Metrics Credit Partners, an Australian firm that manages about A$40bn of loans to companies, halted trading in three of its funds listed on the Australian stock exchange before trading began on Monday and cut the values it says those funds are worth. It did so because its auditor, KPMG, made "different decisions" about the assumptions and probability weightings behind those values. The measure that falls is net tangible assets, what each fund holds minus what it owes: down 12.16% at the Metrics Real Estate Multi-Strategy Fund, 10.08% at the Metrics Income Opportunities Trust and 1.99% at the Metrics Master Income Trust (Reuters). Funds revise their values all the time. An auditor knocking a tenth off one over a weekend is not that.

Read the middle number. A loan that has never been sold on a market does not have a price. It has a set of assumptions: will the borrower pay, when, and how much comes back if not. Each carries a probability the manager assigns. Change those weightings and a tenth of a fund disappears without a single borrower missing a payment. Nothing in the portfolio changed over the weekend. The person holding the pen did.

The most telling figure is what buyers already thought. The real-estate fund's units last changed hands at A$1.68 while the fund said they were worth A$2.53, a discount of 34%. Even against the new figure of A$2.22, the traded price is 24% lower. People who owned the listed units had been saying for months that the values were too high. The auditor has now agreed with them about a third of the way. Metrics says it lent nothing to Bathla, the property developer that collapsed last month owing A$3bn to some 40 lenders, and Australia's financial regulator warned the whole sector about unrealistic valuations only last week.

Hours later, on the other side of the world, staff at the SEC, the agency that polices American markets, put out what they called a "critical reminder" about valuing assets that do not trade, urging "particular care" for private lending funds because their loans have no ready quoted price. It contains no new rule. It does record that at midyear more than $14.5bn of investors' money was stuck in more than a dozen private lending funds that had stopped paying it back (Bloomberg).

Metrics is the visible version of this. Its funds are listed, so the gap between what the manager said and what buyers would pay sat on a screen for anyone to see. The same kind of assumptions sit in funds nobody can trade, on insurers' balance sheets (the lists of what they own and owe) and in retirement accounts, where no auditor has yet said no. For the meter this is the discovery kind: the world did not get more dangerous on Monday. Our measure of how much of the pile is overvalued got better.

Brent crude, the oil price most of the world quotes against, rose above $108 a barrel on Monday after Donald Trump rejected Iran's weekend proposal to reopen the Strait of Hormuz. The interest rate the US government pays to borrow for ten years rose to 5.23%, its highest since 2007, and closed at 5.24% by our own record of the day; the thirty-year rate finished at 5.56% and the two-year at 4.92%. The futures market, where traders bet on where rates are headed, now expects the Federal Reserve, America's central bank, to raise rates by a quarter point twice more by January (FT). Rates like these last existed before the 2008 crisis, and the reason for them this time is a war rather than a budget.

What is new since Friday is where it is happening. Britain's ten-year rate is 5.42%, near its highest since 2008. German government bonds at 3.65% mean the dearest German borrowing since 2011. France pays what it paid in 2008. Japan's two-year rate touched 1.98%, above anywhere it has traded since 1995. This is no longer a story about American deficits and the volume of debt Washington is selling, though those have not gone away. It is oil feeding expectations of inflation in every market at once.

The chain is simple and unpleasant. A war closes a strait; energy gets dearer; central banks that raised rates this month are expected to raise them again; and every bond that pays a fixed sum each year is worth less, because new ones pay more. Gold spells it out. Gold futures fell 3.1% to $4,189, a seven-week low, and silver fell 4.7%, in the middle of a war (WSJ). Gold is where people usually run when there is shooting. A metal that pays nothing cannot compete with a government bond paying 5.2%, and when the war is the reason for the 5.2%, the safe place and the dangerous one turn out to be the same bet.

Lenders to companies have begun to notice. The extra interest that shaky companies pay compared with the government stood at 2.93 percentage points on Friday, the latest reading we have, up 0.13 on the day and 0.33 on the month. Funds holding the bonds of safer companies, which trade like shares, fell another 0.8% on Monday. Shares themselves have not noticed: the S&P 500, the index of America's largest companies, is 1.2% off its record, and the VIX, the market's gauge of how much turbulence traders expect over the next month, is 15.8, where anything under 20 counts as calm. That gap, bonds at a nineteen-year extreme and stocks a percent from a record, is the thing to watch. One of them is wrong about the next six months. For the meter this is ignition, not fragility: nothing new was stacked, but the spark moved closer.

Wealthy individuals and the family offices that manage their money have started borrowing against their stakes in buyout funds, because the funds have stopped paying out, the FT reports. Firms that buy companies using borrowed money have returned far less cash to their investors over the past four years than in the decade before, and family offices had 20% of their money in those funds and in private lending in 2025, up from 16% in 2019, according to UBS. The whole market for this kind of loan, which goes by the name net asset value lending, or NAV lending, is about $150bn, with an average deal of $150mn (FT). That is small. What matters is what the loans rest on.

It works like this. You own a stake in a fund. The fund's manager says the stake is worth, say, $100mn. A bank or a private lending firm lends you a fraction of that against the stake. You get cash without selling the stake to another investor, who would demand a discount and force you to book a loss. Goldman's Ben Williams puts the client attitude plainly: "it's inefficient to have an asset that you can't sweat". In other words, clients want to get some use out of a holding they cannot easily sell.

Look at what secures the loan. Not the cash the fund pays out, not a price on a stock exchange, but the manager's own estimate of value, the same category of number an auditor in Sydney rewrote by 10% this morning. The borrower is choosing not to test that estimate against a real buyer. The lender is accepting it as the thing it keeps if the borrower cannot pay. Both are betting that a value nobody has sold at is the value it would sell at.

It is also a third layer of debt on the same companies. The company the fund owns carries the loans used to buy it. The fund itself may have borrowed against its holdings. Now the fund's investor borrows against the fund. Each layer is sized against the one below, and none of them trades. The market is small, and Ares's Scott Rosen calls the part aimed at individuals early. But a product that exists to avoid admitting a loss is a product that stores losses, and the news here is that it has reached the rich-individual end of the pipe. For the meter this is fragility: genuinely new borrowed money on the pile, though at $150bn not yet much of it.

Private loans now make up about 45% of the roughly $42bn of debt investments held by Delaware Life, the insurer Mark Walter bought 13 years ago, up from 9% in 2014, according to an analysis by the Wall Street Journal. Two of its three largest investments are loans to companies Walter is affiliated with; the third has business ties to his firm, Guggenheim Partners (WSJ). Going from a tenth to nearly half in private loans is the whole industry's direction of travel. Lending nearly half of that to the owner's own businesses is not.

This is the cleanest single picture yet of the shift Crash Lab tracks. A life insurer takes premiums today and promises to pay its customers decades from now. The rules require it to hold debt graded safe, and safe debt pays too little to interest a private lending manager. So the managers package private loans into instruments that earn a safe grade from one of the three firms that rate how likely a borrower is to pay you back, while still paying high interest. The insurer keeps the gap between that interest and what it owes its customers as profit. The Journal calls this one of Wall Street's biggest trends, and it is: private lending firms have been buying life insurers precisely to get hold of this balance sheet, the list of what the insurer owns and owes.

Walter's version adds a twist the rest of the industry mostly avoids in public. The loans went to his own businesses, concentrating money meant for policyholders in what the paper calls a web of intertwined investments. Delaware Life and a second insurer have since relabeled about $20bn of investments as deals with related parties, and the group is under federal investigation. Walter and the companies deny wrongdoing and say there was no fraud.

Two things the story does not settle. Whether the loans to affiliated companies have gone bad; nobody has said they have. And how private loans get valued on an insurer's regulatory books when no market trades them. An insurer's assets are valued by the insurer, reviewed by a state regulator, and tested only when a lot of customers want their money at the same time. That is a slower test than a listed fund's discount, and a harder one to see coming. For the meter this is measurement rather than new risk: the loans were made years ago, and what changed this week is that somebody counted them.

A report released Monday by the Senate Permanent Subcommittee on Investigations says USDT, the digital dollar token issued by Tether, has become a primary means of payment for the Iranian regime and a source of funding for its proxies, such as Hezbollah. Investigators looked at 846 digital wallets sanctioned by the US and Israel over links to Iran and found 84% of them dealt exclusively or almost exclusively in USDT. The subcommittee sent the report to the Justice and Treasury Departments. Tether did not respond to requests for comment; it has worked with law enforcement to freeze some wallets linked to the regime (WSJ). Congressional reports are common. One with wallet-by-wallet numbers, handed to prosecutors in the middle of a war, is not.

USDT is about 60% of all such tokens by total value. In practice it is a dollar that moves outside the banking system, backed by reserves the issuer holds. That is the entire attraction for Iran, which Washington has spent a war now in its eighth month trying to cut off from dollars, including an August campaign that the Treasury Department, which borrows the money the government spends, named Operation Economic Outcast. It is also the entire risk for everyone using USDT for honest reasons: the qualities that make it useful under sanctions make it a target for the people doing the sanctioning.

The report comes from Senate Democrats, not the administration, and it follows last week's seizure by the Justice Department of $84mn from a payments firm that handled money for Tether and Bitfinex. What is new is a record made by Congress, with wallet-level figures, delivered to the two departments that could act, during an active war. Nothing has been done yet. The referral is the fact.

Separately, Bitget, a crypto exchange with 120mn users, raised its estimate of what it lost to a hack to $387.5mn from the $351.6mn it disclosed on September 24, after counting transfers on the Zcash and TRON networks it had missed. That makes it 2026's largest crypto theft. Withdrawals, frozen since the attack, reopen network by network this week; the exchange points to a protection fund of 5,500 bitcoin, about $464mn (BFMTV). Two questions about the pipes in one day: whether the biggest dollar token survives its politics, and whether the coins an exchange keeps locked offline cover the hole in the ones it kept connected. Neither moves the meter today. A referral is not an action, and we will say so when it becomes one.

Meta will sell bonds in Europe for the first time this fall to pay for its AI build-out, according to investors familiar with its plans, the FT reports. It joins a queue. Goldman Sachs counts about $500bn raised by AI-linked companies this year, roughly $200bn of it by the five giant cloud-computing companies, Amazon, Alphabet, Meta, Microsoft and Oracle, which are expected to sell more than $1tn of new debt over the next few years (FT). A trillion dollars of borrowing by five companies is not a use of the bond market. It is a change to it.

Three things in the piece show a market being reshaped rather than merely used. Other companies now time their bond sales to avoid clashing with the five. Many are borrowing for shorter periods because the market for bonds that do not come due for twenty or thirty years is glutted with tech paper. And Kevin Warsh, who chairs the Federal Reserve, America's central bank, and Scott Bessent, the Treasury secretary, are both arguing that the five are competing for money with the $31tn market in US government bonds, which puts some of the rise in the government's ten-year borrowing rate to 5.24% at the door of Menlo Park and Redmond rather than Washington. Staff at the European Central Bank asked last month whether euro area markets can "smoothly handle such large and concentrated debt inflows". Meta's arrival is the test.

The idea to hold onto: money spent building things out of a company's own cash is a bet the company loses if it is wrong. The same spending paid for with a bond is a bet the company loses and its lenders share. At $1tn of new borrowing it is a bet the whole credit market shares, and the higher the interest, the less room there is for the AI revenue to arrive late. PGIM's Greg Peters calls the quantum "historic" and "transformational". Quantum here means the sheer amount.

The stock market's verdict differs by borrower. Meta is up 25% on the month even after falling 3.8% on Monday. Oracle, the most debt-dependent of the five, is down 10% on the week and 18% from its high. Same trade, different balance sheets (the lists of what each owns and owes), and the market has started to price the difference. This is fragility, added slowly: more borrowed money resting on one assumption about when the revenue shows up.

The interest rate the US government pays to borrow for thirty years touched its highest level since 2004 on Friday (5.53%), and the ten-year rate reached a level last seen in June 2007 on Thursday (5.22%) before settling at 5.18% (Bloomberg, FT). This is a fast move. As recently as early July the thirty-year rate was below 5%. Friday's trigger was a consumer sentiment survey from the University of Michigan that came in stronger than expected, which means fewer reasons to cut rates. The week's trigger was worse: an auction of five-year government debt on September 23 drew so little demand that it ranked second-worst on record, and a seven-year auction cleared at the highest interest rate since 1993. Governments hold these auctions constantly; a bad one is a sign that lenders want more for their money. Markets now put roughly 70% odds on the Federal Reserve, America's central bank, raising rates again in October, and 16 of its 18 rate-setters have penciled in at least one more increase this year (Nikkei).

Two things about the shape of it matter. First, the short-term rates are rising fastest: two-year borrowing costs 4.87%, five-year more than 5%. Ashok Bhatia at the investment firm Neuberger calls this "a textbook hiking-cycle pattern", meaning it looks the way markets usually do when a central bank is raising rates. Five years is the term most heavily indebted companies borrow for when they refinance, so this is the rate that hurts. SoftBank paid up to 9.75% this week for debt of the kind considered too shaky to be safe. Prospect Capital is paying about 8% to replace debt that cost it 3.364%. Second, it is happening everywhere: Japan's ten-year rate hit 3.115%, last seen in August 1996; Germany's reached 3.62%, its highest since 2009; Britain's is 5.36%.

Who was on the wrong side? Gennadiy Goldberg at TD described an "enormous wipeout in positioning over the past few days" with "very significant losses". In plain terms, a large crowd had bet that long-term rates would fall, and they were forced to sell those bets at a loss all at once. Citi closed a basket of emerging-market bets that depended on calm in American bonds, and did it right after the weak five-year auction (Bloomberg). A bank shutting a trade the same day as an auction is a bank that got hurt by it. And Beth Hammack, who runs the Cleveland branch of the Federal Reserve, said on Friday that the US government now has to compete for funding with AI projects (Bloomberg). Central bankers do not say that lightly. It is the argument this site has been making.

And stocks? The S&P 500 closed Friday 0.7% from the record it set in August. The Nasdaq 100 is 0.6% from a high set on September 22. The VIX, the market's gauge of how much turbulence traders expect over the next month, is 14.9; under 20 is calm. The extra interest shaky companies pay over the government has widened only a tenth of a percentage point in two weeks, to 2.8 points (280 basis points). Here is what that adds up to. The bond market is pricing a cycle of rate rises hitting a system stuffed with borrowed money. The stock market is pricing something else. One of them is wrong. That disagreement is why ignition moved this week and why it did not move more: the fuse is shorter, but the spark has not jumped to shares or company debt.

Oracle has told the landlord of its biggest planned AI campus that it will delay paying full rent, and the lease it signed was written specifically so that it could not do that. Project Jupiter is a 2.45-gigawatt site in New Mexico being built by Stack Infrastructure, a company owned by the investment firm Blue Owl, to run computers for OpenAI. Oracle's lease has what lawyers call "hell-or-high-water" terms: it cannot be terminated, and Oracle must pay rent whether or not it has the electricity to run the place (WSJ). This week Oracle sent Stack a force majeure notice, a claim that events beyond its control excuse it from performing on time, pushing back the date at which it pays full rent. The lease allows a delay of up to three years. The first phase had been due in the third quarter of 2028.

The reasons are on the ground, not in the spreadsheets. Oracle reworked a plan to power the site with gas turbines and diesel generators after local resistance, but has not obtained state air permits. A 17-mile gas pipeline crossing state land has been rejected twice. And Oracle removed a key land and power partner from the project. Oracle says Jupiter "remains on our planned schedule". Blue Owl says the notice "does not change the financial commitments". Both statements can be true and the debt can still have a problem.

Here is the problem. Around $18bn of loans were made against this development, and the FT reported on September 18 that they were trading near 90 cents on the dollar (FT Alphaville). Loans considered safe trade at or near 100 cents. Ninety means the lenders already doubted something before this week. Those loans were made on the assumption that Oracle's rent would arrive and pay the interest. The lease says Oracle must pay eventually; force majeure says not yet. Between "eventually" and "not yet" sit the interest payments, and nothing published says who covers that gap. The law firm Quinn Emanuel, in a note to clients quoted by Axios, says force majeure "is no longer a back-end boilerplate provision" but a tool that decides whether a delay "cascades through the project's construction, customer, power, insurance, and financing documents". In other words, the clause nobody used to read is now the one everything depends on. Johnathan Owen at the fund manager TwentyFour says the desks that buy the debt of safe companies are "aggressively going through these docs". Those desks are used to lending to Procter & Gamble, not to holes in the ground.

The politics are spreading too. Texas's attorney general opened an investigation into hundreds of data centers over how they report water use, days after the governor halted new permits (Bloomberg). Oracle's shares fell as much as 8% on Thursday and are 15.7% below their high. Blue Owl is down 22% in 20 days. The same Friday, a filing showed that Larry Ellison, Oracle's founder, has pledged 67 million more of his Oracle shares as security for personal loans, about $9.2bn worth, taking the total to 36% of everything he owns in the company, while he helps fund his son's $111bn bid for Warner Bros (Bloomberg). A founder borrowing against a third of his stake in a falling stock is not a crash on its own. It is another way the same shares are holding up more than one thing.

This story does not make a crash more likely today. The risk in these loans was already in our fragility count, and the loans were already at 90. What changed is that the contract everybody was relying on has been tested for the first time, and the answer so far is "not yet". That is a risk being discovered, not a risk growing. If the loans mark down further, or a lender sets money aside against unpaid rent, that would be growth.

Prospect Capital, a fund that borrows money, lends it to mid-sized private companies and passes the interest to shareholders, is raising $500m of five-year debt at an interest rate of about 8% (Bloomberg). It is the first time it has sold this kind of bond since 2021, and the money will replace debt that costs it 3.364% and comes due in November. That is the whole story in two numbers: the same loan, taken out again, at more than twice the price. Moody's, one of the three firms that grade how likely a borrower is to pay you back, rated the new bonds Ba1 on Tuesday, one notch below the line that separates safe borrowers from shaky ones, citing two years of falling earnings and a riskier loan book than its peers. Prospect lost its safe rating in 2024. The company says the deal lifts the average interest it pays on all its debt by up to half a percentage point, to about 6.5%, "still well below" the roughly 11% it charges the companies it lends to.

That sentence is the mechanism, so it is worth sitting with. A fund like Prospect earns the gap between the 11% coming in and the 6.5% going out, minus what its managers take in fees, minus the loans that go bad. Every time an old bond comes due and is replaced, a 3% cost becomes an 8% cost. The other side cannot follow, because a company already paying 11% mostly cannot pay 13% without going under. Bloomberg notes that with five-year government borrowing costs at a near two-decade high there is "less room" for private lending funds to earn anything after fees and normal losses, and that even the funds with safe ratings are paying more. This is not confined to funds. To sell a €2.8bn loan financing Platinum's buyout of Nestlé's water business, banks had to raise the interest by a full percentage point, to four percentage points above the benchmark rate (400 basis points), and cut the price to 97 cents on the dollar to find buyers (Bloomberg).

On the other side of the ledger, some good news, and we say so plainly. This morning's FT reports that the rush of savers trying to pull money out of these funds is slowing (FT). Ares's income fund received withdrawal requests for 13.1% of its shares in the third quarter, down from 14.4%. Apollo's $15bn fund had requests for 14.7%, from 16.8%. BlackRock's HPS fund 11.5%, from 13.3%. Blackstone's $43bn fund held steady near 10%. All four have limited payouts to 5% of the fund per quarter since the second quarter, and on average have met less than half of what was asked. Peter Troisi at Barclays: "too early to call a bottom". We agree, and we record the easing anyway. Requests are still two to three times what the funds will pay out. A queue that is shrinking is still a queue.

The share prices tell you what investors think the problem is. Blackstone is down 17.5% in 20 days, Blue Owl 22%, Ares 13%, KKR 12%. Those are the firms that run the funds. The funds themselves, the ones that hold the loans and trade on the stock market, have barely moved: Ares Capital is down 1%, Blackstone Secured Lending 0.2%. Investors are marking down the managers' fee income, not the loans. Either the loans are fine, or nobody has tested them yet. For the meter, this is a wash. Refinancing at 8% adds pressure to a pile we already count; the easing in withdrawals takes some out. Fragility holds.

Two office towers in Philadelphia that were valued at $471m in 2019 are being sold for $70m, a fall of 85%, and the loss reaches investors who held what was rated the safest debt in the market. The buildings are Centre Square, at 1500 Market Street. A judge approved the sale last month and it closes in the coming weeks (Bloomberg). The $368m loan on the towers had been bundled into bonds and sold to investors in slices, ranked by who gets paid first. The slice once rated AAA, the top grade, is expected to recover around 44 cents on the dollar after fees and other claims are paid. The seven slices beneath it get nothing. Pimco, one of the world's largest bond managers, is the biggest holder, with roughly $58m of the debt at face value, about half of it in the top slice. Bloomberg estimates its loss at more than $35m.

The mechanism is worth spelling out, because AAA does not mean what most people think. This bond was backed by one building and one borrower. The AAA rating never said the building was safe. It said the holder was first in line, and that the slices below would absorb losses first. That cushion was sized on the assumption that a downtown office tower could not lose most of its value. An 85% loss is bigger than any cushion anyone built. For seven years the 2019 valuation stood on the books, untouched. It was tested only when a court forced a sale. That is the pattern we keep flagging: a price is not a price until somebody has to sell at it.

Alan Todd, who follows these bonds at Bank of America and ran his own numbers on the recovery, says office values in Chicago, downtown Los Angeles, Portland and Denver "have not bounced back" and "you're definitely going to see more of this" as loans come due over the next year or two.

For our running argument this cuts an interesting way. Bonds like this one are where prices get tested in public: there are trustee reports, ratings, a court date. The same kind of building sits inside private real-estate lending funds and on the balance sheets of insurers, the lists of what they own and owe, with none of that machinery. We already count that risk in fragility, and today's number does not add to it. The world did not get more dangerous this morning; our measurement of it improved. What the sale does is set a reference price. The next time a private lender values an older office loan at close to full value, somebody will ask why the AAA next door got 44.

Foreign investors bought a record $942bn of American shares and funds, net of sales, in the twelve months to July. That is the largest twelve-month total since the Treasury Department, which borrows the money the government spends, began keeping the data in 1985 (FT). The second quarter alone was $426bn, up 62% on a year earlier and above the previous quarterly record of $299bn set in 2022. Foreign buying of US government debt went the other way: $188bn in the second quarter, down from $314bn in the first. China's reported holdings of US government bonds are $618bn, the lowest since August 2008, as Beijing buys gold and the bonds of government-backed mortgage agencies instead.

This is the missing link between the two markets we described above. Overseas money is still coming to America, but it is going into stocks rather than bonds. That is why the government is having to pay 5.18% to borrow for ten years while the S&P 500 sits within a percent of its high: the next foreign dollar wants Nvidia, not a government IOU. Brad Setser, who tracks these flows, thinks some of the second quarter was catch-up after a weak first quarter. The FT's sources describe Korean and Taiwanese investors who own as much Samsung and Hynix as their rules at home allow, and are buying American shares to spread the risk. The monthly figures matter here: $110bn in April, $182bn in June, and $3.7bn in July.

A dollar propped up by stock purchases is a dollar whose support is repriced every afternoon at the close. The dollar index, which tracks the dollar against a basket of other currencies, is at 101, up 1.3% in 20 days. Gold is down 4.6% over the same period, to $4,321. The yen is the mirror image. Japan's finance minister confirmed on Friday that Donald Trump raised the weak yen with Prime Minister Takaichi on Tuesday, and that she and Scott Bessent, the US Treasury secretary, will keep coordinating on exchange rates. This follows a $96bn intervention in July and August, backed by the US, to prop the yen up (FT, MoF). The Bank of Japan raised rates to 1.25% on September 18 and Japan now pays more than 3.1% to borrow for ten years, and the yen is still ¥158 to the dollar. Higher Japanese rates have not pulled Japanese money home. Higher American rates have not attracted foreign buyers of American bonds. Money is chasing the stock market's returns, and those returns depend on AI.

This is fragility, not ignition. Nothing here is a spark. It tells you that the thing holding the dollar up is the thing most exposed to a fall in a handful of AI stocks, which makes the pile bigger, not the fuse shorter. The July collapse to $3.7bn is the number to watch when August's data arrives. One month is noise. Two would be a trend.

Three regulatory moves landed within a day of each other, pointing in different directions. The Federal Reserve, America's central bank, is considering raising the size at which a bank triggers stress tests and stricter oversight: the lower line from $100bn of assets toward $150bn, the top line from $700bn toward $1tn, with formal proposals possible later this year (Bloomberg, citing Reuters). Those lines were set in 2019. JPMorgan and Citi are so far above them that nothing changes. US Bancorp and Capital One get room to grow before the heaviest rules apply.

The same morning, a senior official at Japan's Financial Services Agency, which supervises that country's banks and insurers, said it is stepping up scrutiny of the loans Japan's biggest banks and life insurers are making to build AI data centers, mostly in the United States. It is examining how they manage the risk and cited "uncertainty over whether hyperscalers can generate adequate returns" (Bloomberg). Hyperscalers are the giant cloud companies, Microsoft and Amazon and their peers, who are the tenants. The agency says it does not want to discourage the lending. Regulators say that when they are worried and do not want to cause a panic.

And US state insurance regulators wrote back to Senator Warren. The letter from their national body, the NAIC, says the "continuing modernization of these tools does not reflect an insurance regulatory vacuum requiring enhanced federal guardrails", and points to a tighter definition of what counts as a bond, the power to challenge credit ratings paid for privately, and more reporting on private investments (Bloomberg). The exchange began when insurers controlled by Mark Walter of Guggenheim disclosed that more than $20bn of loans should have been labeled as made to related companies, and were not.

Read together: supervision is being extended where the risk is new, Japanese insurers funding American data centers, and loosened where it is old, at mid-sized American banks. Whether $150bn is the right line is a question we cannot answer from the reporting. What we can say is that the market has its own view of what 5% five-year rates do to a bank's balance sheet, the list of what it owns and owes. Schwab, the bank that got into trouble in 2023 by holding long bonds against short-term deposits, is down 8% in a month. Capital One is down 7.6%. None of this moves the meter much this week. The Japanese scrutiny slightly improves how well the AI lending is understood, which is the direction we want. The higher bank thresholds slightly reduce it. We include the story because the pattern is worth knowing, not because it changed the odds.

The next test is Friday's jobs report. The consensus is 100,000 new jobs and unemployment at 4.2%. Bank of America expects 60,000, which it says would ease the pressure for a rate rise in October (FT).

The Federal Reserve, America's central bank, which sets interest rates, is expected to raise them on Wednesday for the first time in three years, and on Sunday the president who appointed its chair demanded the opposite in public. "We should be paying the lowest interest rate in the world, regardless of their formulas," Donald Trump told reporters in Ireland. Traders put the odds of a rise at 86%, with a second one expected before the end of the year (Bloomberg). The chair is Kevin Warsh, Trump's own pick, and a president leaning on his own appointee three days before a decision is not routine.

The set-up is unusual in three ways at once. The Fed has held rates steady for five straight meetings, and in July three of its officials (Hammack, Kashkari and Logan) voted against the majority in favor of a quarter-point rise, so the committee was split before Friday's inflation figures for August arrived (prices up 0.4% on the headline measure, 0.3% on the core measure that strips out the most volatile items). The Treasury Department, which borrows the money the government spends, spent last week buying back its own ten-to-twenty-year bonds at triple its usual size to hold down the rates on long-dated debt. And John Williams of the New York Fed says the central bank will stop shrinking its balance sheet, the list of what it owns and owes, in December and is "getting closer" to buying bonds again for what he calls technical reasons: the cash banks keep at the Fed has fallen to its lowest in 39 weeks ($2.895tn), while the Treasury's own cash pile there has risen to $967.9bn.

So: short-term rates are about to go up, the government's long-dated debt is being bought to push its rate down, and the president is asking for the opposite of both. The thing to watch is the extra interest lenders demand for tying money up for a long time when they cannot tell what policy will be (the term premium). A rate rise delivered by a chair under public attack from the man who appointed him can push long-term rates up rather than down, if investors decide the rise is a one-off made under duress while inflation is not, with Brent crude at $107 after the strike on the Saudi pipeline, tariffs, and a war in its seventh month. That is our inference; Warsh has not said it. The interest rate the government pays to borrow for thirty years closed Friday near a 19-year high (5.354%), before any of this.

Which of the three the market believes is the question Wednesday answers. The prices in front of us are Friday's; nothing in them has reacted to Sunday. This is an ignition story, not a fragility one: nothing has been added to the pile, but the moment that could set it off is three days away, and we cannot yet say which way it goes.

Marsh, the world's biggest insurance broker, says AI data centers are about to become the next big users of captive insurance, in which a company sets up its own in-house insurer instead of buying cover from someone else. That matters because these buildings are mostly built with other people's money, and a lender to a building normally relies on an outside insurer standing behind it. Companies have run captives for decades; what is new is that the buildings involved sit behind roughly $500bn of AI-related debt sold this year.

"What you're going to see is an explosive growth in the use of captives to take on the portfolio risks for data centers," Michael Serricchio, who runs Marsh's captive business in the US and Canada, told Bloomberg. "Inadvertently, some of the risk for build-outs, construction, surety, property and liability will end up in their captive." He would not name companies.

The reason is capacity. The facilities are larger than anything traditional property insurers have taken on, and many sit on cheap land that comes with tornadoes, floods and drought. When the market will not take a risk at a price the owner likes, the owner keeps it and calls the retained risk an insurer. Oil companies pioneered captives fifty years ago for exactly this reason, so the structure itself is ordinary.

Why it matters here is the debt. These campuses are increasingly financed through a separate company set up to hold the building off the parent's books, in which the tech giant owns about 20% and investment funds or bond investors put up the other 80%. Meta's Hyperion campus, a $27bn package arranged by Morgan Stanley with funds run by Blue Owl holding 80%, is the template (Reuters Breakingviews). Lenders to a building normally require insurance on it. If that insurance is written by a captive owned by the sponsor, the lender's protection against the roof coming off is, in the end, the sponsor's willingness and ability to pay itself. That is fine while the sponsor is Meta. It is a different proposition for a smaller AI cloud company or a bitcoin miner turned landlord.

What the reporting does not say: how much money these captives hold, whether the lenders to those separate companies are accepting captive cover, or how much of the physical risk behind the roughly $500bn of AI-related debt sold this year sits in structures like this. This is fragility, not a spark, and it is the pattern we already count: a risk that used to be priced by an outsider moves inside the structure, where nobody outside puts a price on it. It does not move the number today because it is the kind of risk the number already holds.

Two sets of figures published in the past week describe the same second quarter for BDCs, the funds that borrow money, lend it to mid-sized private companies, and pass the interest to shareholders, and they appear to disagree. The dollar total of loans that have stopped paying interest fell; the typical fund's share of such loans jumped. Both are right, and the reason is that the trouble is in smaller funds lending to smaller companies, which a dollar total barely notices.

The data firm Octus counted $9.5bn of debt on which interest has stopped arriving (non-accrual) across 173 BDCs, down 5% from $10bn in the first quarter, or 1.9% of what the loans cost. KBRA, a firm that grades borrowers, reported that for the median fund among those with a fixed life (the "non-perpetual" ones) the figure jumped to 2.75% of cost from 1.81%, a rise of 0.94 percentage points in a single quarter, which is a lot for three months. A Reuters cut of 44 BDCs put the total at 3.4% of cost, up from 2.5% at year-end, with portfolios valued at 97.57% of cost (matterfact).

The reconciliation is arithmetic. A dollar total is dominated by the largest funds, which lend to the largest borrowers; a median gives every fund one vote. The dollars fell and the median rose because the deterioration is in smaller funds lending to smaller companies. Houlihan Lokey's data make the same point from the borrower's side: among companies earning $10–20m a year, 12% of loans are now valued below 90 cents on the dollar, up from around 1%. Across the sector, 11.8% of loans by size paid at least some interest by adding to the debt rather than handing over cash; interest switched to that arrangement after the loan was made, because the borrower could not pay ("amended PIK"), was 1.6% of interest dollars.

The stock market has already sorted the funds the same way. The nine BDCs whose shares trade above the stated value of what they hold carry average non-accruals of 0.9%; the ten trading more than 30% below it carry 2.6%. Ares Capital and Blackstone Secured Lending are within 2.5% of their highs; Blue Owl's OBDC and FS KKR are 6–7% off theirs. Roughly $772m of interest income across the sector is owed by borrowers who are no longer paying it.

None of this changes our fragility number. It is June data reaching us in September: the world did not get more dangerous this week, our picture of it got sharper. What it changes is where to look. The system-wide headline can fall while the worst cases get worse, and the worst cases are where the smaller managers and the funds sold to ordinary savers with limited withdrawal rights live.

More American car loans are seriously behind on payments than at any point since the record began, including the depths of the last crisis: about one in every eighteen dollars of car-loan balances was at least ninety days late in the second quarter (5.5%, per the New York Fed, above the 5.3% peak of 2009) (The Money Overview). The stress sits where it always sits, among young and low-income borrowers. The lenders who assumed it would arrive are holding up better than the lenders who did not.

Fitch, one of the three firms that grade how likely a borrower is to pay you back, says the share of bundled loans to the riskiest car buyers (loans packaged and sold to investors) at least sixty days late touched 6.90% in January, a 32-year high. Borrowers aged 18 to 29 are seriously behind at 4.83%; those aged 60 to 69 at 1.77%. Fitch calls it a "K-shaped economy", with affordability pressure "weighing disproportionately on lower-income, highly leveraged borrowers". In other words, the squeeze is falling on the people with the least money and the most debt.

Two things in the same reporting cut the other way. The rate on those bundled subprime loans has come down from 6.5% at the end of 2025 to 5.8% at the end of June, and the ninety-day figure itself edged down to 5.49% from 5.60% in the first quarter. A ninety-day delinquency is a backlog: it counts loans that went bad months ago and have not yet been written off. Whether the pipeline behind it is still filling, these numbers do not say.

The share prices are not panicking about the borrowers you would expect. Credit Acceptance, which lends to the riskiest car buyers of all, closed 0.7% off its high and is up 3.6% in twenty days. Capital One is 8% off its high, Ally 6%, Synchrony 6%, Affirm 9%. Our reading, and it is inference: Credit Acceptance buys loans from dealers at a discount that assumes heavy losses from the start, so record delinquency is closer to its plan than a surprise, whereas the lenders to safer borrowers were priced for a customer who was going to be fine. The market is saying the stress is real but sitting where it was expected to sit.

That is reassuring only up to a point. The lenders who assumed the worst are holding up; the lenders who did not are being repriced. That is the pattern of the early stage of every credit cycle. This does not move the odds: it is second-quarter measurement of risk we already carry, and the household beat has been the quietest of our eight because the stress is visible, priced, and so far confined to the borrowers who were always marginal.

The Dutch central bank moved 86 metric tons of gold out of the United States and Canada this month, calling the move a hedge against "extreme systemic risks". Central banks moving their gold across borders is rare enough that each case makes news, and this one came with a reason attached. France has already taken all of its gold out of New York, Germany has debated it, and Norway's sovereign wealth fund, the world's largest, says it will cut its holdings of US government bonds amid the turbulence in that market (Axios). None of it has yet shown up in prices, which is the most useful fact in the story.

Read against that, a study from the New York Fed published the same week is almost a rebuttal. The dollar's share of the reserves central banks hold fell from 64% in 2015 to 56% in 2025, but the researchers find the decline is driven by a handful of countries: China and Russia in 2015–19, then China, Russia, Mexico and Morocco, which together account for nearly all of the 2.3-point drop between 2019 and 2023. Roughly as many countries increased their dollar holdings as reduced them (Bloomberg).

Both can be true, because they measure different things. Moving bars from a vault in Manhattan to one in Amsterdam does not sell a single dollar; it takes the bars out of US jurisdiction. That is a statement about trust in the custodian, not about the currency. Norway's plan is different: it reduces demand for US government bonds at the margin, in a week when the Treasury Department could not find sellers for the full $6bn of bonds it wanted to buy back. And the New York Fed's finding covers reserves through 2025; it says nothing about what allies decided in September 2026.

The prices side with the New York Fed for now. The dollar index, which tracks the dollar against a basket of other currencies, is flat on the week at 99.1 and down half a percent on the month. Gold, at $4,382, fell 2.1% in five days while long-term government borrowing rates went to 19-year highs, which is the opposite of investors running for safety. The yen is up 3.75% in twenty days on the prospect of Japan's central bank raising rates, which is an interest-rate story, not a flight. If the world were losing faith in the dollar as the currency it keeps its savings in, gold would have risen as bonds fell. It fell with them. This does not change the odds; it is one more reason to believe the slow part of the risk is still slow.

Traders in futures markets now put the odds of the Federal Reserve, America's central bank, raising interest rates next week above 85%, according to the FT, up from the 70% we quoted last run. Friday's data showed inflation stuck at 3.4%, with oil above $100 a barrel and federal debt past $40tn, and Governor Christopher Waller had already said a rise 'may be appropriate' if August's improvement proved fleeting. In the same week, the president of the Fed's New York branch, John Williams, said the Fed is 'getting closer' to buying bonds again. Raising rates is how a central bank tightens and buying bonds is normally how it loosens, so the two arriving in the same season is the thing worth understanding.

Williams describes the purchases as technical, with 'no implications for monetary policy', and says the Fed will stop letting its bond holdings shrink in December (sweep). The plumbing explains why both can be true at once. Banks keep cash on deposit at the Fed, called reserves, and that pile fell to $2.895tn in the week to September 2, the lowest in 39 weeks. It fell because the Treasury Department, which borrows the money the government spends, has built its own account at the Fed up to $967.9bn, a four-month high, and every dollar the Treasury holds there is a dollar not in the banking system.

There used to be a cushion. The Fed's overnight facility for spare cash, where lenders park money with the Fed against bonds put up as security, absorbed drains like this from 2022 to 2024. It is now down to $626mn, empty in all but name. With no cushion left, keeping overnight interest rates inside the Fed's target range means the Fed has to add to what it owns, whatever it decides about the policy rate.

So the plausible autumn picture is a Fed that raises rates, a Fed that buys short-term government IOUs, and a Treasury that buys back government IOUs that do not come due for twenty or thirty years at up to $14.5bn a week. None of that contradicts itself. The difficulty is what it looks like. Chair Kevin Warsh told the Jackson Hole conference two weeks ago that the Fed's balance sheet, the list of what it owns and owes, should be reserved for genuine crises and otherwise used sparingly, if at all. A restart labeled 'technical' three months later will be read by some as exactly what he said he would not do.

The market has priced half of this. Three-month government IOUs pay 3.91%, already above the 3.62% that banks pay to borrow cash overnight against bonds put up as security (SOFR), which is traders betting on the rise. What nobody has yet had to price is the Fed explaining purchases and a tightening in the same statement.

On the odds of a crash, our read is that this is the plumbing working: the Fed noticed the cushion was gone and is refilling it before anything seizes. It adds nothing to the pile of borrowed money and brings no spark closer, which is why both halves of the meter hold. Plenty of people we respect will read a central bank buying bonds while it raises rates as a central bank that has lost control of its own signal, and if the market takes that view next week, ignition is where it would show.