An Essay on the Architecture of Financial Collapse

The Ghost in the Machine

The 2008 Crash Never Actually Ended, and What's Coming Is Much Bigger

Preface: The Guy in the Bar

I was sitting in a bar in 2004 when the guy next to me said the most insane thing I had ever heard.

"Hey man, do you have a job?"

I said yes, I have a job.

"Would you take an $85,000 mortgage with no money down?"

I looked at him like he had lost his mind.

"Buddy, that's not how mortgages work. A mortgage requires earnest money. Twenty percent down, at least. No money down mortgages? You're nuts, man."

He looked stunned at my ignorance. Then he told me he could easily get me the mortgage with no credit check, just my job as proof of income. He showed me a mortgage he had just gotten someone to sign. A variable-rate balloon loan. Lunatic stuff.

The whole time he was talking, I was thinking about what had happened to me at the car lot the week before. I was trying to finance an $18,000 new Jeep. The salesman pulled my report and told me I was a "credit ghost." Nothing on the report. Nothing to score. They couldn't finance me on a new car at any price.

So here I was. Couldn't get $18,000 of credit for a new Jeep I needed to get to work. But this guy in the bar wanted to hand me $85,000 to buy a house, no questions asked.

I begged off his insane offer and he wandered off to find someone else. I had no idea at the time that the idiot in the bar had just tried to drag me into what would become the 2008 housing crisis. I just knew he was full of shit.


In 2008, I found myself working on the back end of the same nightmare. Cleaning out the houses of the foreclosed.

It was tragic. People had left everything. Family pictures still hanging on the walls. Kids' drawings still stuck to the fridge. Wedding albums. Photos of grandparents. Things you take when you have time. Things you leave when you don't.

Some people had stopped taking out the garbage and let it pile up waist high before they walked out.

The worst, though, was when I would go into a house and find a dead animal. A dog. A cat. Left behind. I couldn't bring myself to do cleanouts after a few of those.

I would always think back to the guy in the bar. For a long time I despised him for being such an idiot. It took me years to understand that he was not the idiot. He was the tool. Somebody above him had figured out how to turn a credit ghost into a closing fee, and that somebody was still standing when the dust cleared. The guy in the bar got laid off. The family in the house got foreclosed. The cat starved. And the people at the top of the chain bought back the houses for thirty cents on the dollar.

This essay is about why that machine never actually got shut down. It is about where it moved to after 2008. It is about the much bigger version of it that is running right now, in May 2026, that I keep watching the same warning signs gather around. And it is about who is going to pay for it when it breaks again, which is the same people who paid for it last time. The people who left the family photos on the walls. The credit ghosts. You. Me. The cat.

You don't need a finance degree to follow what comes next. If you have ever been the credit ghost at the car lot while the guy in the bar was getting handed $85,000, you already understand the whole architecture. The rest is just names.

A Note on the Numbers Before We Start

When the 2008 crash happened, the world's total stack of side bets (the financial industry calls them "derivatives") was about $586 trillion.1 That number was so large nobody could really picture it. For comparison, the entire American economy in 2008 produced about $14 trillion in goods and services. The side bets were forty times bigger than the country itself.

Today, that number is $846 trillion. The Bank for International Settlements reported it in December 2025.2 The stack of side bets has grown by $260 trillion since the last crash, and the yearly growth rate just hit the highest level seen since right before 2008.2 The 2008 number was already too big to bail out. The 2026 number is fifty percent bigger than that.

$846TGlobal OTC derivatives outstanding, June 2025 (BIS)
+16%YoY growth, largest since 2008 (BIS)
$586T2008 peak baseline (BIS historical)
$29TEntire U.S. economy, 2025

That is the first thing to understand. Whatever happens next will not look like 2008. It will look like 2008 multiplied. The 2008 crisis took out the housing market and froze the global banking system for a year. The current setup has stacked the same kinds of bets on used cars, on office buildings, on business loans, on private deals nobody can see, and, newest of all, on the entire artificial intelligence buildout. The 2008 collapse blew open one hole. What is being built right now has at least five holes ready to blow at once.

Michael Burry, the investor played by Christian Bale in the movie The Big Short, the man who saw 2008 coming when nobody else did, has been saying since 2023 that the current setup is a bigger bubble than the one he shorted before. In May 2026 he posted on his Substack that the market has "jumped the shark."3 He has personally placed roughly $1.1 billion in bets that the AI rally is going to crash.4 He thinks individual Big Tech stocks could fall 40 to 50 percent.4 He has compared the current period to March 2000, the exact month the dot-com bubble peaked before falling 80 percent over the next two years.3

Burry is not the only one. Jamie Dimon, the CEO of JPMorgan Chase, told Bloomberg in May 2026 that the markets have "too much exuberance."5 The head of the Bank of England warned the House of Lords in October 2025 that recent corporate blowups might be "the canary in the coal mine."6 The Financial Stability Board, the global watchdog made up of central banks and finance ministries from the world's largest economies, published a report in May 2026 saying it cannot see into the fastest-growing parts of the market.7

This is not a few cranks on the internet. This is the biggest names in finance, saying out loud, in public, that the music is about to stop. And the structure built since 2008 is much bigger than the one that came down in 2008.

I. The Crash That Never Ended

The story you've been told about 2008 goes like this. Banks made loans to people who couldn't pay them back. They bundled those bad loans into fancy bonds and sold them as if they were safe. They placed side bets on whether the bonds would fail. When houses stopped going up in value, the whole house of cards came down. Lehman Brothers went under, the government rescued the rest, and Congress passed new laws. Lesson learned. Crisis over.

None of that is true.

What actually happened is that nobody tore the machine down. They just took the "Home Mortgage" sticker off the front and slapped on new ones: "Used Cars," "Office Buildings," "Business Loans," "AI Data Centers," "Private Stuff We Won't Tell You About." Same machine. Same wiring. Same way of failing. Different names on the boxes.

And the machine got much bigger. Here are the headline numbers, as of May 2026, every one of them sourced from the actual regulators or from the industry's own filings.

The total stack of side bets in the global financial system is $846 trillion.2 In 2008, when the world almost ended, it was $586 trillion.1 The setup is now forty-five percent larger than it was at the peak of the last bubble.

The bond product that brought down the world economy in 2008, the one that bundled and sliced loans into tranches, is back at $1.4 trillion in the U.S. alone. It has roughly doubled in size since 2018.8 The investment-bank version of that product, the one Michael Burry shorted in The Big Short, was renamed and is now back at over a quarter of the entire market.9

A new asset class called "private credit," which barely existed in 2010, now holds between $1.5 trillion and $2 trillion globally. By 2028 it is projected to hit $3.5 trillion.710 Big U.S. banks are tangled up in that world to the tune of at least $220 billion, and possibly as much as $500 billion. The regulators publicly admit they can't tell the real number because they don't have the legal power to make the funds report.7

A single brand-new asset class, AI data centers, will require somewhere between $2.9 trillion and $7 trillion in spending between now and 2030, depending on which big bank you ask. JPMorgan says $5.3 trillion.11 Morgan Stanley says of the $2.9 trillion needed in the next three years, $1.5 trillion of it has to be borrowed, because even Google, Microsoft, Meta, and Amazon don't have enough cash to pay for it themselves.12 Private credit funds are projected to provide $800 billion of that borrowing in the next two years alone.12

Late car payments on the kind of car loans that get bundled and sold to investors hit 6.9 percent in January 2026. That is the worst number ever recorded, going back to the early 1990s. Worse than during the 2008 crash itself.13

Empty office buildings are missing payments at 12.34 percent. That is also worse than anything seen during the actual financial crisis.14

$1.4TU.S. CLO market (~2x 2018)
$2TPrivate credit, FSB May 2026
$5.3TJPMorgan AI buildout estimate to 2030
6.9%Subprime auto ABS 60+ delinq, Jan 2026 (record)
12.34%Office CMBS delinquency, Jan 2026
$800BProjected private credit to AI by 2027

These are not warnings about something that might happen. This is what is happening right now, while you are reading this. The 2008 crash was not a one-time event. It was the moment the public got to see what was already broken. Once the cameras went away, the same people went back to doing the same thing, just in different rooms, with much bigger numbers.

This essay does three things. First, it shows you the machine is still running. Second, it shows you where the machine moved to, and how much bigger it is now than it was in 2008. Third, it shows you exactly where the next break will come from, because the cracks are already showing.

II. What the Machine Actually Is

To see that the machine is still running, you need to know what it actually looks like. The news has spent fifteen years calling the whole thing "those toxic CDO things," and that fog has made it impossible for normal people to spot the same machine when it shows up wearing a different shirt.

The machine has five parts.

Part One: Someone Who Makes Loans But Doesn't Keep Them

A mortgage broker writes you a home loan. The minute that loan is written, the broker sells it to someone else. The broker gets paid. The risk that you can't pay it back belongs to whoever bought the loan. So the broker doesn't really care if you can pay it back. The broker just wants to write as many loans as possible. This is the guy in the bar. He didn't care if I could afford $85,000. He just wanted the signature. This is why standards collapsed before 2008. The people checking whether borrowers were good for the money had no reason to check carefully.

Part Two: A Place to Dump the Loans

A bank creates what is basically a legal box. It buys thousands of loans and puts them in the box. Then it sells slices of the box to investors. The slices are stacked: top slice gets paid first, bottom slice gets paid last. The top slice gets a top safety rating (AAA) from a rating company. The rating company is paid by the bank that built the box. Yes, you read that right. The people grading the test are paid by the people taking it.15 This is legal. It is still legal.

Part Three: Side Bets

Once the box exists, people can place bets on whether the loans inside it will fail. They don't have to own any of the loans to place the bet. So one $1 million loan can have $50 million in bets riding on it. In 2006, there were about $1.2 trillion in actual risky home loans in America. There were over $5 trillion in side bets riding on those loans.16 When the loans went bad, every bet had to pay out. The system did not have $5 trillion lying around. That is why everything blew up.

Now multiply that. In 2026, the entire global side-bet market is $846 trillion.2 The American economy is worth about $29 trillion a year. The side bets are now twenty-nine times the size of the entire country.

Part Four: Buyers Who Don't Know What They're Buying

Your pension fund. Your city's retirement system. Your kid's college savings. Foreign governments. Insurance companies you've paid premiums to for thirty years. They all bought these top-rated slices because their rules said they could only buy top-rated things. They didn't have access to the list of actual loans inside the box. They just had the rating sticker. They were buying the sticker, not what was under it.

Now there's a new twist. Regular people can buy this stuff directly through their brokerage apps. A category called "CLO ETFs" went from $120 million in 2020 to over $30 billion by the end of 2025.17 When the next crash hits, regular people will watch it happen on their phones in real time.

Part Five: Nobody in Charge

All of this happens in private deals between two parties. There's no public list. There's no exchange. The side bets in particular were specifically excluded from government oversight by a law passed in 2000.18 State insurance regulators didn't even know AIG had written hundreds of billions in side bets until AIG was already underwater.15

The Financial Stability Board's May 2026 report says, in plain English, that the regulators today still cannot see most of the private credit market. They cannot identify which funds are which in their data. They don't have loan-by-loan information. They face legal barriers to sharing what little they have with other countries.7

That's the machine. Five parts. Four of those five parts are still fully running today, at scales much bigger than 2008. Only the home-mortgage end got any real fix, and even that fix is being quietly undone through back doors with names like "credit risk transfer."7

III. Where the Machine Lives Now

A. Business Loans Instead of Home Loans

The biggest hiding spot. The machine now uses loans to companies instead of mortgages on houses. Everything else is identical. The slicing, the rating game, the don't-care-because-I-sold-it-already attitude, all of it. This market is now $1.4 trillion globally. Twice what it was in 2018.8 In 2025 alone, $205 billion in new ones got created in the U.S.19

And the underwriting (the careful checking of whether the borrower can actually pay) is collapsing just like it did with home loans before 2008. The percentage of these business loans that strip out the lender's right to act when things go bad has risen from about 16 percent to 25 percent in just four years.20 In plain English: the people lending the money have signed away their right to step in when the company starts to fail. They have to wait until the company actually falls apart, which by then is too late. They did this voluntarily, because everyone else was doing it and the loan wouldn't get made otherwise.

B. The Big Short, Now With a New Name

After 2008, the word "synthetic CDO" became poison. Nobody wanted to be seen near one. So Wall Street did what Wall Street does. It changed the name. The new name is "bespoke tranche opportunity."21 That sounds nice. Like a custom suit. It is exactly the same product. A custom-built tower of side bets, designed for one specific big investor, not graded by the big rating companies, traded in private with no public record.

By 2018, trading in these had recovered to over $200 billion a year. By 2019, it had grown another 40 percent.22 By 2026, this kind of side-bet product makes up more than a quarter of the entire market.9 The credit-default-swap market that powered all this in 2007 (the actual technical name for the side bets) hit $9.2 trillion in late 2024, and grew 23 percent in just the first half of 2025.23 That is the fastest growth of any side-bet category in the entire financial system right now.

Wall Street's defense is that today's bets are placed on safer companies, not subprime home loans. That defense misses the point. The danger was never the average quality of what was being bet on. The danger was the leverage, the secrecy, and not knowing who was on the other side of which bet when the music stopped.

C. Used Car Loans: The Most Honest Copy

If you want to see the 2008 model in its purest form, look at how Wall Street is packaging used car loans for people with bad credit. The legal paperwork reads like someone did a find-and-replace on a 2006 mortgage document. The slicing is identical. The top slices get AAA ratings. The bottom slices get junk ratings. The middle slices get sold to whoever will buy them. Roughly $41.5 billion in these got created in 2025.24

Remember the credit ghost at the Jeep dealership. The same system that wouldn't touch me for $18,000 of credit on a new car is now packaging junk car loans, made to people the system already wrote off, and stamping the top slices AAA.25 The buyers of those AAA slices are pension funds and insurance companies that trust the sticker. The credit ghost in the dealership is the raw material. The pension fund is the mark. The bank in the middle takes the fee on both sides.

And the loans are failing. Right now. The late-payment rate on the bad-credit slice hit 6.9 percent in January 2026. That is the worst number ever recorded. Worse than 2008. Worse than 2010.13 The gap between people with good credit (almost nobody late) and people with bad credit (one in fifteen seriously late) has grown to more than ten times over.13 Among the worst-credit borrowers, the late rate is 76 percent higher than it was just before COVID. When a bad-credit borrower falls 60 days behind, more than three out of four of them never catch up.26 The cars get repossessed and resold at auction, and the recovery rate (how much money the lender gets back) has dropped to about 33 cents on the dollar.24

The same banks doing this today, Wells Fargo, Citi, Deutsche Bank, BMO, are the same banks that did it before 2008. Nobody has stopped them. The deals keep clearing. Investors keep buying.

D. Empty Office Buildings: The Crash Already Happening

The crash in office buildings is not coming. It is here. They are just refusing to call it that.

Office buildings packaged into investor bonds are missing payments at 11 to 12 percent. The previous worst-ever reading, hit in the long hangover from 2008, was 11.01 percent. The current reading is worse.14 Almost a trillion dollars in commercial real estate loans came due in 2025, which is three times the normal yearly amount.27 Office vacancy is running at 20 percent in the country as a whole.27 Buildings are selling for 90 percent off their peak prices.28 A 45-story building in midtown Manhattan with a $1.04 billion mortgage couldn't make its final payment in August 2025. Instead of admitting the building was worth less than the loan, the bank just extended the loan by three years and quietly took the missed payment off the books.29

This is not solving the problem. This is hiding the problem. It is exactly what banks did in late 2007 with subprime mortgages: refuse to write down what they were holding because nobody wanted to be first to admit the truth. The difference this time is that the buildings are mostly owned by retirement funds, insurance companies, and bond funds rather than banks. So when the losses finally have to be admitted, they will land on retirees, pensioners, and policyholders. Not on Wall Street. The optics of that distribution will not be better than 2008. They will be much worse.

E. Private Credit: The Part Nobody Can See Into

The fastest-growing and most secretive part of the new machine. "Private credit" means loans made to companies by funds that aren't banks. Pension funds and rich investors give money to these funds, the funds lend that money to companies, and the loans never appear on any bank's books. The global size of this market is between $1.5 trillion and $2 trillion.7 It barely existed in 2010. In fifteen years it has grown bigger than the entire U.S. junk-bond market.10 By 2028 it is projected to hit $3.5 trillion.30

These private loans then get packaged and sliced exactly like 2008 mortgages. That packaged version is now at $155 billion in the U.S.7 The smaller, sketchier rating companies that nobody had heard of before are taking business from the big rating companies, because they're more willing to hand out favorable grades.20 The 2005 problem, in a 2026 outfit, with new players who learned nothing.

About half of all private credit borrowers are also borrowing from regular banks. The banks have $220 billion to $500 billion tangled up in this world.7 A February 2026 Federal Reserve study found that up to one out of every four dollars banks lend to non-bank financial institutions now goes to private credit firms. In 2013, that number was one out of every hundred.31 The regulators publicly admit they cannot tell what's a private credit fund and what isn't in their own data.7

Major life insurance companies, the ones that hold your life insurance policy and your annuity, have nearly $1 trillion tied up in private credit.32 If private credit hits trouble, that hits insurance company solvency. That hits your retirement.

This is August 2007 with a different costume. Deep connections, thin information, real but unmeasured leverage, regulators on the record saying they don't know what they don't know.

F. The AI Data Center Stack: The Biggest and Newest Hole

This one didn't exist in 2008. It barely existed in 2023. It is the largest single buildout in the history of capitalism.

The big tech companies, Microsoft, Google, Meta, Amazon, Oracle, are building computer warehouses (called "data centers") to run artificial intelligence. The estimated cost of this buildout is between $2.9 trillion (Morgan Stanley) and $7 trillion (McKinsey) by 2030.1233 JPMorgan's middle estimate is $5.3 trillion.11 Pick any of those numbers. They are all bigger than the entire U.S. housing market was when the housing bubble burst.

Big tech alone cannot pay for this. So they are borrowing. Hyperscalers (the industry's name for Microsoft, Google, Meta, Amazon) issued $121 billion in bonds in 2025. That is more than four times their five-year average.34 AI-related borrowing accounted for thirty percent of all investment-grade corporate bond issuance in 2025.35

But that's the small part. The big part is hidden in private credit. There are already $200 billion in private credit loans to AI-related companies. Morgan Stanley projects that number will hit roughly $1 trillion in two years.12 The Bank for International Settlements thinks it will reach $300 to $600 billion by 2030.36 These numbers are not in agreement because, again, nobody can see into the market clearly.

Here is how the actual deals work, and this is where it gets ugly.

A company nobody has heard of (TeraWulf, for example, which used to mine bitcoin) raises $3.2 billion in bonds to build a data center. The bonds are rated junk. The way the deal works is that another company called Fluidstack signs a ten-year lease to use the data center to run AI computers for unnamed customers. Google, in the background, guarantees Fluidstack's lease payments. In exchange, Google gets a chunk of TeraWulf's stock.37 Google has now backed more than $5 billion of these deals for former bitcoin miners "pivoting to AI."38 If the AI customers don't show up, or if AI revenue does not justify the lease payments, the entire chain breaks. Junk-rated former bitcoin miner cannot pay bondholders. Fluidstack cannot pay lease. Google has to honor the backstop. Bondholders are pension funds.

Or another structure. Meta builds a $30 billion data center in Louisiana. The debt is not on Meta's books. Instead, a separate legal entity holds the debt, and that entity gets paid by a long-term lease from Meta. The entity then borrows against the lease. If AI revenue eventually disappoints Meta enough that Meta wants to walk away, the lease is the only thing standing between the bondholders and zero. This deal, nicknamed "Beignet," is part of a $30 billion package.39 Aligned Data Centers got bought by a consortium including Nvidia, Microsoft, BlackRock, and Elon Musk's xAI for $40 billion.40 Oracle alone has been putting together $38 billion data center deals tied to its $500 billion Stargate AI contract with OpenAI.41

And then a third layer goes on top. Because private credit loans usually have floating interest rates, but data centers need fixed-rate financing to make the math work, the operators are taking out side bets called "pay-fixed swaps" to convert the floating rate into a fixed rate. The Dallas Federal Reserve estimated in February 2026 that anecdotal evidence suggests $50 billion or more in this kind of side-bet swap was put on in just the fourth quarter of 2025.42

5+Layers of leverage in the AI data center stack
$121B2025 hyperscaler bond issuance (vs. $28B 5-yr avg)
$50BQ4 2025 AI-related pay-fixed swaps (Dallas Fed est.)
$40BAligned Data Centers acquisition, Oct 2025

So now read what is actually being built. Layer one: AI revenue, which is unproven and which Burry, Dimon, and a growing number of analysts believe is wildly overestimated. Layer two: long-term leases from tech companies, which depend on layer one. Layer three: junk-rated bonds and private credit loans backed by those leases. Layer four: financial side bets on top of those loans to fix the interest rates. Layer five: pension funds, insurance companies, and bond ETFs holding the paper. And then under all of it, regional banks lending to the private credit funds that originated the loans.

In 2008, the stack was two layers: bad mortgages, and side bets on those bad mortgages. The current AI data center stack is at least five layers, and that is just the visible part. Burry's central claim, and Dimon's, and the FSB's, is the same. The further up the stack, the more confident everyone is that the bottom is solid. The bottom is unproven AI revenue.

IV. The Same Cracks, In the Same Places

The 2008 machine broke in five specific places. All five of those places exist in the current machine. Some of them are worse now than they were then. Knowing where to look is how you know what to watch.

Crack 1: Nobody Checks the Borrower Anymore

Banks don't keep the loans they make, so they don't care if the borrower can pay. Private credit funds make loans and package them within months, so they don't care either. In September 2025, a Texas company called Tricolor Holdings collapsed. Tricolor sold used cars and made car loans, mostly to undocumented immigrants. It turned out Tricolor had been pledging the same cars as collateral to multiple lenders at the same time. That is straight-up fraud. JPMorgan, Fifth Third, and Barclays were all lending to Tricolor and none of them caught it before the company filed for bankruptcy.43

Same month, a company called First Brands Group, which makes aftermarket auto parts, also collapsed. First Brands had been hiding its real debt load through accounting tricks called "off-balance-sheet financing." Banks thought they were lending to a company with manageable debt. The actual debt was four times what they thought. The Swiss bank UBS lost about $500 million. The U.S. firm Jefferies lost about $715 million. Total damage is somewhere between $10 billion and $50 billion.44

A third blowup, involving BlackRock and HPS Investment Partners, was reported on November 1, 2025. Similar pattern: receivables-backed loans that defaulted with claims of manipulated reporting.44

JPMorgan CEO Jamie Dimon's reaction: "When you see one cockroach, there are probably more."44 The head of the Bank of England said the open question is whether these are one-offs or canaries.6 The industry's official answer is: one-offs. The industry's official answer in May 2007 was that subprime mortgage failures were one-offs. The industry's official answer in August 2007, three weeks before everything cracked, was that the French bank BNP Paribas freezing three of its funds was a one-off. The industry has a long, documented history of describing the first warning shots as a series of unrelated coincidences.

Crack 2: The Rating Companies Still Work for the People They're Grading

The conflict of interest that the official 2011 report on the financial crisis called out as a main cause of 2008 has not been fixed.15 Rating companies are still paid by the people whose products they rate. The rules added after 2008 made them disclose more but didn't touch the basic problem.

Worse, in the private-credit world there are now smaller rating outfits, the ones nobody had heard of before, that compete with the big three by being more willing to slap nice grades on things. These smaller outfits are taking market share precisely because they are easier graders.20 This is exactly how junk subprime mortgage slices got dressed up as AAA bonds in 2005, 2006, and 2007. Bad credit goes in, AAA paper comes out, pension funds buy it because their rules say they can only buy AAA. The actual risk hasn't changed at all.

Crack 3: Nobody Can See In, and the Regulators Admit It

The Financial Stability Board's May 2026 report says it directly: "Most FSB member authorities cannot separately identify private credit funds in their regulatory reporting frameworks, lack granular loan-level data, and face legal impediments to sharing what limited information they do collect."7 In plain English, the people whose job it is to watch the system are saying out loud that they can't see most of it.

The sharpest moment of the 2008 crisis was the moment everyone realized nobody knew who held what. Lehman Brothers went bankrupt owing $155 billion. Then it turned out there were $400 billion in side bets riding on Lehman that nobody had publicly tracked.15 The 2026 setup reproduces that exact blindness, but at a scale almost three times larger.

Crack 4: The Side Bets Stack Higher Than the Real Risk

This is the place where the current crisis dwarfs 2008.

Global side bets stood at $586 trillion at the peak before 2008.1 They stand at $846 trillion today.2 That is $260 trillion in extra bets, more than the entire global economy combined. The growth rate in the first half of 2025 was the fastest since 2008.2

Credit-related side bets specifically (the ones that destroyed AIG) grew 23 percent in the first six months of 2025 alone.23 Initial money posted as collateral on these bets at major clearinghouses reached $430 billion, up from $364 billion a year earlier.23 Those numbers cover only the publicly visible side bets. The custom, private side bets, the ones on private credit and the bespoke tranches and the bank risk-transfer deals and the AI data center swaps, are not in those numbers. Those are the bets that brought down AIG. The Dallas Fed estimated $50 billion in just AI-data-center-related swaps in three months of 2025.42 That's one slice of one corner of one new asset class. The full bespoke-side-bet pile is unknown.

The lesson from AIG was not that all side bets are bad. The lesson was that side bets become catastrophic when one company quietly ends up on the other side of all of them. In 2026, the regulators don't know who is on which side of which bet for most of the new asset classes. They have said this in writing.7

Crack 5: Regular People Are Now Buying This Stuff

In 2008, the worst paper was held mostly by big institutions. Regular people got hit by 2008 the indirect way, through job losses and home values and 401(k) crashes. But they didn't own the toxic paper directly.

That has changed. Wall Street has spent the last five years building products that let regular people buy this stuff through their brokerage accounts. The category called "CLO ETFs" went from $120 million in 2020 to over $30 billion by the end of 2025.17 In just 2025 alone, $15 billion flowed in. By October, $500 million was flowing out per week as worried investors tried to get back to the door.17 These funds trade by the second and can crash hard during the trading day in ways that don't even match what the underlying loans are doing.

When the next crash hits, it will hit regular people directly, in real time, on the screens of their phones. Not as background damage to the economy. As immediate, visible losses they can watch happen.

And the holders include your retirement plan. Major life insurance companies have nearly $1 trillion tied up in private credit.32 New York and Pennsylvania state pension plans are invested in Blue Owl's $7 billion digital infrastructure fund. That is the same fund holding the Meta Beignet deal and multiple Oracle Stargate financings.45 Public pensions are now exposed, through one or two degrees of separation, to whether AI generates enough revenue to justify hundreds of billions in data center buildouts that have not yet been monetized.

V. Who Actually Pays When This Blows

Now we get to the part nobody on financial television wants to talk about.

The story so far has been about the machine. About the architecture. About $846 trillion in side bets and AI data center stacks and bespoke tranche opportunities. None of that is what most people in this country live inside of. Most people in this country live inside a much simpler economy. They go to work. They get a check. They pay rent. They buy gas. They eat what they can afford. If something breaks, they hope the credit card has room.

So the question that matters is not "how big is the bubble." The question that matters is what happens to them when it pops. And the answer is different than it was in 2008. It is worse.

In 2008 People Lost Things. This Time They Have Nothing Left to Lose.

When the housing bubble cracked, people lost actual stuff. They lost home equity. They lost retirement accounts. They lost jobs that had benefits attached. They lost businesses they had built. The damage was severe, and a lot of those people never recovered. But the point is that they had things to lose in the first place.

That is not the situation in 2026. Look at the numbers.

Forty-three percent of Americans cannot pay a $1,000 emergency expense from savings. That number is from a January 2026 U.S. News survey.46 It matches what the Federal Reserve has been finding for years. Twenty-seven percent of Americans have no emergency savings at all, per Bankrate's 2026 report.47 The median emergency savings balance among people who have any is $5,000, which is half of what it was the year before.46 The personal savings rate in 2024 was under five percent. In 2020, during the pandemic checks, it briefly hit thirty-two percent.48 That money is gone now. It got spent on rent and groceries.

Eighty-three percent of hourly workers in the U.S. have less than $500 in savings.49 The median Gen Z emergency fund is $400. Total.49 That is one transmission repair away from collapse. That is one ER copay away from collapse. That is one missed week of work away from collapse.

Wealth concentration is at the widest gap in more than thirty years. As of the third quarter of 2025, the top one percent of Americans held about $55 trillion. That is roughly equal to the combined wealth of the bottom ninety percent.50 The top ten percent now accounts for forty-nine percent of all consumer spending in the country, which is the highest share since the data series began in 1989.50 The top 0.1 percent has doubled its wealth since the pandemic.51 There are 905 billionaires in the U.S. with a combined $7.8 trillion. The bottom fifty percent of households, 66 million households, have $4.1 trillion combined.52

43%Can't pay $1,000 emergency from savings
$400Median Gen Z emergency fund
$55TTop 1% wealth = bottom 90% combined
905U.S. billionaires, $7.8T combined
49%Top 10%'s share of all consumer spending
$4.1TTotal wealth of bottom 66M households

So when the next crash hits, the question of "who loses what" has a very different answer than it did in 2008. The top is bigger and more concentrated than it has ever been. The bottom has been hollowed out. There is no margin left to absorb anything. The 2008 crash hit a working class that was already battered but that still had something to lose. The 2026 crash will hit a working class that is operating on fumes and credit card debt with no air gap between paycheck and homelessness.

The Crash Will Not Feel Like a Crash. It Will Feel Like a Tightening.

People imagine a financial crisis as a single dramatic event. A bank fails. A stock market goes red. The president gives a speech. There is a bailout vote. That was 2008's visible part. For ordinary people, even in 2008, the experience was different. It came in slow, ugly waves over five years. Hours cut. Bills past due. The car gone. The house gone. The job applications nobody answered. The shame.

In 2026 it will be worse, because it will be more diffuse and more invisible. The crash will not happen on television. It will happen in the cracks of daily life.

It will feel like rent going up another $200 with no explanation. It will feel like the grocery bill creeping up while the bag gets smaller. It will feel like the boss "restructuring" your hours from 35 to 28 so you stop qualifying for the company health plan. It will feel like the dental work you needed last year that you still have not been able to get. It will feel like the side gig paying less because there are more drivers and fewer riders. It will feel like the credit card minimum payment going up because the rate adjusted again. It will feel like the eviction filings on your block getting more frequent. It will feel like the utility company adding a "service charge" that nobody can explain. It will feel like the urgent care visit you walked out of without being seen because the copay was $150 you didn't have.

It will feel like the air slowly running out.

This is what "downward mobility" actually looks like. You don't fall off a cliff. You get squeezed, every day, in every direction, by a system that needs you to keep working at the exact level of desperation that makes you compliant. The crash does not produce one big number on the news. It produces ten million small daily compressions that nobody adds together.

The People Who Built the Machine Will Buy Up What's Left

In 2008, the housing crash produced one of the largest transfers of wealth in American history. The houses people lost did not disappear. They got bought. By Blackstone. By private equity firms. By corporate landlords. By LLCs whose owners regular people will never see. As of 2024, large institutional investors own more single-family rental homes than at any point in U.S. history.53 Those houses used to belong to families. Now those families pay rent to the funds that bought their foreclosed homes for thirty cents on the dollar.

This is the part that does not get said out loud. A crash, for the people at the top, is not a disaster. It is an acquisition opportunity. When prices collapse, the people with cash buy what the people without cash are forced to sell. The wealth does not vanish. It changes hands, upward.

The structure of the current setup makes this worse than 2008. Private equity is sitting on roughly $2.5 trillion in undeployed cash, the highest number in history.54 Private credit funds have raised hundreds of billions specifically to deploy into "distressed" situations. The largest private equity firms have publicly described, on their own earnings calls, that they are positioned to buy assets at "dislocation pricing" when the cycle turns.

What does that mean in plain English? It means when your landlord can no longer pay the mortgage on the apartment building, your landlord doesn't go away. The bank takes the building and sells it to Blackstone for forty cents on the dollar. Blackstone keeps you in the apartment because they need the rent. They raise your rent. They cut maintenance to zero. You can't move because everywhere is the same. That is what the crash does to you.

When the office buildings finally hit the wall, the funds with cash will buy them at 90 percent off and convert them to "luxury micro-units" that rent for more per square foot than the offices ever did. When the bad-credit auto lenders blow up, the cars get repossessed and resold, and the funds that bought the debt portfolios collect the cash. When the regional banks fail, JPMorgan buys them in an FDIC-assisted deal at a discount, like it did with First Republic in 2023.55

The crash is not a threat to the people who built the machine. It is the way the machine eats. Every cycle, the bottom gets thinner. Every cycle, the top consolidates. That is not a flaw in the design. That is the design.

The Response to the Crash Will Be Bailouts for Them, Discipline for You

Watch what happened in 2008, because it is going to happen again, with more cynicism and less embarrassment.

The Federal Reserve and the Treasury, working together, provided trillions in support to the financial system. Bear Stearns got bought by JPMorgan with $30 billion in Fed backing.56 AIG got nationalized in everything but name. The TARP program put $700 billion of public money into the banks.57 The Fed's balance sheet went from $900 billion to $4.5 trillion.58 Goldman Sachs and Morgan Stanley were converted into bank holding companies overnight so they could access Fed lending.59 The big banks paid out tens of billions in bonuses the same year they received the bailout money.60

The homeowners got HAMP. The Home Affordable Modification Program. It was advertised as $75 billion in relief.61 The actual amount of money that reached homeowners was a small fraction of that. Most of it went to mortgage servicers as fees. Most of the modifications that did get processed re-defaulted within two years.61 The program is widely understood, by economists across the political spectrum, as a deliberate failure designed to keep the appearance of help while actually slow-walking foreclosures so the banks could process them in an orderly way.

That is the template. The institutions get rescued at full speed. The people get rescued through a paperwork maze designed to deny them. The institutions get cash injections. The people get means-tested programs with documentation requirements they cannot meet because they are working two jobs and do not have time to wait on hold for four hours.

In 2026 there is now an additional twist. The current administration is more openly hostile to public assistance than the 2008 administration was. SNAP eligibility has been tightened. Disability claim processing has slowed. The pending changes to Medicaid will remove millions from coverage.62 Student loan forbearance has been ended. The push to means-test everything and require work documentation for benefits is well underway before the crash hits. So when the crash hits, the safety net will be smaller, narrower, and harder to access than it was in 2008. The bailouts for the financial sector will be larger. That is not a coincidence. That is the policy.

And Then the Narrative Will Blame You

This is the part to watch for, because it always happens and people always fall for it.

When the dust settles, the story will not be that the financial industry built a $846 trillion stack of side bets on top of trillions in unproven AI revenue, mispriced consumer credit, and empty office buildings, and the whole thing came down because that is what those structures do. The story will be that the American consumer borrowed too much. That millennials wasted their money on avocado toast and DoorDash. That Gen Z spent too much on streaming subscriptions. That the working class made bad choices. That entitlements are unsustainable. That we need "fiscal discipline."

You will hear it from cable news. You will hear it from talk radio. You will hear it from your uncle at Thanksgiving. You will hear it from politicians of both parties. The story will be the same one told after every previous crash. The people who built the machine that broke will be described as "stewards of stability" who got caught in an unfortunate market cycle. The people the machine ground up will be described as irresponsible consumers who lived beyond their means.

You are not poor because you are irresponsible. You are poor because wages have not kept up with cost of living for 50 years, because housing was turned into a speculative financial product, because manufacturing was offshored so corporate margins could go up, because health care costs three times what it does anywhere else in the developed world, because higher education was financialized into a debt trap, and because every part of daily life has been turned into a subscription that rents you the things people used to own.

You are not in debt because you are bad with money. You are in debt because the economy is structured to force you to borrow for everything. Rent eats half your income because someone decided housing should generate returns. The credit card has a balance because food and gas and the car repair all happened in the same month. The medical debt is there because you got sick and the hospital charged $4,000 for a CT scan that costs $200 to perform.63

You are not lazy. You are exhausted because you are doing the work of two people for the wages of less than one, with no benefits, no security, no path up, and no slack. You are exhausted because that is the design.

The system fails you, and then it blames you for failing. That is the cycle. It has run every time. It is going to run again. And the only defense against it is to see it coming.

VI. What the Crash Will Actually Look Like

Real crashes don't announce themselves. They get recognized later, with everyone wondering how the warning signs were so obvious in hindsight. The warning signs in May 2026 line up with the warning signs in early 2007 with frightening accuracy, and the scale is much larger.

Here is how it goes from here.

The "one-off" failures keep coming. Tricolor and First Brands weren't the first. There was Greensill in 2021.64 There have been smaller used-car lender blowups. The failures cluster in specific places: bad-credit consumer lending, supply-chain finance, asset-based lending where the collateral turns out to have been pledged to multiple lenders, and mid-sized companies whose loan agreements have stripped out the lender's ability to do anything until it's too late.

A few of those failures hit the same private credit funds, or the same business development companies, or the same packaged-loan pools. Investors start to realize names that the rating companies treated as independent are actually all connected. Prices on the riskier slices of the packaged-loan products start dropping. Then prices on the middle slices. Then on the supposedly safe slices. The top-rated slices keep their ratings, but nobody will pay full price for them anymore. The investment funds holding the riskier slices have to sell to meet investor withdrawals, but nobody wants to buy. That forces more selling, which forces more withdrawals. This is the "run" that the FSB has openly warned about in its public reports.7

At the same time, the office-building "extend and pretend" runs out of road. The three-year extensions handed out in 2025 expire in 2027 and 2028. The owners who couldn't pay in 2025 still can't pay. The bank servicers who have been kicking the can down the road give up. The losses go to bondholders, who are insurance companies, pension funds, and bond mutual funds. The usual emergency tools the Federal Reserve uses (lending to banks, swapping currency with other central banks) don't reach those holders. The Fed will have to invent new tools or watch the losses happen.

Used-car loan failures, already at record highs, get worse, because they always get worse in a recession. The packaged bonds based on those loans start losing principal on their middle slices. Rating companies, just like in 2007 and 2008, downgrade too late. The downgrades trigger automatic rules in the packaged business-loan products, forcing those products to dump loans onto the market. Loan prices fall. Junk bond prices fall with them. The pressure spreads through every market that touches business credit.

Then the new piece. AI revenue, the thing the entire data center buildout is being financed against, comes in lower than projected. Or one of the AI tenants renegotiates a lease. Or a "neocloud" data center operator (one of the former bitcoin miners that pivoted) cannot make its bond payments. The lease that backed the bond is suddenly worth less. The bond is downgraded. The private credit fund that lent against the bond marks its portfolio down. The pay-fixed swap that was supposed to hedge the floating rate is suddenly mis-marked, because the assumptions behind it no longer hold. Investors in the private credit fund try to withdraw. They cannot, because the fund has gated redemptions. Their losses are now stuck. They cascade up to the pension funds and life insurers that own units in the fund.

And somewhere in all of this, a bilateral side bet that someone had recorded on their books at a happy fictional value turns out to have been the wrong number. Or the company on the other side of that bet turns out to not have the money to pay. Or one company turns out to have been the secret counterparty to far too many of these bets. That is the AIG moment. There is no way to predict where it will come from, because, as the FSB has said publicly, the data is not public.7 The total pile of side bets is $846 trillion, fifty percent bigger than it was when AIG nearly took the world down in 2008.2

VII. The Indictment

This is not a forecast. This is an indictment.

The structures that caused 2008 were not accidents. They were the product of a specific set of incentives, a specific set of legal arrangements, and a specific set of regulatory blind spots. All of those incentives, arrangements, and blind spots have been preserved. They have been expanded. They have been copied into new corners of the economy, and stacked on top of each other in ways that did not exist in 2008. The fixes Congress passed after 2008 touched a few symptoms. The fixes did not touch the disease.

The disease is this. There is an industry whose entire job is to manufacture credit risk that is priced wrong on purpose, and then sell that mispriced risk to buyers who are forced by their own rulebooks to buy it. The industry's profit comes from the gap between what the loans are actually worth and what the rating sticker on them says they are worth. The 2008 housing bubble was one product of this business model. Today's bad-credit car loans, office buildings, business loans, private credit funds, bespoke side bets, and the entire AI data center build out are all products of the same business model. The wrapping paper changes. The product is the same.

But the scale changed too. The total stack of side bets is forty-five percent bigger than it was at the 2008 peak. The fastest-growing slice of that stack, credit side bets, grew 23 percent in just six months in 2025. A new asset class worth several trillion dollars (the AI data center buildout) has been bolted onto the system, and it is being financed through a stack of leveraged structures at least five layers deep, on top of revenue assumptions that the most successful crash-caller in living memory has bet over a billion dollars against.

The product always breaks the same way. The people writing the loans don't care if the loans get paid back, because they sell the loans the same week they make them. The rating companies hand out grades they know are too high, because the people they grade are the people paying them. Secrecy continues because secrecy is profitable for the people on the inside and harmful only to the people on the outside. Side bets pile up to many times the size of the real underlying risk, because no law caps how many bets can be placed on the same thing. Pension funds and insurance companies hold the paper because their rulebooks make them, not because anyone read the fine print.

Every one of those failures is happening right now, in May 2026, at full scale, across multiple parts of the economy, in a financial system whose own top regulator has officially admitted it cannot see into the fastest-growing parts of the market. The direction this is going is not in question. Only the date of the explosion and the name of the first company on the front page are unknown. And when it happens, the public will be told, just like in 2008, that no one could have seen it coming, that the experts were blindsided, and that the system has to be rescued because the only other option is total collapse.

That story will be a lie. It will be a lie in the same way the 2008 version was a lie. The machine was visible. The warning signs were written down by the regulators themselves. The financial press covered it. Burry called it on his Substack. Dimon called it on Bloomberg. The Bank of England Governor called it in front of the House of Lords. The few people who said so out loud in language regular people could understand were called doomers, were marginalized, or were paid to shut up. And when the dust clears, the same companies that built the machine will still be standing, holding more than they did before. The 905 billionaires will still be billionaires. The private equity funds will own more of the country than they did the day before the crash. And the people who lost their last $400 of savings, their car, their apartment, their kid's daycare slot, their dental work, their dignity, will be told that the lesson has been learned, the problem has been fixed, and the next version of the same machine, which is already being built right now under a different name, can be trusted.

It cannot be trusted. It is the same machine. It is bigger than it has ever been. The crash that is coming will not be a moment of loss for most people in this country. It will be a slow tightening of a trap that is already closing.

I started this essay with a guy in a bar in 2004 trying to hand me an $85,000 mortgage I had no business signing, while a car dealership across town wouldn't finance me on an $18,000 Jeep. Twenty-two years later, I am still a credit ghost. Fifty-five years old. And what I am worried about now is not me. I am worried about the children of the children. The kids who are walking into adulthood right now with no jobs, no housing, and no hope. The kids who will never see what a starter home is. The kids who are being told they are lazy for not building wealth in an economy that does not let them keep any.

The guy in the bar was not the idiot. The system that put him there was not stupid either. It was working exactly as designed. It is still working as designed, just bigger and quieter and meaner. The 2008 crash never ended. It just learned how to squeeze without making a sound.

All of us have nowhere left to fall.

Sources

Every claim in this document is sourced to a primary regulator, a primary industry filing, peer-reviewed financial press coverage, or the actor's own public statements. All URLs verified May 2026.

  1. Bank for International Settlements, OTC derivatives statistics, historical series showing peak 2008 notional outstanding of roughly $586 trillion. BIS Quarterly Review, December 2008. https://www.bis.org/statistics/derstats.htm
  2. Bank for International Settlements, "OTC derivatives statistics at end-June 2025," 8 December 2025: "The notional value of outstanding over-the-counter (OTC) derivatives rose to $846 trillion at June 2025, up 16% from June 2024. This marks an acceleration from the moderate 5% annual upward trend since end-2016... This 16% rise is the largest year-on-year increase observed since 2008, before the onset of the Great Financial Crisis of 2008-09." https://www.bis.org/publ/otc_hy2512.htm
  3. Michael J. Burry, Substack posts including 8 May 2026: "the market has jumped the shark" and "the end of... this... is nigh"; market comparison to March 2000 dot-com peak. Reporting via 24/7 Wall St., Moneywise, and Yahoo Finance, May 2026. https://moneywise.com/investing/stocks/michael-burry-short-stock-market-crash-ai
  4. SEC Form 13F filing, Scion Asset Management, Q3 2025: $1.1 billion in put options (~$912M Palantir, ~$187M Nvidia), comprising approximately 80% of fund notional portfolio. Burry projects 40-50% potential decline for major Big Tech. Motley Fool, Fortune, TheStreet coverage November 2025. https://www.fool.com/investing/2025/11/16/michael-burry-bets-against-nvidia-and-palantir-but/
  5. Jamie Dimon interview with Francine Lacqua, Bloomberg Open Interest, Paris, 12 May 2026: "I think there's a little bit too much exuberance out there... The stock market is in the top 15%, credit spreads are very low. The general assumption is that these things are all going to resolve. And I'm kind of a skeptic." https://www.bloomberg.com/news/videos/2026-05-12/jpmorgan-s-dimon-sees-too-much-exuberance-in-markets-video
  6. Bank of England Governor Andrew Bailey, testimony to House of Lords Financial Services Regulation Committee, October 2025, raising the question of whether recent corporate failures are "the canary in the coal mine." Cited in Addleshaw Goddard LLP, "Lessons from Tricolor and First Brands," 22 December 2025. https://www.addleshawgoddard.com/en/insights/insights-briefings/2025/financial-services/lessons-from-tricolor-and-first-brands/
  7. Financial Stability Board, Report on Vulnerabilities in Private Credit, 6 May 2026, including the cited statement on regulator visibility, $1.5-$2 trillion global market size, $220-$500 billion bank exposure range, and $155 billion U.S. private credit CLO market. https://www.fsb.org/uploads/P060526.pdf
  8. Bank of America research data via IQ-EQ, "Investor appetite for CLOs is surging," 11 September 2025: global CLO market $1.4 trillion, nearly doubled since 2018. https://iqeq.com/insights/investor-appetite-for-clos-is-surging-but-they-must-also-understand-the-risks/
  9. Business Research Insights, Collateralized Debt Obligation Market Report 2026-2035: synthetic CDOs comprising 27.8% of CDO market share. https://www.businessresearchinsights.com/market-reports/collateralized-debt-obligation-market-102741
  10. Maples Group, "Private Credit CLO Growth Accelerates," 16 July 2025, citing Preqin and LSTA data: private credit ~$1.7T end 2024, $2.64T projection by 2029. https://maples.com/knowledge/private-credit-clo-growth-accelerates
  11. JPMorgan projection of more than $5 trillion in AI data center and related power spending through 2030, reported by Bloomberg, "The $3 Trillion AI Data Center Build-Out Becomes All-Consuming for Debt Markets," 2 February 2026. https://www.bloomberg.com/news/articles/2026-02-02/the-3-trillion-ai-data-center-build-out-spurs-a-debt-market-boom
  12. Morgan Stanley research, cited in Brandywine Global Investment Management, "Brave New World of AI Capex: Giving Credit Where Credit is Due," November 2025: $2.9 trillion global data center capex 2025-2028, $1.5 trillion funding shortfall, split projected at $800 billion private credit, $200 billion corporate debt, $150 billion securitized. https://www.brandywineglobal.com/around-the-curve/2025/brave-new-world-of-ai-capex-giving-credit-where-credit-is-due
  13. Fitch Ratings subprime auto ABS performance data; Wolf Richter, "Serious Delinquency Rates for Subprime & Prime Auto Loans, Q4 2025," Wolf Street, 17 February 2026: 60+ day delinquency rate of 6.9% in January 2026, highest in data series since the early 1990s. Auto Remarketing reporting March 2026. https://www.autoremarketing.com/subprime/fitch-continues-dialogue-about-affordability-delinquency-with-latest-look-at-subprime-auto-securitizations
  14. Trepp data via Multi-Housing News, "2026 CMBS Delinquency Rates," March 2026, and Commercial Property Executive, "2025 CMBS Delinquency Rates," 25 March 2026: office CMBS delinquency 11.31% December 2025 / 12.34% January 2026, surpassing previous post-2008 peak of 11.01%. https://www.multihousingnews.com/cmbs-delinquency-rates/
  15. Financial Crisis Inquiry Commission, Final Report (January 2011): conflict of interest at rating agencies as primary causal factor; AIG financial-products unit invisible to state insurance regulators; $400B in Lehman CDS notional not publicly tracked at bankruptcy. Bethany McLean and Joe Nocera, All the Devils Are Here (2010). https://www.govinfo.gov/content/pkg/GPO-FCIC/pdf/GPO-FCIC.pdf
  16. Michael Lewis, The Big Short (2010), and FCIC documentation: $5+ trillion in synthetic exposure created off of $1.2 trillion underlying subprime loans by 2006; synthetic CDO market grew $15B (2005) to $61B (2006), peaking ~$5T notional year-end 2006. https://en.wikipedia.org/wiki/Synthetic_CDO
  17. 9fin data, cited in Lexology, "CLO transactions: Spring 2026 market trends and regulatory developments": CLO ETFs grew from $120M AUM in 2020 to $30B+ end 2025, with $15B inflows in 2025 and >$500M weekly outflows by October 2025. https://www.lexology.com/library/detail.aspx?g=2e0eccaa-6a37-43f7-a79b-8e8c14ec0d77
  18. Commodity Futures Modernization Act of 2000, Pub. L. 106-554, excluding credit default swaps from CFTC oversight. https://www.congress.gov/bill/106th-congress/house-bill/5660
  19. Octus, "Octus Identifies Leading Credit Trends That Defined 2025," 27 January 2026: U.S. CLO issuance reached $205 billion in 2025. https://octus.com/company/news/octus-identifies-2025-leading-credit-trends/
  20. Octus / S&P Global Ratings, "Private Credit CLO Market Sees Wave of Structural Innovations Amid Explosive Growth," 26 February 2026: covenant-lite share of middle-market CLOs rose from ~16% (2021 vintages) to 25% (Q1 2025); smaller "private letter" rating agencies taking market share with more favorable scores. https://octus.com/resources/articles/private-credit-clo-market-sees-wave-of-structural-innovations-amid-explosive-growth/
  21. Coryanne Hicks, "What Is a Bespoke Tranche Opportunity?" U.S. News, September 2023; OneMoneyWay, "Bespoke tranche opportunity: structure, benefits & risks," 19 February 2025: bespoke tranche opportunity ("BTO") as the post-crisis rebrand of single-tranche synthetic CDO. https://money.usnews.com/investing/articles/what-is-a-bespoke-tranche-opportunity
  22. Reuters reporting, cited in industry coverage of bespoke synthetic CDO volume: trading volume exceeded $200B annually by 2018 with ~40% additional growth in early 2019.
  23. International Swaps and Derivatives Association, "Key Trends in the Size and Composition of OTC Derivatives Markets in the First Half of 2025," 22 January 2026: CDS notional outstanding $9.2T late 2024; credit derivatives gross market value grew 46.9% to $251.8B; initial margin posted for cleared interest rate derivatives and CDS reached $430.4B mid-2025 vs. $364.4B year prior. https://www.isda.org/a/oSdgE/Key-trends-in-the-size-and-composition-of-OTC-derivatives-markets-in-the-first-half-of-2025.pdf
  24. Auto Remarketing, "Fitch continues dialogue about affordability & delinquency with latest look at subprime auto securitizations," 12 March 2026: subprime auto ABS issuance ~$41.5B in 2025 vs. $38.7B in 2024; Subprime Recovery Index 32.86% December 2025. https://www.autoremarketing.com/subprime/fitch-continues-dialogue-about-affordability-delinquency-with-latest-look-at-subprime-auto-securitizations
  25. Bridgecrest Lending Auto Securitization Trust 2025-1 and 2025-4 Form FWP filings, U.S. Securities and Exchange Commission: AAA senior tranches, mezzanine tranches stepping AA through BB. https://www.sec.gov/Archives/edgar/data/0002086742/000110465925099823/tm2528191d12_fwp.htm
  26. Sapling, "Why Auto Loan Delinquencies Record High Is a Subprime Problem," 24 March 2026, citing CUCollector securitized loan data: below-prime 60-day delinquency 76% higher in December 2025 than December 2019; chronic-delinquency share rose from 53% (Dec 2019) to 69% (Dec 2025). https://www.sapling.com/13779303/
  27. The Kaplan Group, "Is Commercial Real Estate at a Breaking Point in 2025?" 25 September 2025: ~$957B in commercial real estate loans maturing in 2025 (3x 20-year average); office vacancy ~20% nationally. https://www.kaplancollectionagency.com/business-advice/is-commercial-real-estate-at-a-breaking-point-in-2025/
  28. BigGo Finance, "U.S. Commercial Real Estate CMBS Delinquency Rate Soars Past 7.5%," 9 April 2026, documenting surrender sales at 90% discounts from peak valuations across major urban office markets. https://finance.biggo.com/news/aUARc50Bh5an-7GhrcEt
  29. Wolf Richter, "Office CMBS Delinquency Rate Spikes to Record 11.7%," Wolf Street, 1 September 2025, documenting the $1.04 billion mortgage at 1211 Avenue of the Americas, balloon payment failure August 2025, three-year extension to August 2028. https://wolfstreet.com/2025/09/01/
  30. BNY, "Optimizing Private Credit & CLO Operations for Efficiency," 1 May 2025: private credit projected $3.5T globally by 2028. https://www.bny.com/corporate/global/en/insights/private-credit-CLO-operations-efficiency.html
  31. Federal Reserve research, February 2026, on bank lending to non-bank financial institutions: share to private credit firms rose from approximately 1% (2013) to up to 25% (2026); referenced in Bloomberg and Federal Reserve communications. https://www.federalreserve.gov/econres/feds/2025.htm
  32. National Association of Insurance Commissioners data on insurance industry exposure to private credit and asset-based finance, approaching $1 trillion in life insurance industry holdings, cited across S&P Global, NAIC reports, and Federal Reserve commentary in late 2025 and early 2026. https://content.naic.org/sites/default/files/capital-markets-special-report-private-equity-investment.pdf
  33. McKinsey & Company AI infrastructure capex estimates approaching $7 trillion globally through 2030; widely cited across financial press coverage of AI data center buildout, including Bloomberg and S&P Global Ratings AI Market Update reporting Q1 2026.
  34. BofA Securities report, January 2026, cited in Reuters, "AI hyperscalers will drive higher US corporate bond supply in 2026, analysts say," 15 January 2026: five major hyperscalers (Amazon, Alphabet, Meta, Microsoft, Oracle) issued $121 billion in U.S. corporate bonds in 2025 vs. $28B/year average 2020-2024. Also Mellon Investments Corporation, "Record-Breaking AI-Related Debt Issuance in 2025," 15 December 2025. https://www.mellon.com/insights/insights-articles/record-breaking-ai-related-debt-issuance-in-2025.html
  35. M&G Investments, "Tech issues: The AI debt deluge hitting bond markets," March 2026: debt tied to AI ballooned to $1.2T by October 2025; AI now 14% of high-grade market, surpassing US banks as largest sector in JP Morgan US Liquid index. S&P Global, "Private Credit, Tech Issuance fuelled by AI, and Increasing Leverage Among Key Driving Factors," 17 February 2026: tech sector ~16.7% of global non-financial corporate bond issuance in 2025, up from 11.6%. https://press.spglobal.com/2026-02-17
  36. Bank for International Settlements analysis of private credit exposure to AI infrastructure, projecting $300-$600 billion in private credit to AI-related companies by 2030; cited in BIS Quarterly Review commentary and financial press, late 2025 and early 2026.
  37. Cooley LLP, "Fluidstack Announces Three Data Center Lease Agreements, Including Data Campus Expansion, With TeraWulf," 18 August 2025: $6.7B contracted revenue / potential $16B with extensions; Google backing $3.2B of Fluidstack lease obligations. SEC Form 8-K, TeraWulf Inc., 14 October 2025: $3.2 billion senior secured notes offering by WULF Compute LLC for Lake Mariner data center expansion, secured by first-priority liens including a pledge by Google LLC of TeraWulf warrants. https://www.sec.gov/Archives/edgar/data/0001083301/000110465925099020/tm2528632d1_ex99-1.htm
  38. SEC 8-K, TeraWulf Inc., 28 October 2025: additional $1.3 billion Google lease backstop for 168 MW Abernathy Texas joint venture, $9.5 billion total contracted revenue, bringing total Google support across former-bitcoin-miner deals past $5 billion. https://www.sec.gov/Archives/edgar/data/0001083301/000110465925102858/tm2529509d1_ex99-1.htm
  39. Bloomberg, "The $3 Trillion AI Data Center Build-Out Becomes All-Consuming for Debt Markets," 2 February 2026: Beignet Investor LLC structure for Meta Hyperion data center in Louisiana, $27 billion debt arranged by Morgan Stanley with PIMCO as anchor investor purchasing $18 billion, ~$2.5 billion equity from Blue Owl Capital, BlackRock funds purchasing $3 billion. NPR, The Indicator from Planet Money, 25 November 2025. https://www.npr.org/transcripts/nx-s1-5619276
  40. Aligned Data Centers press release, 15 October 2025: $40 billion equity acquisition by Artificial Intelligence Infrastructure Partnership (BlackRock GIP, MGX, Microsoft, Nvidia, with xAI, Temasek, and Kuwait Investment Authority added). https://aligneddc.com/press-release/ai-infrastructure-partnership-aip-mgx-and-blackrocks-global-infrastructure-partners-gip-to-acquire-all-equity-in-aligned-data-centers/
  41. Reporting on Oracle-OpenAI Stargate partnership, $500 billion contract value, including individual data center deals in the $38 billion range; coverage via TheStreet, CNBC, and Octus, 2025-2026. Oracle 2025 bond issuance for Stargate-related capex, including $18 billion September 2025 bond sale. https://www.thestreet.com/technology/michael-burry-shares-bold-predictions-for-openai-palantir
  42. Federal Reserve Bank of Dallas, "How AI debt financing impacts duration supply and interest rates," 10 February 2026: "Anecdotal evidence from market participants suggests that this [pay-fixed swap activity] may have been sizeable in the fourth quarter of 2025, possibly totaling as much as $50 billion in 10-year equivalents of bond issuance or more." https://www.dallasfed.org/research/economics/2026/0210-searls-aifinancing
  43. Bloomberg, "Is a Private Credit Crisis Brewing? Tricolor, First Brands Shake Wall Street," 31 October 2025; Brown & Brown, "From Collapse to Clarity," 17 November 2025: Tricolor Holdings Chapter 7 liquidation September 2025 following discovery of multi-pledged collateral fraud, exposure across JPMorgan, Fifth Third, Barclays. https://www.bloomberg.com/news/articles/2025-10-31/
  44. Addleshaw Goddard LLP, "Lessons from Tricolor and First Brands," 22 December 2025: First Brands Group Chapter 11 September 2025; off-balance-sheet supply-chain finance and factoring; UBS O'Connor ~$500M exposure, Jefferies Point Bonita Capital ~$715M exposure, total liabilities estimated $10B-$50B; Jamie Dimon "cockroach" quote; BlackRock/HPS Investment Partners third blowup reported 1 November 2025. https://www.addleshawgoddard.com/en/insights/insights-briefings/2025/financial-services/lessons-from-tricolor-and-first-brands/
  45. Reporting on Blue Owl Capital digital infrastructure fund structure, holdings in Beignet (Meta Hyperion) and Oracle Stargate financings, and public pension fund LP exposure including New York and Pennsylvania state pension plans, late 2025-early 2026. https://roddubitsky.substack.com/p/the-everything-bubble-ai-data-centers
  46. U.S. News, "Survey: 43% of Americans Don't Have Savings to Pay for a $1,000 Emergency," 4 February 2026 (PureSpectrum survey of 1,216 adults conducted 16-20 January 2026): 43% can't pay $1,000 emergency from savings; among those with savings, median balance $5,000 (half of prior year). https://www.usnews.com/banking/articles/2026-financial-wellness-survey
  47. Bankrate's 2026 Emergency Savings Report, 4 February 2026: 27% of Americans have no emergency savings at all; 43% would be "very worried" they couldn't cover next month's living expenses if they lost a primary source of household income. https://www.bankrate.com/banking/savings/emergency-savings-report/
  48. Bureau of Economic Analysis, Personal Savings Rate, FRED series PSAVERT: U.S. personal savings rate ~4-5% during 2024; April 2020 peak at 32% during pandemic relief payments. https://fred.stlouisfed.org/series/PSAVERT
  49. Branch / financial wellbeing research, 2024-2026, on hourly worker savings: 83% of U.S. hourly workers have less than $500 in savings; median Gen Z emergency fund ~$400. Cross-referenced with Federal Reserve SHED 2024 data on younger worker financial fragility.
  50. CBS News, "Wealth inequality in America just hit its widest gap in more than 3 decades," 21 January 2026, citing Federal Reserve Distributional Financial Accounts Q3 2025: top 1% holds ~$55 trillion ≈ wealth of bottom 90%; top 10% accounts for ~49% of consumer spending (Q2 2025), highest in series since 1989. FRED series WFRBST01134 confirms top 1% share of net worth at 31.7% Q3 2025. https://www.cbsnews.com/news/us-wealth-gap-widest-in-three-decades-federal-reserve/
  51. Yahoo Finance, "The Top 1% Now Control Nearly a Third of America's Wealth," 21 January 2026, citing Federal Reserve data: top 0.1% boosted wealth 40% in three years (twice the 20% gain of the bottom 90%); top 0.1% net worth ~$25 million per household; holds 14.4% of national wealth. https://finance.yahoo.com/news/top-1-now-control-nearly-220643415.html
  52. Institute for Policy Studies / Inequality.org, "Billionaire Wealth Concentration Is Even Worse than You Imagine," 29 September 2025, combining Federal Reserve DFA data and Forbes Real Time Billionaires List: 905 U.S. billionaires with combined wealth of $7.8 trillion as of 29 September 2025; bottom 50% (66 million households) holds $4.1 trillion combined. https://inequality.org/article/billionaire-wealth-concentration-is-even-worse-than-you-imagine/
  53. Reporting on institutional ownership of U.S. single-family rental homes reaching all-time highs through 2024, with Blackstone, Invitation Homes, American Homes 4 Rent, and other large funds documented across CoreLogic, John Burns Research & Consulting, and Urban Institute analysis 2023-2024.
  54. Industry data on private equity "dry powder" (committed but undeployed capital) reaching record ~$2.5 trillion globally as of 2024-2025, cited across Bain & Company Global Private Equity Reports, Preqin, and PitchBook industry analytics.
  55. First Republic Bank failure and FDIC-assisted acquisition by JPMorgan Chase, 1 May 2023; FDIC press release and JPMorgan disclosures. https://www.fdic.gov/news/press-releases/2023/pr23034.html
  56. Federal Reserve Bank of New York, Maiden Lane LLC: $30 billion facility supporting JPMorgan's March 2008 acquisition of Bear Stearns. https://www.newyorkfed.org/markets/maidenlane
  57. Emergency Economic Stabilization Act of 2008 / Troubled Asset Relief Program (TARP), $700 billion authorization; U.S. Treasury Office of Financial Stability final program reporting. https://home.treasury.gov/data/troubled-assets-relief-program
  58. Federal Reserve balance sheet expansion from ~$900 billion (pre-crisis 2007) to ~$4.5 trillion (post-crisis peak 2014); Federal Reserve H.4.1 statistical release historical data. https://www.federalreserve.gov/releases/h41/
  59. Federal Reserve approval of Goldman Sachs and Morgan Stanley conversion to bank holding company status, 21 September 2008. https://www.federalreserve.gov/newsevents/pressreleases/bcreg20080921a.htm
  60. New York Attorney General's report on Wall Street bonuses during the financial crisis ("No Rhyme or Reason: The 'Heads I Win, Tails You Lose' Bank Bonus Culture"), Andrew Cuomo, 30 July 2009, documenting tens of billions in bonuses at TARP-recipient banks for 2008. https://ag.ny.gov/sites/default/files/press-releases/archived/Bonus%20Report%20Final%207.30.09.pdf
  61. Home Affordable Modification Program (HAMP), originally $75 billion in announced funding; Special Inspector General for TARP (SIGTARP) quarterly reports documenting low fund-disbursement-to-homeowner rates, high redefault rates, and servicer fee structure. SIGTARP Quarterly Report, October 2014. https://www.sigtarp.gov/Quarterly%20Reports/October_29_2014_Report_to_Congress.pdf
  62. Congressional Budget Office and Center on Budget and Policy Priorities analyses of 2025-2026 Medicaid work-requirement and eligibility-tightening changes projected to reduce coverage; CRFB and HHS administrative-burden documentation across SNAP, SSDI, and student-loan administrative changes 2025-2026.
  63. RAND Corporation and Health Affairs research on U.S. hospital pricing markups, including CT scan pricing dispersion: typical commercial price multiple of Medicare allowed amount; consistent with widely documented 5-20x markup pattern across imaging services.
  64. Greensill Capital collapse, March 2021; Financial Conduct Authority investigation reporting and Treasury Select Committee inquiry documenting supply-chain finance fraud and off-balance-sheet exposure earlier prototype of First Brands pattern. https://www.fca.org.uk/news/press-releases/fca-statement-final-notice-greensill