Treasury doubles its buybacks while the Fed quietly stops buying

Archived reading, published Wed, 19 Aug 2026 22:23:03 UTC (7 days ago). This is not the current state of the meter.

See the live reading →

CRASH-O-METER

Cracking

how close are we

Fragility89

how much tinder is stacked up — moves slowly

Ignition43

how close a spark is — moves fast

Crash Lab watches the machinery under the current boom: the debt paying for AI data centres, private credit, the leverage that has moved out of banks into places nobody has to mark, the bond market and the dollar.

Every four hours it reads the day’s reporting from 17 sources and rewrites this page.

Status: Held at 66. The one genuinely new structural fact this window is a policy contradiction rather than a break: the New York Fed has halted reserve-management purchases entirely through 14 September while the Treasury doubles its long-bond buybacks from 9 September. Gold's 4.9% day and the dollar's three-month low are a currency repricing, not a funding event — VIX is 14.9, high-yield spreads are unchanged at 275bp and the 30-year fell 10bp — so ignition does not move.

Reporting from 18 Aug to 19 Aug

Crash points

Eight places it could go, scored 0–10. Tap one for the explainer.

What’s moving the needle

What changed in the last few hours, and what each one says about the plumbing.

Fed, Treasury and policy19 Augmedium

One hand buys bonds, the other stops

The government is managing the long end by shifting its borrowing to the front end at exactly the moment the Fed has stopped supplying reserves there.

On 19 August the Treasury said it would increase its buybacks of 10- to 30-year government bonds "by at least double" — from $2bn to at least $4bn per operation, running 9 September to 4 November, which analysts put at roughly $32bn a quarter (Bloomberg). The 30-year fell 10 basis points to 5.18% on the news, back from its highest level since 2007. Six days earlier, the New York Fed announced the opposite move at the other end of the curve. Reserve management purchases — the Desk's routine buying of Treasury bills to keep bank reserves "ample" — were cut to zero for 14 August to 14 September, having already been tapered to about $10bn a month. These are different instruments and it is worth being precise about why. A Treasury buyback creates no money. Treasury retires long bonds and funds the purchase by issuing short bills: the total debt is unchanged, the average maturity shortens, and someone else has to hold more bills. A Fed reserve-management purchase does create reserves. So the government is currently taking duration risk out of private hands and simultaneously not adding the bank reserves that would normally help finance the resulting bill supply. What has to stay true for this to work: money funds and banks keep absorbing bills at the front end without the repo market tightening. Watch SOFR, which sits at 3.65%. It has not moved. If it starts printing above the Fed's administered rates in the second week of September, when buybacks begin and reserve purchases are still off, that is the tell.

The dollar, gold and reserve status19 Aughigh

The long end got relief, the currency paid

A policy that lowers long yields by weakening the currency has not reduced the risk, only moved it to a different price.

Gold rose 4.92% on Tuesday to $4,580 an ounce. Silver rose 4.47%. The dollar index fell 0.86% to 98.80, a three-month low, and is down 2.3% over twenty days. The euro and yen ETFs both rose about 0.9%. Put that next to the previous dispatch and you have the trade in one line: the Treasury announced it would buy back its own long bonds, the 30-year fell 10bp, and the price of that relief was paid in the currency. Bonds and the dollar are supposed to be two views of the same asset. When yields fall because a large official buyer appears rather than because inflation expectations fall, foreign holders are being told something about how the debt will be managed, not about how it will be repaid. Bloomberg's own columnist called the move "desperate" and noted the buyback programme was explicitly created in 2024 as market-functioning plumbing and described by the previous administration as "not intended to ameliorate periods of acute market stress" (Bloomberg). It is being used in one now. We should be careful about how much to read into one day. Gold has been rising for months — up 10.5% in twenty days — and part of Tuesday's move is oil, the Iran war and Japanese and European long yields at multi-decade highs, not a referendum on Washington. But the direction of the correlation matters. When a Treasury announcement lowers yields and lowers the dollar at the same time, the bond market is no longer the safe asset in the trade. Gold is.

Hidden leverage and shadow banking19 Augmedium

Why chip volatility got cheap: nobody needs the hedge

Cheap volatility is being read as calm when it is partly the residue of a hedge unwind after a fund failed.

Implied volatility on the VanEck Semiconductor ETF collapsed from 65% last month to 40% on Monday, the lowest since February. The obvious reading is that everyone has calmed down about AI chips. The actual mechanism, per one trader who runs semiconductor options for clients, is plumbing. "Bank exposure to leveraged ETF and Situational Awareness this summer got to the point they felt very exposed to jump risk in semiconductor names and that caused hedging and volatility to go way up," Zed Francis of Convexitas told CNBC. "Now they don't need those hedges, and I believe unwinding of those hedges has made volatility in the sector inexpensive." This is worth understanding because it generalises. When a dealer sells leveraged ETF exposure or finances a big directional fund, it buys options to cover the jump risk it has taken on. That buying raises implied volatility. When the fund blows up — Situational Awareness, the $20bn-plus AI fund run by a 24-year-old, cost Jane Street $1.5bn in July according to Reuters — the dealer's exposure disappears and it sells the hedges back. Implied volatility falls. Nothing about the world got safer; a buyer of insurance left the market. So a low VIX-style reading in a sector can be the aftershock of a blowup rather than evidence there wasn't one. Someone appears to agree: the biggest single options trade in the whole market on Monday was a $129m purchase of 20,100 deep in-the-money November puts on the semiconductor ETF — over a third of the day's premium in the fund, and almost certainly a fresh synthetic short.

Household credit19 Aughigh

The long end reaches the kitchen table

The fiscal argument at the long end of the Treasury curve is now setting the mortgage rate, and the housing market has stopped clearing.

Pending US home sales fell 2.3% in July to an index of 71.2 — the lowest since January and the joint second-worst reading in data going back to 2001. Economists had expected no change. All four regions fell; the West dropped 4.7% (Bloomberg). This is the transmission channel that makes the bond dispatch matter to people who do not own bonds. The 30-year mortgage tracks long Treasury yields plus a spread. Thirty-year Treasuries hit 5.33% this week, the highest since 2007; the average 30-year mortgage is sitting just below its one-year high of 6.81%, with the Washington Post putting key rates back around 7%. NAR's chief economist: "The highest mortgage rates of the year hit right in the middle of summer, and that's pulling back contract signings." The second-order effect is a lock-in. Anyone holding a 3% mortgage does not sell into a 7% market, so supply shrinks, prices stay at records, and the market clears through volume instead of price. That is why home prices can be at record highs while contract signings sit near a twenty-five-year low. It is not a contradiction; it is a frozen market. Underneath, the credit data are worse than the price data. The New York Fed's Q2 release puts overall delinquency at 4.7% of balances, seriously delinquent student loans at 10.6% and auto at 5.5%; Fitch had subprime auto ABS 60-day-plus delinquencies at 5.67% in June. Those are second-hand figures via a data sweep rather than the primary release, so treat the decimals loosely — but the direction is not in doubt.

Crypto and TradFi contagion18 Aughigh

Strategy is selling. Its shares are at a high.

The largest digital-asset treasury company has stopped being a buyer, and its shareholders are pricing it as though it hasn't.

Bitcoin rose 7.29% on Tuesday to $69,398, up 10.2% on the week. Strategy — the company whose entire purpose was to own bitcoin — rose 12.64% to $104.21, its highest of the period. Two ether treasury companies did the same: SharpLink +12.2%, BitMine +10.7%, both at period highs. Here is the disagreement. Strategy has not bought any bitcoin since mid-June. Over the past five weeks it has sold about $2.1bn of common stock, used the proceeds to buy back $347m of its STRC preferred shares and build a $4.8bn cash reserve, and sold $213.3m of bitcoin (Bloomberg). It still owns roughly $59bn of the coin. The flywheel — issue stock at a premium to net asset value, buy coins, report a bigger stack, sustain the premium — is running backwards, and management says so. So shareholders are bidding up a company that is a net seller of the asset they are buying it for. Two readings. One: the stock is now trading as a leveraged bitcoin proxy on beta alone, and a 7% day in the coin mechanically produces a 12% day in the equity regardless of what the treasury is doing. Two: the market thinks the pivot to capital management — cash reserve, preferred buybacks, no new leverage — makes the company more solvent, so it deserves the re-rating. We lean to the first. The stock is down roughly 73% over a year against bitcoin's 45%. Nothing about a 12% day reverses that arithmetic; it just tells you the sensitivity is still there in both directions.

Private credit and BDCs18 Augmedium

A $486m loan becomes a $672m loan on the same dirt

Each refinancing that moves from a bank to a non-bank moves a valuation from a supervised process to a private one.

Starwood Property Trust and Realterm have lent $672m against 78 industrial outdoor storage properties — 830 acres of parking lots across 33 markets, owned by affiliates of Stonemont Financial and Cerberus. The money refinances an existing $486m loan on the same assets (Bloomberg). Small deal, clean illustration of the thesis. Realterm's head of credit says the obvious part out loud: the firm "has been expanding its lending platform as traditional banks pull back". The part worth dwelling on is the 38% increase in loan size against an unchanged pile of land. That is not necessarily aggressive — the borrower may be adding properties, rates and rents have moved, and we do not know the loan-to-value on either deal. But it is how valuation works in this corner of the market. There is no exchange price for 830 acres of truck parking. The mark is whatever the last lender was willing to advance against it, and the last lender is now a non-bank whose own funding comes from investors who see the loan at cost until something goes wrong. When a bank held that loan, a regulator looked at the appraisal and a supervisor could force a writedown. When Starwood holds it, the discipline is the manager's own underwriting plus a rating agency if the paper gets securitised. Sometimes that is better underwriting — Apollo, after all, spotted First Brands' "bogus collateral" a year before anyone else. Sometimes it is just a bigger number on the same lot.

Signs of the times

Stuff you wouldn’t have believed was possible until 2026.

Eric Trump's American Bitcoin told an earnings call it is "fast becoming the leader in Bitcoin" with "the strongest brand of all". A filing a month later disclosed the company had two full-time employees.

19 AugForbes via Capital.gr

Leopold Aschenbrenner enrolled at Columbia at 15, graduated valedictorian at 19, worked at FTX, wrote an essay about AI, and turned it into a fund managing tens of billions. It blew up days before his wedding; Reuters reports Jane Street took a $1.5bn hit tied to it.

19 AugNew York Magazine

The OCC has granted preliminary approval for a national trust bank charter to a subsidiary of World Liberty Financial, the Trump family's crypto firm, letting it issue its own $4bn USD1 stablecoin directly — the fourth-largest in the market — while the crypto market-structure bill it would sit under remains stuck in the Senate partly over the president's crypto interests.

18 AugCinco Días

Australian banks are being hit by a surge in home-loan fraud sourced from hundreds of "template farms" — businesses that sell editable fake payslips and bank statements. The mortgage boom has produced a supply chain for forging the paperwork it requires.

19 AugAustralian Financial Review

The rumour mill

What the crash-callers are saying, checked against real reporting.

Partly true

“The Fed's 'reserve management purchases' are backdoor quantitative easing.”

They do create reserves and expand the balance sheet, but the Fed frames them as plumbing to keep reserves ample rather than easing. The more interesting fact is the other direction: the Desk has scheduled zero of them for 14 August to 14 September.

Claimed by Gregory Mannarino

Partly true

“The Fed has been buying $40bn a month of Treasuries for months while calling it liquidity support.”

Purchases ran at up to that scale earlier in the programme, but the most recent scheduled window was about $10bn and the current one is zero. The claim is directionally right about the framing and wrong about the current number.

Claimed by Thoughtful Money (Adam Taggart)

Partly true

“The Treasury is doing buybacks because rising long-term yields forced Washington to respond, and larger buybacks will follow if yields keep rising.”

The timing is exactly as described — the doubling came days after the 30-year hit a 2007 high — but Treasury's stated rationale is liquidity support, and it has not said it will scale further with yields. Bloomberg's own opinion desk reached the same conclusion the channels did.

Claimed by Gregory Mannarino

Confirmed

“The 30-year Treasury just hit a 19-year high, and government bonds worldwide are hitting all-time high yields.”

The first half is right: 5.31-5.33%, highest since June 2007. The second half is loose — Japanese 10-years are at a 1996 high and 30-year gilts near a post-1998 high, which is not the same as all-time highs.

Claimed by Anthony Pompliano

Earlier developments

Dispatches from previous readings. The same argument, no longer the news.

Private credit and BDCs18 Aughigh

The insurer that lent to its owner's friends

The insurance-plus-private-credit model rests on assets whose prices are set internally; this is the first case where a regulator is forcing them to be swapped for something an outsider valued.

Insurers controlled by Mark Walter — the Guggenheim chief executive who owns the Dodgers and, until last week, the Lakers — disclosed after receiving federal subpoenas that they had made more than $20bn of loans to related parties without disclosing the connections. On Tuesday one of them, Delaware Life, said it had agreed to exchange $6.5bn of assets linked to TWG corporate affiliates for unrelated investments. Even after the swap, Delaware Life and a second Walter insurer are left holding more than $10bn of affiliated investments (FT). Here is why this matters beyond one billionaire. The model at issue — pair a life insurer's annuity liabilities with private credit assets originated by an affiliated manager — is now standard across the industry. It works because annuity money is long, sticky and does not run, so it can fund loans that are illiquid and infrequently priced. The weak point is who sets the price. When the borrower is a related party, the loan's mark is not a market observation; it is an internal agreement between two entities with the same owner. S&P put Delaware Life on negative outlook last month after it restated its accounts and reclassified a large volume of private credit as affiliated. The FT also reports Walter spent the summer negotiating a multibillion-dollar loan from Apollo secured on his Lakers stake, talks overtaken by the $12.5bn sale to Bob Iger and Josh Kushner. A man raising cash against a basketball team is not by itself a systemic event. An insurance balance sheet whose assets turn out to be loans to its owner is closer to one, and the clean-up is happening under a Delaware regulator's supervision rather than in a fire sale. That is why we read it as fragility being measured, not fragility rising.

Hidden leverage and shadow banking18 Aughigh

Two frauds, one mechanism: collateral pledged twice

Non-bank lending has scaled faster than the plumbing that verifies collateral exists, and both recent blow-ups turned on the same trick.

The SEC sued three former executives of Tricolor Holdings on Tuesday over the subprime auto lender's $1.9bn collapse, alleging they raised money on faked documents and used the same vehicles as loan collateral multiple times — double pledging. "These defendants defrauded investors based on bogus collateral and violated the integrity of our private credit markets," said enforcement director David Woodcock (Bloomberg). On the same day, a 112-page filing in the First Brands bankruptcy revealed that Japan's Katsumi Global had tried to sell part of its factored-invoice exposure to First Brands to Apollo — and that Apollo declined "after correctly identifying the numerous red flags," then shorted the debt instead. An internal Apollo note from April 2024, eighteen months before the Chapter 11, had already written that "the cash flow of the business has never matched the Company's claims about its margins" (FT Alphaville). The two cases share a mechanism. In bank lending, collateral is registered, perfected and periodically re-verified by someone whose job depends on it. In the non-bank versions — warehouse lines against car loans, receivables factoring, invoice finance — the collateral is a document, and the check is usually a spreadsheet from the borrower. If the same car or the same invoice is pledged to two lenders, neither finds out until cash stops arriving. What is new here is not that fraud happens. It is that Apollo, the single largest buyer of this kind of paper, saw the problem, declined, and shorted — and the seller kept marketing the position to others. That is a market where diligence is done well by one participant and badly by several. It also tells you the loss is not evenly distributed: someone bought what Apollo refused.

The AI capex bubble18 Aughigh

The AI trade has split into two trades

The equity market is now discriminating between AI companies that fund capex with cash and those that fund it with project debt — which is the distinction that will matter if demand slows.

Look at five days of prices inside the AI complex. CoreWeave −15.6%. Nebius −13.6%, including −9.8% on Wednesday alone. Talen Energy −11.7%. Core Scientific −10.2%, and −20.8% over twenty days. Applied Digital −9.3%. Oracle, the most debt-funded of the hyperscalers, −6.2%. Now the other half. Nvidia is down 2.9% on the week and 3.4% off its high. Microsoft is up 24% in twenty days. Alphabet is flat. Amazon is up 8.6% in twenty days. The Nasdaq 100 is 2.8% below its June record. That is not a sector selling off. That is the market sorting the complex by balance sheet. The companies being sold are the ones that borrow at project level against contracted leases — neoclouds, converted bitcoin miners, merchant power — and the companies being bought are the ones that pay cash out of operating income. The pricing point that made the distinction concrete was last week's QTS "Project Odyssey" bond: $3.9bn, five-year, rated Baa3/BBB−, marketed at about 7.63%, a junk-like yield on an investment-grade rating for a Microsoft-linked data centre (Bloomberg). Curiously, the options market is not paying up for the risk. Implied volatility on the semiconductor ETF SMH collapsed from 65% last month to 40% on Monday, the lowest since February, as banks unwound hedges they had put on around the Situational Awareness blow-up. Into that cheapness, one trader bought $129m of November puts — over a third of the day's total premium in the fund (CNBC). Cheap protection and a large buyer of it is a combination worth watching.

Crypto and TradFi contagion18 Aughigh

The market bid up a company that is selling

The digital-asset treasury model has now been shown to run in reverse, and the stocks rallied anyway.

Strategy shares rose 12.6% on Wednesday, closing at the top of their recent range, while bitcoin rose 6.9% to $69,128. Here is what Strategy actually did over the past week: it sold $333.7m of common stock — not to buy bitcoin, but to repurchase preferred shares and add to a cash reserve. It has bought no bitcoin since the middle of June. Over the past five weeks it has sold about $2.1bn of stock, bought back about $347m of STRC preferred, and sold $213.3m of bitcoin. The reserve now stands at $4.8bn against roughly $59bn of coins (Bloomberg). The original machine worked like this: issue equity above the net asset value of the bitcoin you hold, buy more bitcoin with the proceeds, watch bitcoin-per-share rise, which justifies the premium, which lets you issue again. It is a flywheel that only spins while the stock trades above asset value. It is now running the other way. The company is issuing equity to retire fixed obligations and fund the coupons on its preferred stock. That is a deleveraging, and structurally it makes Strategy safer — the preferred dividends are the hard claim, and there is less of them. It also removes the largest marginal buyer bitcoin has had for six years. Bitcoin's 30-day realised volatility has fallen to a three-year low, which is what a market without its biggest price-insensitive bidder looks like (Bloomberg). The two ether treasury companies we track, BitMine and SharpLink, rose 10.7% and 12.2% on the day. Whether their holders have absorbed what Strategy just demonstrated — that the flywheel reverses — is not obvious from the tape.

Household credit19 Aughigh

Your lawsuit, sliced into bonds

Securitisation is reaching into collateral with no default history, and doing so with fewer disclosure requirements than in 2008.

Consumer legal funders advance cash to personal-injury plaintiffs to cover rent and medical bills while their cases grind through court. The advance is non-recourse: if the case fails, nothing is owed. If it settles — and insurers usually settle — the plaintiff repays with fees and interest averaging 35 to 45 per cent a year. The New York Times found one New York woman who received $76,500 in advances and owed at least $1.4m by the time her case settled. The new part is what happens next. The biggest funders are bundling thousands of these IOUs into asset-backed securities (NYT). Investors buy a claim on the future settlement stream; the funder recycles the cash into new advances. Think about what is being securitised. The collateral is not a car or a house or a credit-card balance with decades of default statistics behind it. It is the expected outcome of litigation — correlated to court backlogs, to insurer settlement policy, to a change in state law on funding disclosure. Price it and you are pricing a legal regime. There is no observable secondary market for a plaintiff's settlement expectation, so the marks come from a model. This lands in the same place as the SEC's decision this month to exempt data-centre securitisations from crisis-era risk-retention and disclosure rules (Telegraph). Both are cases of cash-flow streams with no default history being wrapped into securities that someone will hold at par. Neither is large enough to matter on its own. The pattern is the point: when an economy runs out of ordinary collateral, it invents new collateral, and the new collateral is always the least tested.

Fed, Treasury and policy19 Aughigh

$40tn, and the ceiling arrives a year early

Duration relief at the long end is being financed at the front end just as the Fed withdraws its front-end bid, and the debt ceiling now binds earlier than budgeted.

The US national debt passed $39.9tn on Monday and will cross $40tn this week — months ahead of schedule. Six months ago the CBO expected total borrowing to top out at $39.4tn this fiscal year. The gap is partly billions in lost revenue from tariffs that were subsequently invalidated (Washington Post). The consequence is a date. Congress set the debt limit at $41.1tn last year; budget analysts now think borrowing reaches it by early next year rather than late. Annual interest is projected to exceed $1tn, roughly the Pentagon budget. Set that against two other things that happened in the same week. Treasury announced it will at least double buybacks of 10-to-30-year bonds, taking purchases to roughly $32bn a quarter from 9 September; the 30-year fell 10bp to 5.18% and the dollar hit a three-month low (Bloomberg). And the New York Fed said on 13 August it would halt reserve management purchases entirely from 14 August to 14 September, having already tapered them to $10bn a month. So: the Treasury is retiring long bonds it must replace with shorter paper, while the Fed steps back from buying short paper. The net effect is more bills into a market with one fewer buyer, at a moment when the borrowing ceiling is approaching faster than anyone planned. Deutsche Bank's George Saravelos called the buyback "increasing administration unease." Gold rose 4.9% on the day, its largest move in this cycle, and the dollar index is down 2.3% in twenty days. The bond market got its relief. The currency paid for it.

Bond market dysfunctionhigh

The Treasury becomes a buyer of its own debt

The fiscal authority is now actively managing the price of its own long debt, which is what you do when you are not confident the market will clear it.

On Wednesday the US Treasury said it was "increasing, by at least double, the size of liquidity support buyback operations" for bonds maturing in 10 to 20 and 20 to 30 years. Repurchases of $2bn become "at least" $4bn. The 30-year yield fell about 9 basis points to 5.2%, the 10-year 7bp to 4.64%, and the dollar had its worst day in three months (FT, Bloomberg). What a buyback is: the Treasury takes cash and buys back bonds it has already issued. It is not QE — no money is created, the Fed is not involved, and the debt does not go away. It gets refinanced. And because the Treasury has a deficit to fund, the cash to buy the long bonds has to come from issuing something else. Market participants expect that something to be bills. "Given the amount of issuance required to fund the deficit, this implies greater issuance at the short end of the curve," said Daniel Murray of EFG International. So the mechanism is: shorten the average maturity of the national debt to relieve pressure at the long end. That works, and it has an obvious cost. Every dollar moved from a 30-year bond to a three-month bill is a dollar the government has to reprice every three months. The US spent 2020-21 not doing this, and is now paying for it. Doing the reverse buys relief today by buying more rollover risk tomorrow. The context is a fortnight of ugly auctions: $42bn of 10-years cleared at 4.683%, the highest since 2007; $25bn of 30-years at 5.216%, the highest since 2001. A $16bn 20-year sale is next.

The dollar, gold and reserve statusmedium

The adjustment moved to the currency

Managing the long end without fixing the deficit converts a bond problem into a currency problem, and the currency market has started to notice.

Gold rose 4.1% on Wednesday to $4,545 an ounce. Silver rose 3.5%. The dollar index fell 0.69% on the day and is down 2.15% over twenty days. This happened on the same afternoon that the Treasury announced it would double its long-bond buybacks, and it is the more interesting half of the story. The logic is worth spelling out. A government with a large deficit and a long-end problem has three places to put the adjustment: yields, the currency, or the real economy. If the fiscal authority steps in to keep yields down — and Bessent has said repeatedly that 10-year yields are his benchmark of success — the adjustment does not vanish. It relocates. Holders of long-dated dollar assets who wanted 5.33% to compensate them for inflation and fiscal risk, and are now offered 5.18% because the issuer is bidding, respond by holding fewer dollar assets. Gold is the cleanest expression of that: it has no yield to lose and no issuer to disappoint. We would not read one day as a verdict. But it is not one day. Gold is up 9.6% in twenty days and sitting at the top of its range; the euro and yen are both stronger; the broad dollar index is down 1.35% over the same period. The 30-year TIPS yield reached 3.09%, the highest since 2008 — that is the real rate, stripped of inflation, and it is the number that says investors want more compensation for lending long regardless of what happens to prices. A Treasury that has to defend the long end and a gold price at records are the same fact seen from two angles.

The AI capex bubblehigh

A $14bn data centre with $450m of cover

The catastrophic-loss risk on the largest AI projects is not being insured; it is being handed silently to bondholders.

Meta and BlackRock's one-gigawatt data centre campus in El Paso, Texas — internally called Sopaipilla, cost around $14bn — has bought insurance in the hundreds of millions of dollars, on advice from the broker Marsh. The FT has the schedule: up to $427mn of all-risk property cover during construction and $450mn once operating, for a premium of about $5mn a year rising 2% annually; $645mn against terrorism; $218mn for rent abatement if construction runs late; and general liability with a $50mn cap per event and $50mn in aggregate, for a $1mn premium. The project is not insured against a total loss. Such cover has become prohibitively expensive, the FT reports, because insurers are unwilling to build large aggregate exposures to single sites now routinely valued in the tens of billions. This matters because of who is holding the residual. Lenders to smaller data centres normally demand full insurance; as deal sizes grew, partial cover became acceptable. So the gap between $450mn of cover and $14bn of asset is not sitting with an insurer, priced and reserved against. It sits with whoever owns the debt. Meta priced $12.5bn of bonds for the El Paso project in July, at a wider yield than a comparable 2025 deal. That is the running Crash Lab thesis in one transaction. The tail risk on a single gigawatt-scale asset has been moved out of the insurance industry — which would have to mark and reserve it — into a bond that trades at par until it doesn't. Nobody has to write it down. Nobody has to hold capital against it.

Private credit and BDCshigh

Private credit said no to Sophos

The assumption that private credit will always refinance a sponsor's portfolio company is the load-bearing assumption of the whole asset class, and it just failed a $2bn test.

Thoma Bravo-backed Sophos needs to refinance or extend more than $2bn of loans, possibly next month. It spent months trying to get private credit funds to take the paper. Bloomberg reports those efforts "faltered," and the company is now going back to its existing leveraged-loan lenders — offering a higher coupon, amortisation payments and a tighter covenant package to get it done. Thoma Bravo has told investors it will not inject fresh capital, rebuffing a request from lenders worried about what AI does to a cybersecurity software business. Read the direction of travel carefully, because it is backwards. For five years the story has been that private credit is where syndicated loans go when the banks won't take them: a permanent, patient bid that doesn't require a market. Here the private credit bid did not show up, and a sponsor-owned borrower is going back to the public loan market with worse terms to get refinanced. That is not a default. It is repricing, and it is the healthy kind — lenders extracting amortisation and covenants is lenders doing their job. It is also the third Thoma Bravo software credit in three months to need this. The firm made major concessions in July to get a $5bn Proofpoint refinancing over the line, and in June creditors took control of Medallia outright. Set against the median non-accrual rate at the twenty largest listed BDCs hitting 2.8% of cost in Q2 — the highest since 2017 — the pattern is consistent: private credit is being more selective at the margin, in software specifically, and the market is not paying attention. BDC shares are up: ARCC +5.3% and FSK +10.9% over twenty days.

Hidden leverage and shadow bankingmedium

$830bn of borrowed money sits in the bond market

The largest single leveraged position in global finance is a bet on Treasury market functioning, at the moment Treasury market functioning is the question.

A piece of Fed research now puts the Treasury cash-futures basis trade at roughly $830bn as of September 2025 — about double its previous peak in early 2020, and around 35% of hedge funds' total long Treasury exposure. It is financed, in the researchers' description, almost entirely through repo. If you have not met the basis trade: Treasury futures usually trade slightly rich to the actual bonds, because asset managers want duration exposure without tying up cash. Hedge funds harvest that gap by buying the cash bond and shorting the future. The spread is tiny — a few basis points — so to make it a business you borrow, in the repo market, against the bond you just bought, and do it again. Leverage of fifty or a hundred to one is normal. The position is not a bet on rates going up or down. It is a bet that the two prices stay close together, and that repo funding stays available. Both of those held in March 2020 for about a week, and then didn't, and the Fed had to buy $1tn of Treasuries in a fortnight to stop it. The position is now twice as large, in a $29tn market, at a moment when the long end has just needed the Treasury Department to step in as a buyer. One mitigating point, and it is real: the growth of sponsored repo has made this funding cheaper and more stable than it was in 2020. The trade is also genuinely useful — it is a large part of why Treasury futures and cash prices track each other at all. This is a tier-2 summary of Fed research, not a Fed press release. Treat the exact figure as approximate.

Household credithigh

Subprime lenders' shares up, subprime borrowers not

Consumer credit stress is now visible in the data and invisible in the prices of the companies underwriting it.

The SEC sued three former executives of Tricolor Holdings on Tuesday over the subprime auto lender's $1.9bn collapse. The allegation is specific and worth knowing: the company raised more than $1.9bn while making false claims about its health, faking documents, and pledging the same vehicles as collateral for multiple loans — double pledging. "These defendants defrauded investors based on bogus collateral and violated the integrity of our private credit markets," said the SEC's enforcement director (Bloomberg). Two have pleaded guilty and are cooperating. Double pledging is the failure mode that makes asset-backed lending different from other lending. The whole structure rests on a lender being able to verify that a specific car secures a specific loan and no other. If nobody checks, the collateral is a spreadsheet entry. Meanwhile the underlying borrower is deteriorating. The New York Fed's Q2 data has overall delinquency at 4.7% of balances, seriously delinquent auto at 5.5%, and seriously delinquent student loans at 10.6%. Fitch put 60+ day delinquencies in subprime auto ABS at 5.67% in June. Pending home sales fell 2.3% in July to 71.2, the lowest since January and matching the second-worst reading in data back to 2001. Now look at the share prices. Oportun is up 37.8% over twenty days. Synchrony is up 10.9%, Capital One 10.35%, Affirm 5.7% in a week. Only the deep-subprime specialists — Credit Acceptance down 4.7%, Ally down 2.1% — are being marked down at all. Equity holders are being paid a spread that assumes losses stay where they are. The loss data says otherwise. One of the two is wrong.

The AI capex bubblehigh

An investment-grade bond at a junk price

The first sustained sign that the marginal lender to AI infrastructure is charging a risk premium, while the rest of credit stays asleep.

Blackstone-backed QTS Realty began marketing about $3.9bn of five-year notes to fund construction of a data centre tied to Microsoft. Initial price talk: a yield of roughly 7.63%. The notes are expected to be rated Baa3 by Moody's and BBB- by Fitch — investment grade. The yield is where junk bonds trade. The deal, dubbed Project Odyssey, was upsized by roughly $1bn after the high coupon pulled in about $10bn of indications of interest (Bloomberg). Hold that against the rest of the credit market. Broad high-yield spreads are 270 basis points over Treasuries; investment grade is 81. Neither has moved meaningfully in a month. So this is not credit repricing in general. It is investors demanding a specific premium for one thing: a single-tenant building full of chips, financed for five years, against a lease whose economics depend on demand for compute in 2031. The rating agencies are pricing the tenant. The market is pricing the building. That gap is the whole argument. A Microsoft-linked cash flow deserves a high-grade rating; a purpose-built facility with one customer and a technology cycle measured in quarters does not obviously deserve to be treated as real estate. Investors split the difference at 7.63%. Equities are making the same distinction more brutally. Over five days: CoreWeave -16.6%, Nebius -16.7%, Applied Digital -13.6%, Talen Energy -14.3%, Core Scientific -12.1%, Oracle -9.7%. Over twenty days Microsoft is up 23.8%. The market will fund the tenant. It is getting fussy about the landlord.

Hidden leverage and shadow bankingmedium

$1.65tn of debt in the footnotes

The measured debt of the AI buildout is now smaller than the unmeasured debt, which is the exact condition Crash Lab was set up to track.

Nikkei went through the filings of Alphabet, Microsoft, Amazon, Meta and Oracle and put their off-balance-sheet obligations at about $1.65tn — more than the roughly $1.35tn of debt these companies actually carry on their balance sheets. Four years ago the figure was roughly an eighth of that. Meta's share is about $420bn, nearly triple its recorded debt. Oracle's is $273.3bn as of the end of May, up more than thirtyfold in four years (Nikkei). The mechanism is dull and legitimate, which is exactly why it works. Under lease accounting, a data centre lease that has not yet commenced does not go on the balance sheet. Neither do long-term purchase commitments for GPUs and servers that have not been delivered. Both appear in the notes to the accounts. So a company can commit to two decades of payments on a facility being built by someone else, disclose it in a footnote, and show an unremarkable debt load until the day the lease starts. That is the design of the Meta–Blue Owl Hyperion structure: a $27bn bond raised in a special purpose vehicle against a 20-year Meta lease promise, and not on Meta's balance sheet. Goldman puts sector-wide leasing commitments at about $1.5tn, of which roughly $1tn is invisible in headline statements. What has to stay true: these leases pay for themselves out of AI revenue that mostly does not exist yet. If it doesn't, the obligation does not disappear — it arrives on the balance sheet on schedule, one commencement date at a time.

Private credit and BDCshigh

The Lakers stake and the $20bn of related-party loans

Insurance balance sheets are where private credit's marks go to avoid being tested, and a federal prosecutor is now testing them.

Mark Walter held talks with Apollo through the summer about a multibillion-dollar loan secured on his stake in the Los Angeles Lakers, the FT reports. The talks were overtaken by an offer from Bob Iger and Josh Kushner, and Walter is now selling the franchise instead, for $12.5bn (FT). Why a billionaire needed the cash is the interesting part. After receiving grand jury subpoenas from the Southern District of New York in February, insurers controlled by Walter's TWG Group disclosed more than $20bn of loans to related parties whose connections they had not disclosed. On Tuesday, Delaware Life said it had agreed to swap $6.5bn of assets linked to TWG affiliates for unrelated investments. After the swap, Delaware Life and another TWG insurer still hold more than $10bn of affiliated investments. S&P put Delaware Life on negative outlook last month after it restated its annual statements and reclassified a large volume of private credit as related-party. The plumbing: an annuity writer collects premiums from retail savers, and must invest them in assets a state insurance regulator will accept. If those assets are originated by an affiliate of the owner, the owner has effectively borrowed from his own policyholders — at a price he set, marked at a value he influenced, against liabilities that come due decades from now. This pairing of insurance float with affiliated private credit is not a Walter eccentricity. It is the industry template, adopted across a multi-trillion-dollar business. The subpoenas are about disclosure. The model is about who bears the loss.

Bond market dysfunctionhigh

Two bad auctions and a 30-year global repricing

Every valuation in the AI complex and every private credit mark discounts against a long rate that has now repriced globally, in an orderly way, without anyone forcing it.

The US Treasury sold $25bn of 30-year bonds at a stopping yield of 5.216%, the highest at auction since 2001. The day before, $42bn of 10-year notes cleared at 4.683%, the highest since 2007. By Tuesday the 30-year yield had reached 5.33%, a level last seen in June 2007, and the 30-year TIPS yield hit 3.09%, the highest since 2008 — so this is real yields and term premium moving together, not an inflation scare alone. It is not a US story. The 30-year Bund reached 3.78%, its highest since the eurozone crisis; the 30-year French yield 4.9%, highest since 2008; the 30-year gilt 5.86%. Japan's 10-year touched 2.93%, the highest since 1996, with Deutsche Bank's Shoki Omori noting that 3% is "a critical defence line for fiscal credibility" because it is the rate assumed in the government's own budget (FT). Robin Brooks points out that 10y10y forward yields rose almost everywhere last week despite soft US inflation data — France +14bp, UK +13bp, Japan +11bp. That is a risk premium, not a growth forecast. The crowding matters for our beat. Alphabet, Amazon, Meta and Microsoft are issuing ultra-long bonds to fund data centres into the same limited pool of duration buyers that governments need. Morgan Stanley expects $250–300bn of AI-related debt issuance this year. The oddity: our rate-volatility gauge is down 16% over five days and 31% over twenty. Yields at multi-decade highs, vol collapsing. Orderly repricing, then — until a leveraged holder of duration decides otherwise.

Private credit and BDCsmedium

Non-accruals at a nine-year high, BDCs up 6%

Non-accruals are the one private credit number that is hard to manage, and it is rising while the equity that owns the loans is near its highs.

Across the 20 largest listed US business development companies, the median share of loans on non-accrual reached 2.8% of cost in the second quarter, up from about 2.0% at the end of March. That is the highest since roughly 2017, when the oil crash was working through energy borrowers. The data is Solve's, summarised by the FT and picked up in our web sweep. A non-accrual is a straightforward admission: the lender has stopped booking interest income on a loan because it no longer expects to be paid. In a BDC, which distributes almost all its income as dividends, that flows directly into dividend coverage. The individual disclosures show what it looks like up close. Kayne Anderson's BDC moved 4over and the last-out tranche of Diverzify to non-accrual, taking its rate to 2.7%; its first-lien loan to American Soccer Co. is marked 44% below cost. Now look at the prices. Over twenty days: Ares Capital +5.7%, Blackstone Secured Lending +5.9%, Blue Owl Capital Corp +6.5%, FS KKR +10.9%. The largest BDCs trade within 1-5% of their recent highs. The managers have done better still — Blue Owl Asset Management +30.8% in twenty days, Ares Management +19.3%, Blackstone +19.2%. That is the disagreement worth naming. The loan data says credit quality at the bottom of the middle market is the worst in nine years. The equity market says the fee streams are fine, because managers earn on assets under management, not on outcomes. Both can be true for a while. Only one of them is about the loans.

Household credithigh

Personal injury lawsuits, now available as a bond

Securitisation is reaching collateral with no loss history at exactly the point where the mainstream consumer books are running at post-2008 stress levels.

The New York Times has documented what consumer legal funders do with their loan books. These firms advance cash to personal injury plaintiffs — rent, food, medical bills — while cases grind through court. If the case fails, the plaintiff owes nothing. If it settles, which most do, the plaintiff repays with fees and interest averaging 35 to 45% a year. One New York woman who said she turned her ankle on a sidewalk received $76,500 in advances and owed at least $1.4mn by the time her case settled. The funders are now bundling thousands of these IOUs into asset-backed securities (NYT). Investors buy a claim on future settlement payouts; the funder gets its capital back immediately and lends it out again. Songwriting royalties, fast-food franchises and data centres have already been securitised this year. Litigation advances are simply the next collateral pool. The reason this is worth your attention is not the interest rate. It is what it says about the search for yield reaching collateral whose loss rates nobody has a cycle of history for. A pool of these advances has an unknowable recovery rate, because it depends on the settlement behaviour of insurance companies and the pace of court calendars. Meanwhile the ordinary consumer books are already strained. Credit card balances hit $1.263tn in Q2, and the flow of balances into 90-day delinquency was 6.97% — the tenth consecutive quarter above the level that marked the onset of the Great Recession. Subprime auto ABS 60-plus-day delinquencies were 5.67% in June. Pending home sales fell 2.3% in July, the second-worst reading since 2001.

What would change our mind

The specific, observable things that would move the number - in either direction.

SOFR printing meaningfully above the Fed's administered rates in the second week of September, when buybacks start with reserve-management purchases still at zero — that would turn a policy contradiction into a funding event.

Would move the number

High-yield OAS above 350bp, or a data-centre bond or ABS deal pulled from the market after the QTS deal cleared at 7.63%.

Would move the number

The 30-year Treasury back above 5.33% after buybacks actually begin — proof the buyer was too small to matter, which would be worse than not trying.

Would move the number

Gold and the dollar rallying together, or a BDC cutting its dividend or restating NAV — the first would tell me the debasement bid was tactical, the second that the non-accrual data has reached investor cash flows.

Would move the number