Microsoft-linked data centre bond marketed at 7.63%, a junk price

Archived reading, published Wed, 19 Aug 2026 14:22:18 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

+1 since the last reading

Fragility89

how much tinder is stacked up — moves slowly

Ignition45

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 26 sources and rewrites this page.

Why it moved: Up one, 66 to 67. The new thing this window is a price: a Baa3/BBB- rated, Microsoft-linked data centre bond marketed at about 7.63% while broad high yield sits at 270bp, alongside the debt-funded end of the AI complex falling 12-17% in five days. Nothing broke at the funding level, equities are 1% off a record and rate volatility keeps falling, so it is one point and not five.

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.

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.

Crypto and TradFi contagionhigh

The flywheel, running backwards

A structurally pro-cyclical seller has replaced the market's largest structural buyer, and the equity has not fully accepted it.

Strategy sold $333.7mn of common stock last week. It did not buy any bitcoin with the proceeds. It bought back preferred shares and added to a cash reserve, now $4.8bn. Michael Saylor's company has not bought bitcoin since the middle of June. Over the past five weeks it has sold about $2.1bn of common stock, repurchased roughly $347mn of its STRC preferred, and sold $213.3mn of bitcoin (Bloomberg). Understand what the machine was. Strategy issued equity and preferred stock at a premium to the value of the coins it held, used the cash to buy more coins, and the resulting increase in coins per share justified the premium. It only worked in one direction. When the premium compressed, issuance stopped being accretive, and the preferred dividends — which are cash obligations — kept coming. So the company now sells stock to buy back its own preferred, and sells bitcoin to pay interest. It still owns about $59bn of bitcoin. It is no longer a buyer. As Caroline Mauron of Orbit Markets put it: "Strategy's trading is pro-cyclical. They are able to buy more when it goes up, and they are forced to sell when it goes down." For six years Strategy was the largest marginal buyer of bitcoin. That bid is gone, and bitcoin's 30-day realised volatility has fallen to a three-year low. The coin is $65,333, up 0.9% over twenty days. The stock is down about 73% over a year against bitcoin's 45%. Note what the equity did on the day: MSTR +6.0%, and it sits 3.8% below its recent high. Somebody still likes the capital-management story.

Signs of the times

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

Eric Trump's American Bitcoin told an earnings call it was "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. Forbes describes the model as selling shares at a Trump-name valuation, buying bitcoin with the proceeds, and citing the larger bitcoin pile as proof the plan works.

Forbes (via Capital.gr)

Erebor Bank, founded by Palmer Luckey and backed by Peter Thiel, was chartered only this year and is already raising capital at an $8bn valuation, nearly double the $4.35bn it fetched late last year. Its stated plan is to serve the concentrated tech and defence client base that sank Silicon Valley Bank, with blockchain rails so clients can move their money at any hour.

Bloomberg

A New York woman who said she turned her ankle on a sidewalk took $76,500 of cash advances from a consumer legal funder and owed at least $1.4mn by the time her case settled. Those advances are now being bundled into asset-backed securities.

The New York Times

Leopold Aschenbrenner, who is 24 and enrolled at Columbia aged 15, turned an essay about AI into Situational Awareness LP, a fund managing tens of billions, which imploded days before his wedding; Jane Street blamed the firm by name for losing billions of dollars in July. The Wall Street Journal reports a rescue deal was scrapped at the last minute for a better offer, with investors getting the call from Aschenbrenner himself.

New York Magazine

The SEC has exempted some data centre owners from crisis-era ABS disclosure and risk-retention rules, on the grounds that data centres are not "financial assets" like mortgages or car loans. The change was prompted by a request from Latham & Watkins, a major AI adviser. Jensen Huang routinely calls data centres an "investable asset class".

The Telegraph

The rumour mill

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

Confirmed

“A business financially connected to the Trump family works with a platform distributing Chinese AI models, including from companies the US government has restricted, and Trump family crypto products earn revenue from it.”

Reuters reports that World Liberty Financial, 38% owned by the Trump family, is collaborating with Hong Kong-based WorldClaw, which offers around 43 of 90 models from Chinese firms flagged by the Pentagon, and accepts World Liberty's USD1 stablecoin as payment. Separately, the OCC has granted preliminary trust bank approval to a World Liberty subsidiary, letting it issue USD1 — now the fourth-largest stablecoin at about $4bn — directly.

Claimed by Gregory Mannarino

Confirmed

“Tokyo began intervening to prop up the yen on 30 July and the United States joined the next day — the first US participation in yen support since 1998.”

Japan's finance ministry confirmed coordinated yen buying with the US on 31 July; estimates of the two days' intervention run to roughly $85–88bn. The yen has since given back about half its gains and trades near ¥159, which tells you what coordinated intervention buys you these days.

Claimed by Mark Moss

Partly true

“Free cash flow at the four largest hyperscalers turns broadly negative by 2027 and is already negative for several, forcing them to fund data centres with debt.”

LSEG consensus has Microsoft, Alphabet, Amazon, Meta and Oracle spending more on capex than they generate in free cash flow by 2027, and Bank of America models an aggregate swing from +$180bn in 2025 to -$144bn in 2027 across a broader AI group. But it is not already negative for all four — Alphabet and Microsoft are the exceptions — and not all are debt-funding the buildout.

Claimed by Meet Kevin

Partly true

“TLT has hit a new low for the year and is down 50% from its 2020 peak, with the 10-year at 4.7% and the 30-year at 5.27% marking the highest weekly closes of the year.”

The TLT drawdown is real and documented — more than 50% off the 2020 record, lowest since 2004. The specific weekly closing levels are not corroborated; the yields cited are close to daily prints, and TLT actually rose 1.5% on Wednesday. The long bond is having a historically bad decade, not a dramatic week.

Claimed by Peter Schiff

Earlier developments

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

Private credit and BDCshigh

Nobody wants to refinance Sophos

When the buyer of last resort for LBO debt steps back and the sponsor refuses to add equity, losses get resolved through terms rather than defaults — which is why nothing shows up in spreads until it shows up all at once.

Thoma Bravo's cybersecurity company Sophos is going back to its existing lenders to refinance or extend more than $2bn of loans as soon as next month, after an attempt to line up private credit backing faltered, Bloomberg reports. To get it done, the company may have to offer a higher coupon, amortisation payments and a tighter covenant package. Thoma Bravo, which bought Sophos in 2020, has indicated it will not inject fresh capital — rebuffing a request from investors worried about what AI does to Sophos's business. That last sentence is the mechanism. In a leveraged buyout the sponsor's equity is the cushion under the debt, but nothing obliges the sponsor to top it up. When a sponsor says no, the lenders' choice is between worse terms on the same borrower and a fight. Amortisation and covenants are how lenders take payment for that: the risk gets repriced without a default, without a downgrade, and without showing up in any credit index. This is the third instance in three months from the same sponsor. Blackstone-led creditors took control of Thoma Bravo's Medallia in June. In July, Thoma Bravo had to sweeten a $5bn Proofpoint refinancing after a lender revolt. Now Sophos. Software cash flows are the single largest collateral pool in direct lending, and the thing eating them is the same thing the whole market is long. Median non-accruals at the 20 largest listed BDCs are 2.8% of cost. High-yield spreads are 270 basis points and unchanged on the week. Private credit manager shares fell 3–5% over five days — Blackstone, Apollo, Blue Owl, KKR — while still up 10–25% over a month. The equity is nervous. The credit market is not.

The AI capex bubblehigh

A $14bn data centre, $450m of insurance

The catastrophe risk on the largest new asset class in credit cannot be laid off to insurers, so it stays with bondholders — a form of leverage that appears in no leverage statistic.

Meta and BlackRock's $14bn data centre campus in El Paso, Texas — known as Sopaipilla — has bought a few hundred million dollars of insurance, and is not insured against total loss. On the advice of Marsh, the world's biggest insurance broker, the one-gigawatt site has 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 of rent-abatement cover if construction runs late; and general liability with a $50mn per-event and $50mn aggregate limit for roughly $1mn a year. The FT reported the terms; Marsh, Meta and BlackRock declined to comment. Here is why lenders should care. Data centre debt is increasingly project debt: money lent against one building with one tenant, repaid out of lease payments. The standard protection for such a lender is full insurance, so a fire, a storm or a lawsuit does not wipe out the collateral. The FT notes lenders to smaller data centres still demand that. At this scale, insurers will not build the aggregate exposure — a single campus can hold tens of billions of dollars of chips — so partial cover has become acceptable, and everything above the cap sits with the borrower, which in practice means the creditors. What we do not know: how the lenders to Sopaipilla were compensated for that gap, or whether their documents require cover they cannot buy. What we do know is the direction. The insurance market has decided this risk is too concentrated to underwrite at any sensible price. The bond market has not yet reached the same conclusion.

Fed, Treasury and policymedium

A data centre is not a financial asset

The fastest-growing collateral in structured credit is being exempted from the rules written after the last time structured credit went wrong.

The SEC has exempted some data centre owners from crisis-era rules on asset-backed securities, removing certain disclosure and risk-retention requirements, according to the Telegraph. The change was prompted by a request from Latham & Watkins, a major adviser to the AI sector, and rests on the reasoning that data centres are not "financial assets" like mortgages or car loans. This is one outlet's account of the decision; the underlying documents have not been widely reported elsewhere, so treat the framing carefully. Risk retention is the post-2008 rule that the sponsor of a securitisation keeps roughly 5% of it, so someone who packages loans still owns some of the loss. Disclosure rules are what let a buyer see the collateral. Removing both for a new collateral type is a bet on price discipline instead of rules. Professor Raghavendra Rau of Cambridge Judge told the paper the SEC's position amounts to: if you think a data centre has too concentrated a tenant base or will go obsolete, price the bonds accordingly. So, is it being priced accordingly? Partly. On Tuesday, Blackstone-backed QTS began marketing about $3.9bn of five-year notes to fund a Microsoft-linked project — rated Baa3 by Moody's and BBB– by Fitch, investment grade — at initial talk of roughly 7.63%, which is a junk yield. Investors put in about $10bn of interest at that level, and the deal grew by $1bn. That is discipline of a sort: buyers are demanding a two-notch premium over the rating. It is not the same as knowing what is in the pool. JPMorgan sees data centre securitisation reaching $30bn–$40bn a year, from about $27bn in 2025; Morgan Stanley expects $250bn–$300bn of total AI debt issuance this year.

Bond market dysfunctionhigh

$194bn of losses Japan's insurers don't have to show

The largest patient buyers of long-dated government bonds are sitting on losses that only stay invisible if they never have to sell.

Unrealised losses on domestic bonds at Japan's major life insurers reached ¥30.86tn ($194bn) at the end of June, up 60% year on year, Nikkei found from 13 of the 14 biggest firms. The figure now exceeds their unrealised gains on Japanese stocks, ¥30.03tn. The cause is arithmetic: 30-year JGB yields were in the 3.9% range at end-June, about 2.7 percentage points higher than three years ago. Held to maturity, these losses vanish. That is the honest version of the story and it is mostly right. But there are two ways they stop being theoretical. First, impairment: if a bond's market value falls 50% below acquisition cost, insurers may have to write it down. Some ultra-long bonds bought in the late-2010s at near-zero yields have now fallen that far. Nippon Life booked ¥44bn of impairments in April–June; Meiji Yasuda ¥25.3bn. Second, duration: life insurers try to match asset and liability maturities. Where assets are longer than liabilities, rising rates cut net assets, and the fix is to sell long bonds — turning unrealised losses into realised ones. Policyholder behaviour makes this worse, and Sony Life's lapse and surrender rate rose 0.2 points to 1.4%, partly because yen weakness prompted cancellations of foreign-currency policies. Japan's life insurers are the natural buyer of the world's longest-dated government debt. The 10-year JGB is at 2.93%, the highest since 1996, and the 30-year around 4.06%. If the biggest structural holder of duration becomes a seller of duration, the global long end loses a bid at precisely the moment the US is selling 30-years at 5.22% and Britain's are near 5.86%. The reassuring detail this week: MOVE, the measure of Treasury rate volatility, is down 27% in twenty days. The repricing is severe and so far orderly.

Hidden leverage and shadow bankingmedium

The market maker that lost $15bn in a month

The firms that absorb everyone else's risk are now running hedge-fund-sized directional books, and nobody supervises their risk appetite.

Jane Street lost about $15bn in July — its first losing month in a decade — and the loss has forced a re-description of what the firm is. Bloomberg reports that in its most recent debt offering, Jane Street described its strategy as having evolved to include longer-term bets, in line with a hedge fund. The firm has still made more than $40bn of net trading revenue this year. The numbers in circulation do not fully agree, and the difference matters. Bloomberg puts the firm-level July hit at $15bn. The FT attributes a $35bn July loss to one trader, Leopold Aschenbrenner, on leveraged AI stock bets — a figure that would top the all-time leaderboard of trading losses. Those are presumably gross-versus-net or book-versus-firm distinctions, but we cannot reconcile them from what has been published. Why this belongs in a crash meter rather than a gossip column: a bank cannot lose $15bn in a month on directional trades, because capital rules and supervisors make the position impossible to hold. A non-bank can, and this non-bank is one of the largest providers of liquidity in equities, options and ETFs. The question is not whether Jane Street survived — it plainly did. It is whether a firm that has just taken a loss of that size on its own book is willing to warehouse risk for everybody else in a bad week. Market-making capacity is not a fixed quantity; it is a decision made each morning by people looking at their own P&L. Separately, a summary of Federal Reserve research circulating this week puts the Treasury cash-futures basis trade at roughly $830bn as of September 2025 — about double the early-2020 peak, 35% of hedge funds' long Treasury exposure, financed almost entirely in repo. That is a second-hand figure and dated, but it is the same sector.

Household credithigh

Bogus collateral, twice pledged

Subprime consumer credit is now funded largely outside banks, where the collateral behind the paper is verified by whoever is selling it.

The SEC sued three former executives of Tricolor Holdings on Tuesday, including founder Daniel Chu, over what it calls a multi-year scheme at the subprime auto lender that collapsed owing $1.9bn. The company raised more than $1.9bn while, the SEC alleges, executives faked documents and pledged the same vehicles as collateral for multiple loans — double pledging. Two of the three, the former CFO and a former senior finance director, have pleaded guilty and are cooperating. The SEC's enforcement director said the defendants "defrauded investors based on bogus collateral and violated the integrity of our private credit markets." Bloomberg has the filing. Double pledging is the specific failure worth understanding. Asset-backed lending works because a discrete thing — a car, an invoice — sits behind each loan, and somebody checks. When nobody checks, the same collateral supports several claims and the loss on discovery is a multiple of the loss you thought you had. It is the same mechanic alleged at First Brands, where invoices were factored more than once. Two large private-credit blow-ups in a year turned on collateral that was not there. Meanwhile the borrowers themselves are grinding. New York Fed data for Q2 put credit card balances at $1.263tn, with 12.92% of balances 90 days or more delinquent and the inflow into serious delinquency at 6.97% — a tenth consecutive quarter above the level that marked the onset of the Great Recession. Seriously delinquent student loans are at 10.6%. Subprime auto ABS 60-day delinquencies were 5.67% in June. Pending home sales fell 2.3% in July to the second-worst reading in data back to 2001. And the lenders' shares? Capital One is up 7.8% in twenty days and 2.2% off its high; Synchrony up 12% and effectively at its high. The market is reading this as normalisation, not deterioration. One of the two views is wrong.

Crypto and TradFi contagionhigh

The coin is flat, the equities are not

The crypto treasury structure has flipped from a demand engine into a mechanical seller, and it is happening in public equities rather than in leveraged bank exposure.

Bitcoin is at $64,333, down 0.6% over twenty days. In the same twenty days Coinbase fell 16.9%, Galaxy Digital fell 19.6% and Strategy fell 9.2%. Over a year, Strategy's stock is down about 73% while bitcoin is down 45%. When the wrapper falls three times as far as the thing inside it, the wrapper is the problem. The mechanism is on display in Strategy's weekly filings. It sold $333.7m of common stock last week not to buy bitcoin but to repurchase preferred shares and build cash. Over five weeks it has sold roughly $2.1bn of common, bought back about $347m of its STRC preferred, and sold $213.3m of bitcoin. It has bought no bitcoin since mid-June, and holds a $4.8bn cash reserve against roughly $59bn of coins. Bloomberg has the numbers. A digital-asset treasury company only makes sense if it can issue equity above the value of the coins it holds and buy more coins with the proceeds. Below that line, the machine runs the other way: the equity is diluted to service preferred dividends, and coins are sold to raise cash. "Strategy's trading is pro-cyclical," Caroline Mauron of Orbit Markets told Bloomberg. "They are able to buy more when it goes up, and they are forced to sell when it goes down." The consequence for everyone else is that the largest marginal buyer of bitcoin for six years has left the bid. Bitcoin's 30-day realised volatility is at a three-year low, which is not calm so much as absence. Contagion into TradFi remains limited — the losses are sitting in listed equity, which is where you want them. That is genuinely the good version of this story.

Private credit and BDCshigh

Two point eight per cent

Non-accruals are the earliest hard number in a market that otherwise reports its own valuations, and they moved a lot in one quarter.

The median share of loans on non-accrual at the 20 largest publicly traded business development companies hit 2.8% of cost in the second quarter, up from 2.0% at the end of March — the highest since 2017, when the industry was digesting the oil crash. That is the FT's analysis of data from Solve. (FT) Non-accrual is a specific thing, not a mood. It means the fund has stopped booking interest income on a loan because the borrower has stopped paying, or is expected to. It flows straight through to net investment income, which is what BDCs pay dividends out of. An 80 basis point move in the median in one quarter is fast. FS KKR reported 7.1% of its book troubled. David Golub of Golub Capital told investors there was "elevated credit stress" and added: "We're in a credit cycle. Others denied it for a while. I don't think there's a lot of denial any more." Fitch says private credit defaults hit a record in July. PitchBook LCD data show the biggest listed BDCs shrank again — repayments and sales outran new commitments at vehicles run by KKR, Blue Owl and Apollo's MidCap Financial. Now look at the prices. Ares Capital is up 3.7% over 20 days, Blackstone Secured Lending 3.8%, FS KKR 8.2%. Blue Owl Capital Corp trades 4.2% off its high. High-yield spreads are 270bp, unchanged on the week. So the reporting says a credit cycle has arrived and the market says it hasn't. The reason both can be true for a while: private loans are marked by the manager, not by a screen. A non-accrual is a disclosure event. A markdown is a decision.

The AI capex bubblehigh

Nvidia guarantees the rent

The financeability of the largest AI buildout now rests on one chipmaker's willingness to stand behind its customers' rent.

Nvidia has agreed to guarantee up to $105bn of OpenAI's lease payments on a data-centre campus in Pike County, Ohio being built by SoftBank-backed SB Energy. It is also putting $1.5bn of equity into SB Energy. OpenAI is contracting for roughly 8 gigawatts of capacity; the first 800MW is due in 2028. Nvidia's backing phases in from 2028 to 2030 and runs 20 years, and the $105bn covers only the first 4.25GW. It may take on the second phase too. (Bloomberg) Here is what the guarantee does. Whoever lends against that campus is lending against a stream of lease payments. Without the guarantee, that stream is OpenAI credit — a company with enormous revenue growth and no rating anyone would lend $100bn against for twenty years. With it, the stream is substantially Nvidia credit. Same building, same tenant, radically cheaper money. Nvidia's own numbers make the loop visible: OpenAI's chip purchases could represent about $600bn of Nvidia revenue through 2030, and Nvidia has separately signed MOUs with Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR to mobilise more than $500bn of third-party compute financing, backstopping up to $125bn of it itself. Jensen Huang says this is not circular financing. Nvidia stresses it is not guaranteeing the whole site or all of OpenAI's obligations. What we do not know is the accounting. A guarantee of this type would normally sit as a contingent liability in the footnotes until a payment is missed. We have not seen the trigger, the coverage waterfall, or what Nvidia would recover if it paid. Bond desks are already trying to size roughly $70bn of what Bloomberg calls "shadow AI backstops." This adds a number an order of magnitude larger.

Hidden leverage and shadow bankinghigh

$1.65 trillion in the footnotes

The largest single addition of corporate leverage in a decade is being recorded in places that leverage ratios do not capture.

Nikkei went through the filings of Alphabet, Microsoft, Amazon, Meta and Oracle and totted up the obligations that do not appear on their balance sheets. The figure is $1.65tn — more than the $1.35tn of debt that does appear, and roughly eight times the level of four years ago. Meta's off-balance-sheet total is about $420bn, nearly triple its recorded debt. Oracle's is $273.3bn as of end-May, a more-than-thirtyfold increase in four years. (Nikkei) The mechanism is boring and that is the point. Under lease accounting, an operating lease becomes a balance-sheet liability when the lease commences. A signed twenty-year lease on a data centre that is still a field in Ohio is a commitment disclosed in a footnote. GPUs and servers under long-term purchase contracts that have not yet been delivered are the same. All of it is perfectly legitimate. So the period of maximum commitment — when the contracts are signed, the sites are being poured, and nothing has been delivered — is also the period of minimum balance-sheet disclosure. Leverage ratios during the buildout look like the leverage ratios of the old capital-light software businesses. They are not. The structures reinforce it. Meta's $27bn Hyperion financing sits in a special purpose vehicle backed by a twenty-year Meta lease promise, keeping the debt off Meta's books. Goldman estimates roughly $1.5tn of comparable leasing commitments across the sector, about $1tn of it invisible in financial statements. (Telegraph) Meta's shares are down 15.5% in twenty days. That may be the market doing the footnote arithmetic on its own.

Bond market dysfunctionhigh

Every long end at once

Term premium is rising everywhere at once, and the marginal holder of US duration is a repo-financed relative-value trade rather than a real-money investor.

The 30-year US Treasury yield touched 5.33% on Tuesday, the highest since 2007, having been under 5% at the start of last month. It did not happen alone. The 30-year Bund hit 3.78%, its highest since 2011. French 30-year yields reached 4.9%, the highest since 2008. UK 30-year gilts hit 5.86%. Japanese 30-year yields went to 4.16%, close to a record, and the 10-year JGB to 2.93% — its highest since 1996. (FT) Auctions are where you see whether this is repricing or indigestion. A $25bn 30-year sale cleared at 5.216%, the highest auction yield since 2001. A $42bn 10-year cleared at 4.683%, the highest since 2007. Both tailed. Three things are pushing. Brent above $90 with the Hormuz truce expiring, so an inflation premium. US debt near $40tn with no appetite to address it, so a fiscal premium. And, newly, AI. Investment-grade issuers have sold nearly $1.5tn of bonds this year, up 36%. Nomura reckons the roughly $200bn borrowed by the biggest tech companies alone equals about 25% of the Treasury's net issuance of notes and bonds to private investors — five times its share in 2025. (Bloomberg) What sits underneath is the part that worries us. Federal Reserve research puts the cash-futures Treasury basis trade at roughly $830bn as of September 2025 — about double the early-2020 peak, and around 35% of hedge funds' long Treasury exposure, financed almost entirely in repo. That is not a buyer of duration. It is a levered arbitrage that requires repo to stay cheap and margin to stay stable. The last time it broke, in March 2020, the Fed bought $1trn of Treasuries in three weeks.

The AI capex bubblemedium

The power trade cracks, the credit doesn't

The equity market has started to discount the AI buildout while the credit market is still funding it at investment-grade terms.

Tuesday was ugly for everything that sells shovels to the AI buildout. Talen Energy fell 10.96% in a day and is down 11.3% on the week and 20.6% from its high. Vistra fell 3.85% and is down 13.4% over twenty days. CoreWeave fell 12.07%. Applied Digital fell 8.48%. Core Scientific fell 6.4% and is 21.5% lower over twenty days. IREN fell 6.5%. Nebius fell 7.65% — after rising 28.7% in the preceding five sessions. The Nasdaq fell 1.33% and sits 3% below its June high. So this is not a market selloff. It is a specific repricing of the companies whose value is a levered claim on the buildout proceeding at the assumed pace: the merchant power generators signing long-dated supply contracts, the ex-bitcoin miners converting sheds into GPU halls, the neoclouds borrowing against chips. Now the other side of the ledger. High-yield spreads are 270bp, flat on the week. Investment grade is 81bp, three basis points wider over twenty days. VIX is 15.84. Blackstone-backed QTS priced a $3.9bn data-centre bond tied to Microsoft at about 7.63% — a junk-like yield on a Baa3/BBB- rating — and drew $10bn of indications, letting it upsize the deal by $1bn. (Bloomberg) Our reading, and it is inference: equity holders reprice every afternoon, and credit holders do not have to reprice until a coupon is missed or a lease is broken, which on twenty-year data-centre paper is years away. Equity is the fast instrument. That does not make it right. But when the two disagree this sharply about the same cash flows, one of them is going to move a long way.

Crypto and TradFi contagionhigh

The flywheel spins backwards

A levered vehicle whose only business is holding an asset has become a structural seller of it, which is how these things end rather than how they begin.

Strategy sold $333.7m of common stock last week — not to buy bitcoin, but to repurchase preferred shares and build a cash reserve. Over five weeks it has sold about $2.1bn of common, bought back about $347m of its STRC preferred, and sold $213.3m of bitcoin. It has not bought a coin since mid-June. The cash reserve is now $4.8bn against roughly $59bn of bitcoin. (Bloomberg) The original machine worked like this. Issue equity above the market value of the bitcoin per share; use the proceeds to buy bitcoin; bitcoin per share rises; the premium is justified; issue more equity. It is a perpetual motion device that runs on the premium and nothing else. Remove the premium and every arrow reverses. Issuing equity below NAV now destroys bitcoin per share, so the sensible thing to do with the proceeds is retire the liability that has a hard cash coupon — the preferred — rather than buy more of the asset that has none. When the equity market won't take the paper, the coins go instead. That is a deleveraging, executed in public, one Monday filing at a time. The prices tell the same story more precisely than either number alone. Bitcoin is at $64,124, down 0.9% over twenty days — essentially flat. Strategy is down 9.2% over the same period and fell 5.3% on Tuesday. Over a year the stock is off about 73% while bitcoin is off 45%. The holding company is being marked down relative to the thing it holds. The wider point: the largest marginal buyer of bitcoin for six years is now, in modest size, a seller. Bitcoin's 30-day realised volatility has fallen to a three-year low. Nothing here has broken. Something here has stopped.

Private credit and BDCshigh

The lender of last resort said no

Private credit's function in the system is to absorb debt nobody else will hold; when it starts declining, the marginal borrower has nowhere left to go.

Sophos, the cybersecurity firm Thoma Bravo bought in 2020, needs to refinance or extend more than $2bn of loans as soon as next month. It spent months trying to line up private credit backing. That effort faltered, and the company is now going back to its existing leveraged-loan holders, potentially offering a higher coupon, amortisation payments and a tighter covenant package to get them to stay (Bloomberg). Understand what that inverts. The entire pitch of private credit is that it is the buyer of last resort for corporate debt: it will lend where the syndicated market won't, on terms the syndicated market won't write, at a price. When a $2bn software LBO shops itself to private credit and comes back empty, the order of operations has reversed. The paper is going back to the public loan market — the one place it gets marked every day. Two further details matter. First, Thoma Bravo has told lenders it will not inject fresh capital, rebuffing a request from investors. The sponsor is declining to defend its equity, which means any loss stays with the creditors. Second, the stated worry is AI: lenders are now being asked to underwrite whether a security software product survives a technology shift, over a loan that outlives the forecast. This follows a pattern. In July, Thoma Bravo had to sweeten a $5bn Proofpoint refinancing after a lender revolt. In June, Blackstone-led creditors took control of its survey company Medallia. One firm's portfolio is not a cycle. But this is what a credit cycle looks like at the beginning: not defaults, refusals.

The AI capex bubblehigh

A $14bn building with $427m of cover

The credit quality of AI data-centre debt rests entirely on the tenant's lease, and the insurance market has now declined to stand behind the building itself.

Meta and BlackRock's Sopaipilla campus in El Paso is a one-gigawatt data centre costing about $14bn. On the advice of Marsh, the world's largest insurance broker, the venture has bought up to $427m of all-risk property cover during construction and $450m once it is running, rising 2% a year, for a premium of roughly $5m. Terrorism cover runs to $645m. Rent-abatement cover for construction delay is $218m. General liability is $50m per event and $50m in aggregate, for about $1m a year. The project is not insured against a total loss (FT). So roughly 3% of the asset value is covered. Lenders to smaller data centres normally demand full insurance; as deal sizes have grown, partial coverage has become acceptable, because the insurance market simply will not build that much aggregate exposure to a single site. That is a capacity constraint, not carelessness. But the constraint has to land somewhere, and where it lands is the debt. What we do not know is how the lease treats a casualty. If Meta keeps paying rent on a destroyed building, the bondholders are fine and the equity absorbs it. If the lease abates on total loss, the credit evaporates with the concrete. That single clause determines whether this is a footnote or a hole, and it is not public. This is the shape of the whole AI financing complex in miniature: the tenant's covenant is doing all the work, and the physical asset underneath it is thinly protected. Insurance is one of the few places where a genuinely independent party has to price the tail. It looked at this and said: not for $14bn, not in one spot.

Fed, Treasury and policymedium

Skin in the game, waived

The AI buildout is being financed through exactly the channels — SPVs, vendor guarantees, ABS without retention — that were designed after 2008 to be harder to use.

The SEC has exempted some data-centre owners from crisis-era rules on asset-backed securities, removing certain disclosure and risk-retention obligations. The change came after Latham & Watkins, a major adviser to the AI sector, asked the regulator to clarify whether data centres should be regulated as tightly as other ABS. The SEC's answer: data centres are not "financial assets" like mortgages and car loans (Telegraph). Risk retention is the Dodd-Frank rule requiring whoever assembles a securitisation to keep a slice of it, so that the person choosing the collateral eats some of the loss. It exists because in 2006 nobody who picked the loans kept any. Removing it from a market JPMorgan expects to run at $30–40bn a year, up from about $27bn in 2025, is not by itself dangerous — the market is small next to Morgan Stanley's $250–300bn estimate for total AI debt issuance in 2026. What it signals is the regulatory posture heading into the peak of the buildout. Cambridge's Raghavendra Rau put it plainly: "The SEC has been replacing regulatory constraints with a reliance on market discipline." Stack it with the rest of the week. Nvidia has signed memoranda with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to build compute-financing platforms targeting more than $500bn of third-party capital, with Jensen Huang saying Nvidia could backstop up to $125bn, or a quarter of potential deals (CNBC). Meta's $27bn Hyperion financing sits in an SPV, off its balance sheet, behind a 20-year lease promise. Every one of these structures does the same thing: it moves the exposure to a place where no regulated entity has to hold capital against it, and no quarterly mark reveals it.

Hidden leverage and shadow bankingmedium

The market maker was a hedge fund

Market-making capacity and directional risk have merged inside firms that no regulator capitalises, which means liquidity disappears exactly when it is needed.

Jane Street lost $15bn in July — its first losing month in a decade — on directional bets, according to Bloomberg. The firm has still generated more than $40bn of net trading revenue this year, surpassing all of 2025. In its latest debt offering it described its strategy as evolving to include longer-term positions, in line with a hedge fund (Bloomberg). The FT, ranking it against history's biggest trading losses, put the July figure at $35bn and attributed it to Leopold Aschenbrenner's leveraged AI stock bets. The two numbers are not reconcilable from public information — one may be a strategy's gross loss and the other the firm's net monthly P&L. We do not know. Either way it belongs on the all-time list. Why this matters beyond one firm: Jane Street is a core liquidity provider in ETFs and options. Its balance sheet is what stands between a redemption order and a price. If that same balance sheet is also running leveraged directional exposure to AI equities, then in a selloff the entity you need to make prices is the entity taking losses. Banks are capitalised and supervised precisely to prevent that overlap. Jane Street is not a bank, so nobody sees the overlap until a month like July. It sits alongside a bigger number. Federal Reserve research puts the cash-futures Treasury basis trade at roughly $830bn as of September 2025 — about double the 2020 peak, some 35% of hedge funds' long Treasury exposure, financed almost entirely in repo. The liquidity of the two largest markets on earth now depends on leveraged non-banks whose positions are disclosed after the fact, if at all.

Bond market dysfunctionhigh

Japan's insurers are sitting on $194bn

If Japanese life insurers become sellers of duration rather than buyers, the global long end loses its most reliable price-insensitive bid.

Unrealised losses on domestic bonds at Japan's major life insurers reached ¥30.86tn ($194bn) at the end of June, up 60% year on year. For the first time they exceed the same insurers' unrealised gains on domestic stocks, at ¥30.03tn (Nikkei). The cause is straightforward: 30-year JGB yields reached the 3.9% range by end-June, about 2.7 percentage points above July 2023. The standard reassurance is that unrealised losses vanish if you hold to maturity, and for a life insurer matching 30-year liabilities that is usually right. Three things break it. First, accounting: if a bond's market value falls 50% below cost, impairment can be forced. Nippon Life booked ¥44bn of impairments in April–June; Meiji Yasuda booked ¥25.3bn. Bonds bought in the late-2010s at near-zero yields are now reaching that threshold. Second, duration mismatch. If assets are longer than liabilities, rising rates cut net assets, and the fix is to sell bonds — converting paper losses into real ones. Third, and most important, policyholders. Sony Life's lapse and surrender rate rose 0.2 points year on year to 1.4% in April–June, partly because yen weakness prompted customers to cancel foreign-currency policies. A surrender is a withdrawal. Enough of them and a hold-to-maturity book becomes a sell-to-meet-redemptions book. That is the same arithmetic that killed Silicon Valley Bank in 2023, running in slow motion on a much larger balance sheet. Japanese lifers are structural buyers of ultra-long JGBs and, historically, of foreign bonds including Treasuries. The 10-year JGB is at 2.93%, its highest since 1996. Watch the lapse rate, not the yield.

What would change our mind

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

High-yield OAS above 400bp or investment grade above 120bp — a broad credit repricing rather than the sector-specific one visible in the QTS deal — would push ignition up sharply.

Would move the number

A data centre ABS or neocloud financing pulled for lack of demand, or Project Odyssey-style paper trading materially wider in the secondary market, would mean the marginal lender has gone from expensive to absent.

Would move the number

Median listed-BDC non-accruals above 3.5%, a dividend cut at a top-five BDC, or withdrawal limits at a non-traded BDC or interval fund — the first forced seller in private credit.

Would move the number

The 30-year Treasury back below 5% with a well-covered auction, and rate volatility staying low, would let me cut ignition several points; so would evidence that hyperscaler off-balance-sheet commitments are being funded from cash flow rather than new debt.

Would move the number