Nvidia signed the leases itself and plans to sell them on

Archived reading, published Thu, 27 Aug 2026 10:21:51 UTC (25 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

Fragility87

how much tinder is stacked up — moves slowly

Ignition35

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 29 sources and rewrites this page.

Why it moved: Down one to 61, entirely on the spark side: the VIX, the market's gauge of how much turbulence traders expect over the next month, printed 14.88, the lowest reading in our eight-week trail. The extra interest that shaky borrowers pay over the government is unchanged at 2.70 percentage points, safer corporate borrowers pay just 0.81 percentage points above the government, and the ten-year government bond rate sits at 4.64%. The long-end spark we carried last week keeps fading. Fragility holds at 87 because everything new this window (Nvidia's CFO commentary, second-quarter default rates at the big private lending funds, the exemptions carved into the new clearing rules for cash borrowed overnight against bonds) is sharper measurement of dry tinder already stacked, not fresh tinder.

Reporting from 25 Aug to 27 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.

Private credit and BDCs27 Augmedium

Four big lenders, all at post-2021 default highs

Default rates across the four largest private lenders are at five-year highs while their listed vehicles trade at their highs. The gap is entirely a question of when the prices they put on their own loans catch up.

The second-quarter 2026 numbers are now compiled, and they point one way. Loans that borrowers have stopped paying interest on across the top 20 listed private lending funds reached a median 2.8% of cost, up 0.8 percentage points on the quarter and the highest since 2017. Morningstar DBRS puts the industry-wide average at 3.4% of portfolios, from 3.1% in the fourth quarter of 2025. Blackstone's BCRED saw non-paying loans go from 0.6% to 2.4% in a single quarter, driven by two names: Medallia and Affordable Care International. Funds run by Ares, Blackstone, Blue Owl and Golub all reported loan default rates at their highest since 2021, with Blue Owl's fund at 2.8%, a five-year high. Now look at what the shares did. Blackstone Secured Lending is up 8.3% in twenty days and sits 0.26% off its period high. Ares Capital is up 6.25% and 1.5% off its high. FS KKR is up 13%. The extra interest that shaky borrowers pay over the government is at 2.70 percentage points, near the tightest it has been. The two claims are not quite contradictory, and the reconciliation is the interesting part. These funds' share prices are driven by the value of their assets and the dividend they pay, not by the default rate. At 11-12% portfolio interest rates, 3% of loans going bad still leaves the payout covered. The share price only breaks when the prices the manager puts on the other 97% move, and those prices are set quarterly, by the manager, on loans that do not trade on any exchange. So the market is not disagreeing with the data. It is saying the data does not reach it yet. A caveat we will not bury: these aggregates come from a compilation of second-quarter filings and third-party research rather than a single paper of record, and we treat the fund-level figures accordingly.

The AI capex bubble26 Aughigh

Nvidia is on the lease, and hopes to hand it off

The vendor is no longer just guaranteeing its customers' rent. It is signing leases itself on the assumption it can sell them to a buyer base it is currently helping to invent.

Nvidia's CFO commentary, published alongside Wednesday's results, puts numbers on the Ohio deal. Nvidia has entered guarantees covering the land, power and shell buildout for roughly 4.25 gigawatts at SB Energy's PORTS-Pike campus, a site that will exclusively host Nvidia hardware under 20-year leases to OpenAI. The guarantee obligations are capped at $105bn, become effective in phases, and the first phase is not expected until fiscal 2029. Nvidia also holds an option to provide the same kind of credit support for about 3.8 additional gigawatts as the site scales. The less-noticed sentence is the second one. The same commentary says Nvidia has signed data-center leases of roughly fifteen years, commencing in fiscal 2028 to fiscal 2029, and that it expects to reassign those leases to third parties. That is not a guarantee. That is Nvidia as tenant, warehousing lease obligations on its own books with a stated intention to sell them on. For it to work, a buyer has to exist in 2028 for fifteen-year leases on purpose-built AI shells, a market that does not exist today. The most likely buyers are the six asset managers Nvidia is already organizing into a $500bn compute-financing platform: Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR (The Economist). Worth noting where this started. CNBC reported the original OpenAI discussions ran to $250bn; the WSJ put the revised exposure at less than $120bn; the signed number is $105bn. The direction of travel on that particular figure has been down. The filings do not say how the reassigned leases will be accounted for, or what Nvidia books if no one takes them.

Bond market dysfunction26 Aughigh

One buyer took three-quarters of a Japanese government bond auction

A single buyer that does not care about price is now the swing bidder in the world's second-largest government bond market, and its buying is conditional on stocks going up.

On August 18 Japan's Finance Ministry auctioned about ¥2.5tn of five-year government bonds. More than ¥1.8tn went to unidentified buyers, according to dealers polled by Nikkei QUICK News. The largest identified buyer, Mitsubishi UFJ Morgan Stanley Securities, took ¥265bn, under a seventh of that. "The only explanation is that the whale was buying," one trader told Nikkei. The whale is the Government Pension Investment Fund, which manages ¥300tn ($1.88tn) of assets, making it the largest public pension in the world. Since a 2025 rule change it can bid directly at government bond auctions rather than through dealers. The explicit purpose was to stop traders who track its flows from trading ahead of it. Its identity emerges only after it has gone. The mechanism matters more than the anonymity. The fund runs a fixed four-way split (domestic bonds, domestic stocks, foreign bonds, foreign equities) and rebalances back to it. A global stock rally therefore makes it a mechanical, price-insensitive buyer of Japanese government bonds. In fiscal 2025 it bought a net ¥14tn of Japanese bonds and sold a net ¥10tn of domestic stocks, one of the largest such adjustments in a decade. This is the flow currently absorbing Japanese long-dated bond supply while smaller institutions sell into it at a loss: 17 shinkin banks (local cooperative lenders) posted net losses last fiscal year, up from five, on government bond writedowns. Run it backwards. If global stocks fall, the same arithmetic makes the fund a buyer of stocks and no longer a buyer of bonds. The buyer holding up the long end of a market with rising interest rates disappears precisely when everything else is being sold. And because the bidding is anonymous, nobody outside the ministry will know how much of recent demand was one rebalancing rule.

Hidden leverage and shadow banking26 Augmedium

The clearing mandate has holes exactly where the borrowed money is

The regulatory repair for the 2020 government bond blow-up does not cover the two structures used to finance the largest leveraged positions in the market.

The Office of Financial Research, the Treasury Department's data arm, made two posts in the last window. The first, on August 19, put hedge funds' cash government bond holdings at roughly $2tn, and stressed that because the profits on the trade are tiny, the funds "rely heavily on leverage through repo financing and futures margining." In plain terms: they borrow against the same bonds over and over, using each round of borrowing to buy more (OFR). The second post is the one to keep. Central clearing of cash borrowed overnight against government bonds plateaued in the first quarter of 2026. Two exemptions explain a lot of that. As of May 2026 the SEC, the agency that polices American markets, does not require trades between affiliated firms to be centrally cleared if the affiliate clears its other eligible trades. And the clearinghouse itself does not offer clearing at all for open-ended or flexible-term borrowing without a fixed end date, so those trades are exempt by construction. Open-ended borrowing is the workhorse of the biggest leveraged bet in government bonds: hedge funds borrowing heavily to profit from a small gap between two nearly identical prices. The post-2020 policy fix was to force this activity into a clearinghouse so that a default gets shared across members rather than chased through a chain of one-on-one contracts. The coverage of that fix has stopped growing while the position has roughly doubled its share of the market: regulatory filings for September 2025 show $2.4tn of long positions and $1.6tn of short positions in government bonds, $4tn gross, about 8.5% of privately held US government bonds against roughly 4.5% in early 2023, financed with some $3tn of overnight borrowing. Around 90% of that sits in the 50 largest funds, up from 84%. This is measurement, not new risk. We already count this trade. What is new is the concentration figure and the specific shape of the exemptions. Against it: the Federal Reserve Bank of New York's account of April-May 2025, when roughly $60bn then $40bn of related positions came off, the main leveraged bet held, and overnight funding did not seize. That unwind was a test, and it passed.

Crypto and TradFi contagion26 Aughigh

Legislating for a growth curve that has flattened

The channel by which digital tokens pegged to the dollar could drain bank deposits is being built out by governments at the moment the underlying usage has stopped growing.

The UK Treasury is giving the Bank of England a statutory secondary objective to support innovation in payment systems and digital money, an explicit push to make London a hub for dollar-pegged digital tokens after crypto firms complained the Bank was too conservative (FT). Financial stability remains the primary objective; growth is now the second. Separately, Revolut launched EURR, a euro-pegged token, initially to customers in Denmark, Poland and Portugal (Bloomberg). Here is the awkward fact underneath the policy. The combined market value of stablecoins (digital tokens meant to be worth exactly one dollar) has plateaued in 2026 after years of rapid growth, according to DefiLlama, and Visa's on-chain data, which tries to measure actual usage rather than wash volume, shows transaction volumes cooling this year. Tether and Circle still account for about 85% of tokens in circulation, three years after PayPal launched its own. This matters for the thing regulators actually fear. The systemic worry about stablecoins is deposit substitution: money leaves a bank that lends it into a token backed by government bonds that does not lend it. That transfer is what would disrupt the deposit model the Federal Reserve, America's central bank, will discuss at Jackson Hole this week, whose theme is financial innovation and payments. It requires the tokens to grow. They are not growing. Two readings, and the evidence does not choose between them. Either the risk is smaller than the volume of legislation implies, or growth resumes once incumbents (Revolut, Visa, Klarna, the banks) do the distribution that crypto-native issuers could not. Meanwhile the speculative end is very much awake: bitcoin is at $80,156, up 23.5% in twenty days, and the treasury companies have run harder still (BitMine up 50% and sitting exactly at its high, Strategy up 32%, SharpLink up 36%). That is appetite for borrowed-money bets on price, not for payments.

The AI capex bubble26 Augmedium

The borrowed end is being sold, the funded end is not

The market is already discriminating between AI spending funded by cash flow and AI spending funded by borrowing, before any broad measure of corporate borrowing costs shows stress.

Tuesday's close, and the dispersion inside the AI complex is worth more than any index level. IREN fell 6.2% on the day and 7.6% on the week. Applied Digital fell 6.8%, days after being reported as gauging interest in a roughly $3.5bn bond sale to investors willing to take higher risk for higher interest. Nebius fell 3.6% having upsized its convertible bond (a bond that can be swapped for shares) to $5bn. Core Scientific fell 3.4% and is 27% below its period high. Talen is down 4.9% on the week and 19% off its high; Vistra is 17% off. Against that: Microsoft sits 1.75% off its high, Oracle rose 2.8% on the day, Meta rose 1.1%, and the Nasdaq 100 was flat at -0.08%. The VIX, the market's gauge of expected turbulence, printed 14.88, the lowest reading in our trail. The split runs along a clean line. What is being sold are the firms whose next gigawatt depends on issuing new debt: high-risk bonds, convertibles, project finance secured on GPU chips. What is being held are the firms that can fund from the cash their businesses already generate. This is the market pricing financing risk, not AI demand risk, and it is doing so while the broad market shows nothing at all. The bond market is saying the same thing in the interest rates it demands. QTS sold $3.9bn of bonds for a Microsoft-tied Georgia facility at about 7.23% despite having investment-grade ratings; BlackRock paid 7.53% in July on blue-chip paper for a Texas project (Bloomberg). Investors are demanding the kind of compensation normally reserved for shaky borrowers, on debt issued by safe ones, in this sector. And they are getting it. If that persists, the buildout concentrates into fewer, better-capitalized hands. Whether that is de-risking or the kind of concentration this site keeps warning about depends on whether the periphery's debt gets refinanced or defaulted.

Signs of the times

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

$150bn of rent before the IPO

Anthropic's new $45bn, six-year deal with Nscale follows $50bn with Fluidstack, $45bn with SpaceX and $10bn with Volta Infra, roughly $150bn of compute commitments signed by a company that has not yet listed on a stock exchange. The Nscale contract alone covers 460MW, enough electricity for about 345,000 US homes at any one time.

aicapexpre-ipo

26 AugBloomberg

And there is an option for more

Nvidia's guarantees on the Ohio campus are capped at $105bn and cover about 4.25 gigawatts. The same commentary discloses an option to provide credit support for a further 3.8 gigawatts, nearly as much again, as the site scales.

vendor-financingguarantees

26 AugThe Economist

First loss since 1951

Ehime Shinkin Bank, on the island of Shikoku, lost ¥18.9bn in the year to March, the first net loss since it was founded in 1951, on Japanese government bond sales. Seventeen shinkin banks were loss-making, up from five, and 237 of 245 are sitting on unrealized securities losses.

ratesjapanunrealised-losses

26 AugNikkei Asian Review

Fifty-one billion dollars of unshipped servers

Dell ended its latest quarter with an AI server backlog of $51.3bn in orders it has not yet fulfilled, up from roughly $43bn at the start of the fiscal year, after booking $24.4bn of AI orders in a single quarter. The queue is growing faster than the shipping.

aisupply-chainbacklog

26 AugCrash Lab fact-check of Meet Kevin claim

The rumour mill

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

Confirmed

“Asked by a reporter about further bond market intervention, President Trump said: 'The ultimate intervention is our military. And if we have to use that, we will.'”

Said on camera at Joint Base Andrews on August 21, in direct response to a question about whether he had directed Bessent to intervene in the bond market. Carried by Bloomberg and multiple wires.

Claimed by Heresy Financial

Partly true

“Nvidia's CFO commentary implies total off-balance-sheet guarantees of about $198bn once the option on 3.8 additional gigawatts is included.”

The $105bn cap, the 20-year OpenAI leases and the 3.8GW option are all in Nvidia's own commentary. The $198bn total is the commentator's extrapolation; no filing puts a number on the optional portion.

Claimed by Meet Kevin

Partly true

“Nvidia wants to give OpenAI a $125bn guarantee.”

Talks reportedly ran as high as $250bn before being cut; the WSJ put revised exposure below $120bn and the signed guarantee is capped at $105bn. The direction on this number has been downward, not upward.

Claimed by Wealthion

Partly true

“The Federal Reserve has bought over $330bn of Treasuries in the last twelve months, which is QE by another name.”

The Federal Reserve's weekly balance sheet report shows government bond holdings up roughly $325-340bn year on year, so the number is right. The Fed frames this as a pause in runoff plus reserve management rather than asset purchases; whether the label matters is a separate argument from whether the data are accurate.

Claimed by Palisades Gold Radio

Earlier developments

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

Private credit and BDCs27 Auglow

The exit door is being rationed

Gates are how a problem with loans turns into a problem with cash, without a single borrower failing to pay.

MA Financial, one of the largest private lending firms in Australia's roughly A$200 billion market, has capped withdrawals from its A$2.3 billion property loan fund after a surge of investors asking for their money back. Investors can now only get a portion of what they request each quarter. That is one named fund. The broader number in the same sweep is the one worth arguing about: redemption requests at private credit funds sold to individual investors averaged 14% of total fund value in the second quarter of 2026, up from 13% in the first quarter, against quarterly withdrawal caps that typically allow only 5% out (web sweep). If that is right, gates are not an isolated event. They are the normal state of the retail channel. We flag the sourcing plainly: this comes from a single aggregated analysis, not from a major newspaper or a fund filing. Treat the 14% as a claim, not a fact. The mechanism does not depend on the number being exact. A fund that holds loans which cannot be sold quickly and offers quarterly withdrawals works only while requests stay under the cap. Once they exceed it, everyone gets a fraction of what they asked for, which rewards whoever files first. Every remaining investor learns the queue is real. Nothing has to go bad for this to bite. The publicly traded market sees none of it. Ares Capital sits 1.5% off its high, Blackstone Secured Lending 0.3% off, and FS KKR is up 13% in twenty days. The extra interest rate that shaky borrowers pay over the government is 2.7 percentage points. Either the gates are a local Australian property problem, or the traded versions are looking at the wrong part of the structure.

The AI capex bubble26 Augmedium

Nebius is funded by its own customers

The new cloud-computing funding model turns customer concentration into funding concentration, and neither shows up as borrowed money on anyone's books.

Nebius, a cloud-computing company building data centers for AI, increased a bond offering that converts into shares to $5 billion this week to pay for chips and data centers. It says it could spend up to $25 billion on building things this year. The more interesting disclosures are about where the rest of the money comes from. Roughly 70% of the deals Nebius closed in the second quarter included an upfront payment from the customer, and management expects more than $9 billion of customer prepayments during 2026 against more than $40 billion of total customer commitments. In July it borrowed $775 million, pledging its existing chips and the cash flows from a contract with a highly rated customer as security. So the funding stack is: convertible debt, customer cash paid in advance, and loans secured on chips and contracts. Each layer depends on the same underlying thing: that a handful of AI labs keep paying. That is the structure to hold in mind. A prepayment is not revenue. It is a customer lending money to its supplier. If one large tenant slows down, Nebius loses the contract, the prepayment stream, and the value of the security behind the chip-backed loans, all at the same moment. The three sources of funding are not diversified. They are one source, counted three times. The market is not treating this as a problem. Goldman's Alexander Duval raised his price target to $328 from $286 this week, the highest on Wall Street, and our fact-check confirms it. Nebius is up 44% in twenty days at $213.96. It is also 22.9% below its recent high and fell 3.6% on Wednesday. Both facts are true. The stock has been repricing this business every few weeks all summer.

Crypto and TradFi contagion26 Aughigh

MSCI may cut the wire Saylor climbs

The model where a company buys digital assets using index-fund demand depends on index membership it does not control, and the index provider is now asking whether it should be there at all.

MSCI, the firm whose indexes determine what trillions of dollars of passive funds must own, is consulting on a rule that would kick out companies it judges to be primarily accumulating assets rather than operating a business. Applied as of May, it would have removed Strategy Inc., Japan's Metaplanet, and the uranium vehicle Yellow Cake (Bloomberg). MSCI floated a version of this aimed specifically at crypto in January and dropped it. It is back, framed as a test of whether a company actually operates, rather than whether it holds digital assets. That is harder to lobby against. The mechanism matters more than the coin price. Strategy has financed more than $60 billion of bitcoin purchases by issuing shares and share-linked paper at prices above the value of the bitcoin it holds. That only works while there is reliable demand for the shares. Index membership manufactures a slice of that demand: passive funds must hold the stock because it is in the benchmark, not because anyone chose it. Remove it, and forced buyers become forced sellers, the gap between share price and bitcoin value compresses, and issuing new stock to buy more bitcoin starts destroying value rather than creating it. The flywheel runs backward. Nothing about bitcoin itself changes. What changes is the plumbing that converts index money into coin purchases. The market is not pricing the threat. Strategy is up 18.2% in five days, at $123.14, while bitcoin rose 2.2% over the same stretch. The borrowed-money effect in the vehicle is expanding, not contracting. The consultation is a proposal, with no decision date in the reporting. But the question MSCI is asking is the one the whole digital-asset-treasury sector was built to avoid.

The dollar, gold and reserve status26 Aughigh

Seven billion dollars of insurance in five days

Investors are buying insurance against the policy mix without selling the currency, which tells you they expect erosion, not rupture.

Gold and bitcoin ETFs took in a record $7 billion over five trading days: $3.4 billion into SPDR Gold Shares and $1.5 billion into BlackRock's iShares Bitcoin Trust, both in the top ten US ETFs by weekly inflow (Bloomberg). The stated catalyst is Scott Bessent's plan to at least double the Treasury Department's purchases of government bonds that do not come due for twenty or thirty years. What makes this worth a dispatch is what did not happen alongside it. The dollar index, which tracks the dollar against a basket of other currencies, is at 99.15, down 0.02% on the day and 0.65% over twenty days. The euro is up 1.8% over the same period; the yen 2.5%. The extra interest rate that safe corporate borrowers pay over the government is 0.81 percentage points and has not moved in a month. This is not capital fleeing the dollar. It is dollar-based investors buying assets whose supply no government controls, while keeping their money in dollars. The distinction matters. A crisis in the dollar's status as the world's reserve currency shows up in the currency itself and in foreign demand for US government bonds. A hedge against the policy mix, a Treasury Department suppressing long-term interest rates while inflation runs above target, shows up exactly here: gold at $4,660, up 15.1% in twenty days at a record, bitcoin up 21.5%, and the exchange rate flat. Bernstein's Gautam Chhugani put the argument plainly: the forty-year era of falling rates has ended, leaving governments exposed to debt-service costs at record debt levels. You can agree with that and still note that this trade is a bet on slow erosion over years, not on a break next week. It is the cheapest thing to be right about early.

Fed, Treasury and policy26 Aughigh

Two arms of the state, pulling opposite ways

If the Treasury Department is setting the long-term interest rate, the Federal Reserve's inflation target becomes a statement of intent rather than a constraint.

Major investors have told the FT that Bessent's move to at least double Treasury purchases of long-term government bonds works directly against the Federal Reserve's attempt to bring inflation down (FT). The point is mechanical, not rhetorical. If buybacks succeed in pushing long-term interest rates lower, mortgage rates and corporate borrowing costs fall with them. That is monetary easing, delivered by the Treasury Department, at a moment when several Federal Reserve officials have said the policy rate may need to go up. Kevin Warsh speaks at Jackson Hole on Friday, his first appearance there as chair of the Federal Reserve, America's central bank. Bank of America's Mark Cabana says the market needs details of how the Fed gets inflation back to target after five years of overshooting; the extra interest lenders demand for tying their money up for a long time has risen partly because it has not heard them. Warsh has previously said rising interest rates were healthy, a position now flatly at odds with a Treasury Secretary who says yields "don't reflect the underlying fundamentals." The unresolved question is who controls the interest rate on long-term government debt, and with what tools. Warsh wants to shrink the Fed's holdings of bonds, which pushes long-term rates up. Bessent is buying long-term bonds, which pushes them down, funded by issuing short-term bills. Done together without coordination, they cancel out at a cost. Done together with coordination, it is an accord, and an accord means the central bank has accepted a debt-management objective. For now the market is calm about it: the ten-year rate has drifted back to 4.64% from 4.73% a week ago, and the MOVE index, which measures expected turbulence in government bonds, is down 16.5% in twenty days. The first buyback under the expanded program has not happened yet. It is scheduled for September 9.

Household credit27 Augmedium

The pipeline is emptying, not curing

Household stress is not spreading. It is concentrating, and concentrated losses land on specific lenders rather than the economy.

Two household numbers landed this window that look contradictory and are not. Credit card payments that are thirty or more days late on bank-issued cards were 2.85% in the second quarter, the lowest since the second quarter of 2023. Meanwhile 12.92% of credit card balances were ninety or more days late, down only slightly from 13.1% in the first quarter, a level the New York Fed data put at the highest since 2011. The first is a flow: how many people fell behind this quarter. The second is a stock: how much debt is sitting deep in arrears. When the flow improves while the stock stays pinned at a fifteen-year high, the deeply late balances are not recovering. They are aging toward the point where the lender writes them off, and the improvement at the front of the pipeline is arriving too late to help them. The same split runs through car loans. Payments sixty or more days late on subprime auto loans hit 6.13% in July, the highest July in a series running back to 1994. Prime was 0.49%, more than twelve times lower. The New York Fed puts auto debt ninety or more days past due at 5.5%. So the average consumer is fine and getting slightly better; a defined subset is at record stress. That is not a contradiction, it is the whole story. It also explains the market: Capital One is 4.1% off its high, Synchrony 1.5%, Credit Acceptance up 6.8% in twenty days. Lenders to the riskiest borrowers are priced as though a 6% late-payment rate is simply what that book earns, which, if the assumptions about how much they lose on each bad loan hold, it is.

Private credit and BDCs27 Augmedium

Australia finds the cracks first

This is the first time a national securities regulator has said publicly that private credit funds in its jurisdiction are blocking withdrawals while borrowers fail. That is the sequence everyone has modeled and nobody has watched happen.

Sarah Court, the chair of ASIC (the agency that polices Australian markets), told a gathering in Sydney that the regulator is seeing "the first significant cracks" in Australian private credit (lending to companies by investment funds rather than banks). Several large borrowers have collapsed and several major funds have limited or blocked withdrawals. Two names are attached: the New South Wales property developer Bathla, which appointed administrators this week, and the hospitality operator Jon Adgemis, who went bankrupt earlier. Both were large borrowers from private credit funds. The more useful quote came from the central bank. "People don't know where the leverage is," said Michele Bullock, governor of the Reserve Bank of Australia. "They don't know who is exposed. So, any time that there's a big unknown, you know it's a big chunk of lending, but you don't know anything about it." Bullock also said in August that she did not think there was "a massive worry" about it in Australia. Both things can be true, and the combination is the point: the regulator's stated position is that the sector is large, impossible to see into, and untested, and that it is now being tested. The way the risk reaches ordinary people is through superannuation, Australia's compulsory retirement savings system. Most adult Australians hold private credit through their pension funds, in products that promise savers they can take their money out periodically, even though the underlying loans cannot be sold quickly. Blocking withdrawals is the designed response to that mismatch, and it works, right up to the point where it becomes the signal that something is wrong. Australia is a small, clean version of the structure we track everywhere: loans valued by the manager who made them, funded by savers who think they can leave. Listed private credit in the US shows no trace of this. Ares Capital is 1.5% off its recent high, Blackstone Secured Lending 0.26% off, and the extra interest that shaky borrowers pay over the government is unchanged at 2.7 percentage points. One regulator's warning in one country is not contagion. It is, however, the first regulator to use the word cracks.

The AI capex bubble26 Aughigh

Financing a fax machine

If the chips are not real security for the loans and insurance is not available, the entire $7 trillion is unsecured lending to four or five tenants, dressed up as secured.

The FT puts the number at $7 trillion of data-center spending by 2030, enough, it notes, to feed every person in China for three years. Even the companies with tens of billions in cash cannot fund that internally, so lenders are being asked to underwrite a type of asset that has no history. The specific problem is what backs the loans. Much of this debt is secured on chips. "It's like you're financing a fax machine and then someone invented email," says Carlos Mendez of Crayhill Capital, which has repeatedly turned down deals secured on chip value that lacked structural protections. How long a chip stays useful is an assumption, not a fact, and the loans are written to be paid back over decades: one vehicle described in market commentary this week carries an A+ rating, pays 6.58%, and does not fully mature until 2049. Then there is the half-built problem. "As long as developers have the permits they need, they can continue borrowing," says Eric Klar of White & Case. "But local opposition or the loss of a permit could stop that midway through construction. For lenders who have already put money into the project, the question is: what happens then?" A partially completed data center is not worth a fraction of a completed one. It is worth land and concrete. And the risk cannot be laid off. Insurers have been reluctant to write billions of coverage on the biggest sites, worried about catastrophe and concentration. Strip these away and what is actually supporting the debt is the tenant's promise to pay rent. That is lending to a big tech company wearing project-finance clothes, priced as though the hardware were the security. Data-center bonds yielding 7.23% (QTS) and 7.53% (BlackRock) suggest buyers already suspect as much. In a normal market, bonds backed by real assets from borrowers this large would pay far less.

Bond market dysfunction26 Aughigh

The whale is bidding in the auctions now

Japan's government bond market at the long end is being held up by one rebalancing rule at one pension fund. That rule flips direction if stocks fall.

On August 18, Japan's Finance Ministry auctioned about ¥2.5 trillion of five-year government bonds. More than ¥1.8 trillion went to unidentified buyers, according to dealer interviews by Nikkei QUICK News. The largest identified buyer, Mitsubishi UFJ Morgan Stanley, took ¥265 billion. "The only explanation is that the whale was buying," one trader said. The whale is GPIF, Japan's public pension fund, which manages more than ¥300 trillion (making it the largest of its kind in the world). A 2025 rule change allowed it to bid directly in government bond auctions rather than going through dealers. The stated advantage is that it can buy without traders seeing it coming and driving the price up. Its identity emerges only after it has left. The mechanism worth understanding is why it is buying. GPIF splits its money evenly across four buckets: Japanese stocks, foreign stocks, Japanese bonds, and foreign bonds. Global stocks have rallied, so the bond bucket is underweight relative to the target, so it buys bonds. In the fiscal year through 2025 it bought a net ¥14 trillion of Japanese bonds and sold a net ¥10 trillion of Japanese stocks, one of the largest rebalancings in a decade. Two consequences. First, when 72% of an auction goes to a single buyer that is not shopping for value but filling a quota, the auction stops being information about whether private investors actually want to lend to the Japanese government for five years. Anyone reading Japanese auction results as a signal of demand is reading GPIF's spreadsheet. Second, the mechanism reverses. If global stocks fall, GPIF becomes a seller of bonds and a buyer of stocks, pulling its support from the bond market precisely when the bond market is under stress. Meanwhile the traditional domestic buyers are going the other way. Seventeen shinkin banks (small Japanese cooperative lenders) posted net losses last fiscal year, up from five, and their bond-related losses nearly tripled to ¥424.8 billion as they sold into rising interest rates. The whale is replacing them.

Hidden leverage and shadow banking26 Augmedium

Ninety percent of it sits in fifty funds

The trade where hedge funds borrow heavily to profit from a tiny gap between two nearly identical prices in the US government bond market is now bigger and more concentrated than it was in 2020, and the reform designed to contain it has stopped expanding.

The Form PF numbers circulating this week (from mandatory filings by large hedge funds) put those funds at $2.4 trillion of bets that US government bonds will rise and $1.6 trillion of bets they will fall, for $4 trillion of total exposure. That is about 8.5% of privately held US government bonds by market value, roughly double their share in early 2023. Inside that: $830 billion of the basis trade (hedge funds borrowing heavily to profit from a small gap between two nearly identical prices), close to twice its early-2020 peak, and $305 billion of a related trade on interest-rate swaps, financed with something like $3 trillion of cash borrowed overnight against bonds put up as security. The figure that matters is the concentration. About 90% of the $4 trillion sits in the fifty largest funds, up from 84% in early 2023. Fifty firms, borrowing overnight, holding one-twelfth of the tradeable US government bond market. The Office of Financial Research added the plumbing detail on August 19 and 20. Because the basis trade earns very little per dollar, funds "rely heavily on leverage through repo financing and futures margining." The overnight borrowing lets the same bond be pledged as security down a chain of transactions, with each link borrowing against it again. Separately, the OFR notes that central clearing of overnight bond borrowing (the reform meant to make this safer by putting a clearinghouse between the two sides of every trade) plateaued in the first quarter of 2026. It named two holes: from May 2026 the SEC, the agency that polices American markets, will not require trades between affiliates of the same firm to be centrally cleared if the affiliate clears its other eligible trades. And the clearinghouse does not offer clearing for open-ended loans with no fixed end date, which exempts them from the mandate. Central clearing was the post-2020 answer to exactly this kind of fragility. It has stopped growing, with carve-outs, while the position it was meant to make safe has doubled. None of this is new risk. The data is nearly a year old and we have counted it before. What is new is the shape: fewer holders, more borrowed money, a clearing regime that has stalled. The short-term cash markets that keep trading desks running day to day show nothing yet. The overnight borrowing rate (SOFR) is 3.66%.

Household credit26 Augmedium

The averages are hiding two different consumers

Household credit is splitting into two populations rather than deteriorating as one, which means the overall numbers will keep looking calm right up until the bottom third stops paying.

New York Fed data for the second quarter of 2026 show 12.92% of credit card balances more than 90 days late, barely below the first quarter's 13.1%, which was the highest since 2011. Read that alone and the consumer is in trouble. Read the flow measures, which track new payments going bad rather than old ones sitting there, and the consumer is fine. Card payments 30 or more days late on bank-issued cards were 2.85%, the lowest since the second quarter of 2023. Equifax data put 60-day lateness across all cards at 2.69%, down from 2.87% a year ago. Student-loan balances falling into serious trouble dropped to about 8% in the second quarter, after surging above 16% in 2025. Total household debt was $18.8 trillion, down $13 billion on the quarter. Households are spending 11.2% of their income on debt payments, below both the 12% long-run average and the 11.7% of 2019. These are not contradictory. They are two populations averaged into one number. The pile of deeply overdue balances stays elevated because the households carrying most of the debt cannot get out. The flow improves because most borrowers are not those households. The Atlanta Fed's framing is the sharpest: 65% of credit card debt is held by households who say they are "finding it difficult to get by" or "just getting by." And the extreme case remains car loans to borrowers with poor credit, where payments 60 or more days late hit 6.13% in July against 0.49% for borrowers with good credit, more than twelve times higher. We are taking a point off fragility for this. The student-loan rate halving is a genuine reduction in a risk we have been carrying since last year, not a re-description of it. The lenders agree: Capital One, Synchrony, Ally, and Credit Acceptance are all within 6% of their recent highs.

Fed, Treasury and policy26 Augmedium

Bessent's next two weapons: the government's checking account and the 20-year bond

If the Treasury Department starts using its cash balance to push down long-term interest rates, debt management and monetary policy have merged, and only one of them has an inflation mandate.

The Treasury Department (which borrows the money the government spends) will run its enlarged buyback program from September 9 to November 4, at least $4 billion per operation, buying back government bonds that do not come due for ten to thirty years. The first announcement bought roughly one day of relief: the interest rate the government pays to borrow for thirty years fell about a tenth of a percentage point, then went back to 5.25% and finished the week near 5.28%, close to where it started. Two escalations are now being briefed, and both come from below our usual sourcing standard. CNBC, via Czech outlet Seznam Zprávy, reports officials are weighing funding the buybacks from the Treasury General Account, the government's checking account at the Federal Reserve (America's central bank, which sets interest rates), which currently holds roughly $950 billion. The New York Post's Charles Gasparino reports the plan could extend to temporarily halting sales of the 20-year bond. Treasury has not commented on either. The mechanics matter more than the sourcing. Buying long-term bonds while selling short-term ones is a maneuver the central bank used after 2008 to push down long-term rates without printing money. Doing it from the Treasury side is new. Funding it out of the government's checking account goes further: money leaving that account and entering the banking system adds cash to the financial system. That is monetary easing, done by the Treasury, at a moment when several Federal Reserve officials think rates should be going up. Investors have told the FT the intervention undercuts the Fed's inflation fight and the Treasury's own credibility. Kevin Warsh speaks at Jackson Hole on Friday, his first appearance as Fed chair. Mark Cabana of Bank of America: "Warsh's 'resolute commitment' to lower inflation is not good enough for the market. We need to hear details." The market has already voted. Gold is at a record $4,695, up 2.1% in a day and 16% in twenty. The interest rate on ten-year government bonds, at 4.64%, is calmer than a week ago. Those two prices are answering different questions. Gold says investors do not trust the currency. The bond market says they still trust the payments.

The AI capex bubble26 Aughigh

Two private companies, $45bn, six years

The largest AI compute contracts are now being written between companies with no public accounts, which means the credit risk sits with lenders and cannot be read off any filing.

Anthropic has agreed to pay Nscale $45bn over six years to rent AI computing power from a data centre in West Virginia, Bloomberg reports. The commitment covers about 460 megawatts — enough electricity for roughly 345,000 US homes at any one time — and buys capacity in the first of three buildings at Nscale's Monarch campus, which will total about 1.35 gigawatts. The chips are Nvidia's Vera Rubin, which start coming online late next year. Neither party to this contract is a public company. Nscale is London-based and preparing an IPO as soon as next month, seeking up to $3bn. Anthropic was valued at $965bn in May and is expected to list soon. So a lender asked to finance Monarch is underwriting six years of payments from a private borrower to a private landlord, secured on hardware that does not yet exist, at a site that is one-third built. Add up Anthropic's disclosed compute deals — $50bn with Fluidstack, $45bn with SpaceX, $10bn with Volta, now $45bn with Nscale — and the company has committed roughly $150bn to rent computers. That arithmetic is ours; the components are Bloomberg's. Anthropic does not publish revenue, so there is no public number against which to test it. The tape disagrees with the reporting, and sharply. On the day this deal landed, the listed members of Nscale's cohort fell: Applied Digital −6.8%, IREN −6.2%, Nebius −3.6%, Core Scientific −3.4%. Oracle rose 2.8% and Microsoft 1.0%. Investors are happy to fund neocloud demand when a hyperscaler with earnings sits behind it, and are marking down the companies whose whole equity value is one tenant's promise.

Bond market dysfunction26 Aughigh

Japan is running 2023's bank problem in slow motion

Unrealised bond losses at small, thinly capitalised lenders are the classic route from a rates move to a funding scare, and Japan now has ¥2.6tn of them.

Seventeen of Japan's shinkin banks — community lenders, roughly credit unions — posted a net loss for the fiscal year ended in March, up from five a year earlier, according to a Nikkei analysis of disclosures from 245 institutions. Combined net profit fell 16% to ¥242.4bn ($1.52bn). Bond-related losses across the sector reached ¥424.8bn, nearly triple the year before. Ehime Shinkin Bank lost ¥18.9bn, its first loss since it was founded in 1951. The mechanism will be familiar. Rural deposits kept coming in while loan demand shrank, so during Japan's ultralow-rate decades these lenders put the money into government and municipal bonds. Rates then rose. As of end-March, unrealised losses on securities across the sector were about ¥2.6tn, up 7% on the year; 237 of the 245 institutions were carrying unrealised losses and 205 of those were deeper than the year before. This is the same duration mismatch that broke Silicon Valley Bank, running much more slowly and with a different ending available. Selling crystallises the loss but frees the balance sheet to buy higher-yielding paper — Tajima Shinkin, which lost ¥19.5bn, called its selling "a management decision". Holding avoids the loss but leaves it embedded. The weakest institutions are the ones holding. Shinkin Central Bank injected about ¥4bn of capital into Tochigi Shinkin in 2025. Two things make this worth watching from outside Japan. Listed regional banks posted record profits over the same period, so the pain is concentrated in exactly the institutions with the least capital. And these are marginal holders of JGBs at a moment when the long end is under pressure everywhere.

Crypto and TradFi contagion26 Aughigh

MSCI may define Strategy out of the index

The digital-asset treasury model runs on a share-price premium that passive index demand helps sustain, and an index provider can withdraw it with a rule change.

MSCI is consulting on rules that would exclude from its global equity indexes any company it judges to be "primarily engaged in accumulating assets rather than operating businesses". Bloomberg reports that applying the proposed methodology as of May would have removed Strategy Inc., the Japanese bitcoin treasury company Metaplanet, and Yellow Cake Plc, which holds uranium. This matters more than a normal index tweak, because index membership is load-bearing for the business model. The machine works like this: passive funds tracking the index must own the shares; that demand helps keep the stock above the value of the bitcoin it holds; the company sells new shares into that premium; it uses the proceeds to buy more bitcoin, which raises bitcoin per share, which supports the premium. Strategy has financed more than $60bn of purchases this way. Take out the forced buyers and one blade comes off the flywheel — MSCI's own strategists frame the risk as reduced demand for the shares and strain on the capital structure. The threat is not new; MSCI dropped a crypto-specific version of it in January and replaced it with this broader test of what counts as an operating company. What is new is that the broader test appears to catch the same names anyway. Meanwhile the treasury complex has just had an extraordinary week that bitcoin did not have. Bitcoin is up 0.6% over five days. Strategy is up 18.2%, BitMine 23.1%, SharpLink 17.9%, with two of the three sitting at their period highs. Whatever is repricing these shares, it is not the coins on their balance sheets.

The dollar, gold and reserve status26 Aughigh

$7bn into the assets Washington cannot print

The hedging is happening in gold and bitcoin rather than in the dollar itself, which is why the FX market looks calm while the fiscal anxiety is real.

Gold and bitcoin exchange-traded funds took in a record $7bn over five trading days, Bloomberg reports. SPDR Gold Shares took $3.4bn and iShares Bitcoin Trust $1.5bn; both made the top ten US ETFs by inflows for the week, with the gold fund trailing only a handful of vehicles including the largest S&P 500 tracker. The proximate cause named in the reporting is Scott Bessent's plan to at least double long-dated Treasury buybacks. That is worth sitting with. An intervention designed to calm the bond market pushed money into the two assets whose supply is beyond the reach of the people intervening. Bernstein's Gautam Chhugani frames it as the end of the 40-year era of falling rates, leaving governments exposed to debt-service costs at record debt levels. What is notable is where this is not showing up. The dollar index is 99.14, up 0.2% on the day and down only 0.9% over twenty days. The euro and yen are flat. This is not a currency crisis; it is a specific bid for scarcity running alongside an unremarkable foreign exchange market. Gold is up 13.4% in twenty days, silver 19.1%, bitcoin 22.6%, while the broad dollar has barely moved. One caveat on reading too much into the gold-and-bitcoin pairing: flows into a vehicle are not the same as conviction about a regime, and both assets have momentum of their own. But the two have historically taken turns as the haven trade. Buying both, in size, in the same week, is a statement about the issuer of the currency rather than about either asset.

Fed, Treasury and policy26 Augmedium

The Bank of England is told to promote innovation

Money leaving supervised bank balance sheets for non-bank issuers is the core Crash Lab thesis in its purest form, and it is now being encouraged by statute.

The UK Treasury is giving the Bank of England a statutory "secondary objective" to support innovation in payment systems and digital money, the FT reports, after criticism from crypto firms that the Bank has been too conservative. Financial stability remains the primary objective; the City minister, Lucy Rigby, says the new duty will "support the bank to continue to drive innovation in payments and digital finance". Secondary objectives are not decoration. They give firms a statutory hook to argue that a proposed rule is disproportionate, and they change what a supervisor has to justify. The Bank published draft rules for sterling stablecoins in June, having previously been attacked for planning ownership limits designed to stop deposits leaving banks too quickly. Those limits are exactly the kind of measure the new objective invites the industry to contest. The underlying question is the one Jackson Hole is built around this year: whether tokenised money disrupts the traditional bank deposit model. A deposit funds a loan and is supervised, insured and counted. A stablecoin balance is a claim on a reserve pool held by a non-bank issuer. The stability risk is not that the token breaks; it is what happens to bank funding if enough deposits move, and what a run on an issuer does to the short-term paper it holds as reserves. Separately, Revolut launched a euro-pegged token, EURR, for customers in Denmark, Poland and Portugal. It joins a crowded field: stablecoin market value has plateaued in 2026 and Tether's USDT plus Circle's USDC still account for about 85% of everything in circulation.

Hidden leverage and shadow banking26 Aughigh

The protection with nothing behind it

Capital relief bought with an unfunded promise is only as good as the promise, and the whole point of the structure is that nobody has to fund it until the day everything else is going wrong too.

Insurers signed up to cover the risk of default on €4.7bn of synthetic risk transfers in 2025, up from €2.7bn in 2024, according to survey data from the International Association of Credit Portfolio Managers reported by the FT. That sits inside a European bank risk-transfer market that shifted €579bn of credit risk last year, on Barclays' numbers. So the insurers are a sliver. The sliver doubled. Here is the mechanism. In a significant risk transfer, a bank keeps a pool of loans on its books but sells the first losses on that pool to an outside investor for a fee. The regulator then lets the bank hold less capital against those loans, so it can lend more. When a hedge fund or a credit fund takes that risk, it posts collateral into an account for the life of the deal — the money is there. When a property and casualty insurer takes it, it does not. It writes the cover unfunded, backed by its rating and its balance sheet. Regulators permit this on the reasoning that bank default risk is uncorrelated with the hurricanes and cyber attacks these insurers otherwise underwrite. That is true in almost every year. It is not true in the year the protection is actually needed. Munich Re, The Fidelis Partnership and AIG-backed Convex are among the names writing it. Fitch's Monsur Hussain put the comparison plainly to the FT: credit default swaps are unfunded credit protection, and this is "essentially the same technology", though he added that AIG's volumes were far larger than anything European insurers are doing now. The volumes keep growing. Crescent Capital expects SRT sales to hit a record $45bn this year against $41bn in 2025; AIB is preparing one on €2.5bn of project finance loans, brokered by Howden, worth 25-30bp of core capital.

What would change our mind

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

The extra interest shaky borrowers pay over the government widening through roughly 3.50 percentage points, or safer corporate borrowers through 1.00 percentage point, on consecutive sessions. That would be the first observable sign that the private-credit default data has reached public markets.

Would move the number

A listed private lending fund cutting its dividend or trading below 80% of the value it puts on its own assets, which would mean the prices on the loans, not just the non-paying count, are moving.

Would move the number

Confirmation in a filing that Nvidia's roughly 15-year data-center leases have actually been reassigned to a named buyer at a stated price. That would lower fragility. Silence into fiscal 2028 would raise it.

Would move the number

The overnight borrowing rate printing materially above the rate the Federal Reserve pays banks to park cash, or a failed or heavily undersubscribed auction of long-dated government bonds despite the doubled buybacks. Either would move ignition immediately.

Would move the number