounds like downturn." He named the AI data-center buildout as a possible source of overheating (Yahoo Finance). Normally a hot inflation number pushes government borrowing costs up, not down. That is why the direction matters more than the size.

For three weeks the defining feature of this selloff was that stocks and bonds fell together. That is the condition in which a second bet placed to limit the loss on the first stops working and forced sellers appear; it is why Chile's pension funds are closing out an interest-rate bet that has lost $7.7bn. On Thursday the old relationship came back: shares fell, government debt was bought. "The market has finally moved back to the old flight to quality," said ITC Markets' Kit Lowe. In other words, when investors got scared they ran to government bonds again, which is what they used to do.

What the rally is betting on is less comfortable than it looks. A 12% rout in chip stocks and a Korean bear market are being read as something that cools inflation, or as something the Federal Reserve cannot raise rates into. Goolsbee's logic runs the other way: the building boom is itself the overheating, so a central bank that raises rates to cool it raises the cost of the debt that funds it, CoreWeave's 15%. The IMF counts 14 of 33 major central banks tightening since the second quarter. The rush into safe assets is real and it is the best thing in today's data. It makes a crash less likely, and we say so plainly. It is also a bet that the Federal Reserve will blink before the borrowers do.

Goldman Sachs and UBS are among the global banks becoming more selective about the specialized financing they give to Australian fund managers who lend to companies, according to people familiar with the matter. New lending is going to the largest, top-tier managers; smaller firms and those heavily exposed to Australian real estate are being steered away from. Existing commitments are largely being kept (Bloomberg).

The plumbing matters here. A fund that lends to companies in place of a bank does not only lend; it borrows too. Banks provide the loans at the fund level that let it put money into deals before investors' cash arrives and, the important part, pay back investors who want out without selling loans. That is how a fund holding property debt nobody could sell quickly can tell its investors they may withdraw every quarter. Take the bank's money away and the fund has two options: freeze withdrawals, or sell loans into a market with no buyers, which forces it to admit a price.

Australia is a small test case with the right ingredients. About half of its A$200bn ($140bn) market for this kind of lending is tied to real estate. Bathla Group, a Sydney developer, went bust in August owing A$3.4bn, largely to these lending funds, and the regulator has been halting fund offerings. "The froth is coming out of the market," says Corval Avenue's Belinda Bible.

The lenders are global, and the behavior is the tell. The New York branch of the Federal Reserve last week counted $1.5tn of US bank lending to firms that are not banks; this is that pipe being squeezed at the small end, by the same institutions whose examiners have been asking about it. The publicly traded firms that run these funds keep falling with no American event to explain it: Blackstone down 3.5% on Wednesday and 21% off its high, Blue Owl down 5.4%, Prospect Capital 26% down in a month. Meanwhile the last reading we have of the extra interest shaky companies pay compared with the government, 3.16 percentage points on October 13, has barely moved. The shares of the lenders are being repriced. Their loans are not. This stacks tinder rather than lighting it; the loans still have not been tested at a real price, and the day they are is the one to watch.

Cash Today Inc., a consumer lender in Edmonton founded in 2005, reported more than 44,000 outstanding loans worth about C$229m at the end of last year. When the court-appointed receiver, MNP, inspected its offices last month it found records of 256 active loans worth less than C$900,000. Its creditors and the holders of its IOUs, about 270 of them, have potential claims of more than C$613m, a figure MNP says may be overstated by at least C$190m, which still leaves a gap no honest balance sheet, the list of what a company owns and owes, can close (Bloomberg).

The structure is the smallest possible version of the one this site writes about every day. Investors bought IOUs, called debentures, from a high-cost consumer lender outside the banking system. Some of those agreements required the money to sit in separate accounts until it was used to buy portfolios of loans. Several investors have shown the receiver records of their money going instead to an account belonging to a related company. Cash Today told the receiver it had no bank accounts in its own name, and the bank records agree. MNP cannot yet establish whether investors' money was used as the agreements required, or what the company's assets and debts actually are. One of the two men who controlled it died in the week of September 21; the other resigned as a director on October 1.

C$613m is not a number that threatens the system. What it shows is a mechanism: the thing pledged to the investors was a reported count of loans, and nobody checked the loans existed until the money stopped arriving. Canada's lightly regulated lending industry is where this one surfaced. The same question, who has actually seen the assets behind the income, is the one being asked this week of a far larger set of private lenders: in Australia, in New Jersey, and in the Federal Reserve's visits to the Wall Street banks that lend to funds. This changes the odds of a crash hardly at all. We include it because it is the pattern in miniature.