The Second Derivative: Why No One Understands the AI Boom - Translation

Several weeks ago Jeff Moss, founder of DEF CON, reposted a link to this article into my Mastodon timeline: The Second Derivative: Why No One Understands the AI Boom.

Post by @joebeone@techpolicy.social

View on Mastodon

It’s an incredibly detailed explanation of the economics of the AI boom and how they relate to previous bubbles. Anyone trying to understand this boom and what the Minsky Moment looks like or when it might happen should give it a read.

The thesis is this:

1. The 2008 mortgage crisis was triggered not by a crash in housing valuations (although that certainly happened), but by a decrease in the rate of growth of housing valuations 1-2 years earlier. Home prices were still bananas and rising, but the rate of growth was declining. This broke the 2- and 3-year ARM refinance treadmill.
2. The AI boom is structurally much closer to a real estate boom than a tech boom. Construction and financing of data centers is at the base of all the other financial froth. The tech-bubble layers are the easiest to recognize, but underneath it all is a classic, debt-fueled real estate boom.
3. What it takes for the AI boom to correct isn’t a collapse in AI demand. It’s just a softening. If the rate of growth of successive funding valuations decreases, the treadmill breaks.

The problem is that this article is an extremely steep climb. The amount of finance and economic jargon makes it unapproachable to the non-finance reader.

Below is an AI-translated version that removes the jargon. It’s still a challenging read because this is a necessarily complicated subject, but it should be approachable for the non-finance nerd (that is the nerd who is not a finance nerd). At the end is a table of vocabulary from the article. I can’t promise you won’t be an armchair finance nerd by the end.

Here was my prompt to the agent:

I don’t have a finance background but I do read and reason about complicated topics. Please translate this article for a non-finance reader.

Keep the structure of the article. Sections should remain the same. Concepts discussed in the various sections should remain in those sections.

Include vocabulary from the article at the end.

A note on style: you’re not rewriting the article as if it wasn’t written for finance. You’re interpretting it for the reader. It’s okay to break the fourth wall with “The article is saying” or “the point here is” type style.

https://www.groundbrkr.com/p/the-second-derivative-why-no-one

Response:

Here is a plain-language translation of the article, keeping its original structure and section-by-section logic. I’ve used a “the article is saying” style where it helps, and I’ve added a vocabulary list at the end.

The Market Remembers 2008 Wrong. That Same Mistake is at the Heart of the AI Boom

The article is saying that most people tell the 2008 mortgage crisis story backwards. The usual story is: banks made bad loans, home prices crashed, people owed more than their houses were worth, they stopped paying, and the mortgage-backed securities blew up. Prices fell, so borrowers defaulted.

But the article says the order was actually different. The real engine of the subprime machine was not the price of homes. It was the speed at which prices were rising — and especially whether that speed was increasing or slowing down.

The classic risky mortgage, the 2/28 or 3/27 adjustable-rate loan, was not designed to be paid off normally. It was designed to be refinanced. The borrower got a very low “teaser” rate for two or three years. The hidden assumption, shared by both the bank and the borrower, was that home prices would keep rising fast enough to create new equity. That new equity would let the borrower refinance into another teaser loan before the rate jumped.

It was a treadmill, and the treadmill was powered by rising home prices. While it worked, it worked great: almost 80% of those risky adjustable mortgages made in 2003 had been refinanced away by the end of 2006.

Now here is the key timing. Home prices did not crash in 2006. They decelerated. The year-over-year price gain, which had been rising 15–20% per year, started to slow down. Prices were still going up. They were at record highs. But they were no longer rising faster and faster.

And right then, mortgage delinquencies started rising.

The article’s point is subtle but important. The popular story is not exactly wrong — eventually prices did fall and defaults did follow. But it treats the price crash as something that came from outside, like a meteor. The crash did not come from outside. It came from inside the structure. The whole machine needed prices to keep accelerating to stay ahead of the mortgage resets. Nothing can accelerate forever.

The “second derivative” — the acceleration of price growth — was always going to slow down. When it did, the actual price growth followed it down through zero. Negative equity spread from the weakest borrowers inward. And the defaults that everyone blamed on “falling prices” had actually started a year earlier, when prices were still rising but had stopped rising faster.

II. A Short Theory of Derivatives

The article pauses here to explain a little math, because the whole argument depends on it.

Imagine you are tracking some number. Call it S. There are three ways to describe it:

• The level is just S itself: how big the number is right now.
• The first derivative is like speed: how fast the number is growing.
• The second derivative is like acceleration: whether the growth is speeding up or slowing down.

The article’s point is that markets watch the first two closely and mostly ignore the third. Analysts build models around the level. Traders bet on the speed. Almost nobody builds a strategy around the acceleration.

But the second derivative is where regime change hides. When a financing arrangement is built on an assumption of continued growth — a mortgage that assumes refinancing, a spending plan that assumes expansion, a contract that assumes revenue will keep doubling — the assumption is satisfied not by the number being big, but by growth being sustained.

Growth can still be positive and still make headlines while quietly violating the underlying assumption. “Revenue grew 40%!” sounds great. But if the plan required 70% growth, the gap is opening silently.

There is a window — the article calls it “borrowed time” — between when the acceleration starts slowing and when the actual growth turns negative. During that window, everything still looks fine. Revenue is at record highs. Growth is still positive. Press releases are triumphant. But the machine is already broken; it just has not been told yet.

This is especially dangerous when the financial instruments on top are “negatively convex.” The article explains this with a contrast. A stock valuation can deflate slowly and reflate later. The dot-com bust took two years to bottom, and many companies survived by simply being valued lower.

A leveraged credit structure is different. It earns steady payments on the way up but can suffer huge, sudden losses on the way down. Its rules are step functions: miss one threshold and the whole thing breaks. The article says this distinction — stocks can drift, credit can snap — is the difference between the 2000 dot-com crash and the 2008 mortgage crisis. And it is also the difference between what the market thinks AI is and what AI actually is.

III. The AI Boom is a Credit-Driven Real Estate Cycle

Open any bullish or bearish AI research note and you will see it arguing about levels — revenue in billions, power in gigawatts, total market size — and about speed — is growth 200% or 150%, is enterprise adoption picking up. The bears say the numbers are unsustainable. The bulls say the growth justifies them. Both sides are staring at the level and the speed. Neither is watching the acceleration.

But the entire financing architecture of the last two years is a bet on acceleration — on demand continuing to grow faster — disguised as a bet on the level.

This matters because the dominant analogy in everyone’s head is wrong. People ask, “Is AI a bubble like dot-com?” and that framing produces dot-com answers: maybe the leaders survive, the laggards wash out, stock valuations compress, maybe a 50% drop and recovery. The article says this is the wrong category.

The dot-com crash was a stock-valuation event: too much optimism baked into shares of companies with little debt. The pain was real, but it was stock pain, and stock is patient capital that can be marked down and held. The AI build-out is structurally different. It is increasingly financed not by selling overpriced stock but by locking in future cash flows and borrowing against hardware — long-term must-pay contracts, loans backed by chips, debt vehicles, asset-backed notes sold to insurers.

The article’s punchline: “That is not the architecture of 2000. That is the architecture of 2008.”

There is a deeper misclassification. The market is pricing AI as a technology cycle, but its actual anatomy is a credit-driven real estate cycle. That is why the 2008 mechanics apply.

Technology cycles are driven by innovation and adoption. Their risks are obsolescence and competition. They can lose value slowly as the future gets repriced.

Real estate cycles are mechanical: debt, hard assets, occupancy. They involve building too much with borrowed money, leases disguised as must-pay contracts, long construction lags that guarantee supply arrives after demand has turned. Walk through the AI build-out and every feature looks like property development in disguise: a data center on entitled land, financed with debt against the building, leased to tenants on must-pay terms. The article puts it this way: “This is not a software business that happens to own servers. It is a real estate business that happens to compute.”

Real estate cycles break the same way every time. Not when demand collapses — it rarely does — but when the rate of demand growth slows against the fixed supply the boom just finished building. The second derivative again.

And the credit machine does not lose value gently. It either refinances or it seizes. The instruments that seize — must-pay leases, chip-backed loans, asset-backed notes — are each negatively convex. Within this machine, the financing mechanics split apart. Smaller AI cloud companies borrow non-recourse debt against a specific tenant’s must-pay contract and collapse when the tenant cannot pay. The big tech cloud companies fund with corporate bonds and operating cash flow; for them, a tenant default means a write-down and squeezed margins, not collapse. Oracle sits in between: corporate-funded but dangerously concentrated.

The distinction matters because collapse and write-down are two different wounds, caused by the same slowdown, and both are hiding inside the same $2.1 trillion backlog. To judge the quality of that backlog, you have to look past the headline number to the core exposure. Across the big four cloud platforms, the backlog that Wall Street treats as future revenue is, in reality, a concentrated credit exposure to a handful of cash-burning AI labs and specialized tenants.

The article is saying: the big cloud company has, in economic substance, extended a concentrated infrastructure loan to tenants with no independent operating income. If those tenants default, the backlog evaporates into accounting write-downs, leaving the corporate balance sheet to absorb the fixed costs of customized, rapidly aging capital assets.

IV. The Loan Book Nobody Calls a Loan Book

The article says: look past the “compute is scarce” story and see these agreements for what they really are.

An AI lab signs a contract promising to pay a counterparty tens of billions of dollars over several years for computing power it has not yet used. The counterparty — Oracle, CoreWeave, a big cloud company — books that promise as backlog and borrows against it, raising debt to pour concrete and install chips. Strip the arrangement to its skeleton and it is a loan: the counterparty advances capital in the form of a building full of chips, against the borrower’s promise to pay it back, with interest and principal repayment baked into the must-pay rate. The data center is the collateral. The lab’s contracted payments are the debt service. And the structure works only as long as the borrower can keep funding those payments — which, for a company with no profits, means only as long as it can keep raising money.

This is why the analogy is 2008, not 2000. AI capital spending is not a capital budget. It is a loan book. Leases on paper; debt in substance. The capex is the funded principal; the contracted backlog is the loan receivable. When a big cloud company or AI cloud provider reports record capital spending, the financial press reads it as confidence, as proof of demand. Read it instead as loan volume. Each gigawatt of committed build is a loan extended to whichever tenant signed the must-pay contract beneath it, and the credit quality of that loan is exactly the credit quality of the tenant. The market is celebrating loan growth and calling it revenue growth.

V. The Borrower With No Income

Every subprime cycle has a borrower who can only refinance, never repay. In this one, the article says, that borrower is OpenAI.

OpenAI has committed to pay for computing power on a scale without precedent in corporate history: multi-year, must-pay capacity contracts whose total obligations run to the hundreds of billions of dollars. Against that sits an operating business that does not yet make a profit — revenue is real, large, and growing fast, but it does not cover the company’s own cash losses and is nowhere near covering the contracted payments.

Those payments are therefore not paid out of earnings. They are paid out of financing. And financing, for a borrower in this position, is available on one condition: that each new funding round values the company higher than the last.

For OpenAI, the up-round is not just a sign of progress. It is the funding event itself — the mechanism by which last period’s commitments are paid and next period’s are made signable. The markup is the cash flow.

A company funded by its own rising valuation is solvent not based on how high the valuation is, but based on how fast it is still rising — because cash burn is an accelerating schedule, indifferent to the size of the last round. Each new phase of computing expansion demands an exponentially larger cash injection. When the step-up compresses from 1.91× down to 1.23×, the math breaks: the valuation can hit a record at the exact moment the company’s ability to fund its structural deficit is shrinking. This is not a paradox. It is the arithmetic of a borrower whose liquidity depends on the speed of its own price growth.

And that value is set by the very capital providers who need it to keep rising. OpenAI’s most recent round — reported at roughly $122 billion of fresh capital — set its valuation near $852 billion, with the same names underneath it: Microsoft, SoftBank, Nvidia, Amazon. These are the counterparties whose computing power the proceeds will buy. The valuation goes up because money came in; more money comes in because the valuation went up. The appraiser, the lender, and the buyer are the same three people, passing the same dollar in a circle and marking it higher each time. It works gloriously — for a while — for exactly the reason the teaser mortgage worked: as long as the mark keeps rising, the lab can refinance.

Measured as a level, OpenAI’s valuation is the most remarkable appreciation in private-market history — roughly $86 billion in early 2024, then about $157 billion, $300 billion, $500 billion, and approximately $852 billion by spring 2026. Measured as a rate of change, the same series inverts: the round-over-round step-up ran 1.83×, 1.91×, 1.67×, 1.70×, and falls to roughly 1.23× implied by the reported public-offering target. Private valuations are lumpy — negotiated, episodic, set by a handful of insiders — so no single step is decisive. But the trend is clear: it bends down, and it bends hardest at the one mark set by the deepest, most unforgiving pool of capital — the public market. The implied IPO step-up is both the lowest in the sequence and the hardest to negotiate, and it is the one the structure must actually clear. This arithmetic is also the most likely explanation for OpenAI’s recent IPO delay.

The article emphasizes: this slowdown is not random. It reflects two structural headwinds directly attacking the revenue growth the valuations require: token efficiency and Chinese open-weight models. The industry’s central optimization project — routing simple queries to cheap models and trimming “thinking” tokens — has eroded the token-per-task tailwind that padded revenue. Meanwhile, Chinese open-weight models have repriced the commodity middle of AI usage to near-zero, capturing over 60% of OpenRouter usage at a fraction of the price. Revenue still grows — adoption is real — but the rate of growth is precisely what is under attack, and the rate is what the next valuation needs to clear.

VI. The Lender’s Backlog

Move up one level, from the borrower to the lenders. The credit they have extended takes a specific form: remaining performance obligations — RPO — the contracted revenue a company has under signed agreement. The backlog. Wall Street loves backlog; it reads as visibility, as demand pulled forward and locked in.

On the big cloud company balance sheets, that backlog has swollen into the hundreds of billions apiece, and every quarter the growth in backlog is presented as proof that demand is real and the buildout justified. The larger the backlog, the more secure the story: a company does not build a gigawatt on a hope, it builds it against a contract.

But backlog is not cash in hand. It is a forward contractual commitment — a promise of future payment in exchange for future computing power. And a multi-year commitment is worth exactly the creditworthiness of the entity on the other end. When that entity is solid and cash-generative, the backlog is what it claims to be: high-quality future revenue, merely delayed. When that entity is a pre-profit company that loses tens of billions a year and can pay only by continuously refinancing its own valuation, the backlog is something else entirely. It is a risky commitment, used to justify massive, undepreciated capital spending, reported to shareholders as structural strength.

Now judge the credit quality of that book. Of roughly $2.1 trillion in aggregate contracted backlog across the four big platforms, about half — on the order of $1.05 trillion — is owed by OpenAI and Anthropic. Microsoft’s book is about 49% these two names; Oracle’s is 54%, with roughly $300 billion owed by OpenAI alone; Google’s is 43%; Amazon’s is 51%.

The article repeats the point: the big cloud company has, in economic substance, extended a concentrated, unsecured loan to cash-burning tenants. The backlog that Wall Street values as future revenue is, in reality, a credit exposure to borrowers with no operating income.

Here the bulls raise their strongest objection. Yes, the frontier labs burn cash now — but so did Amazon, so did every great compounding business in its infrastructure-building phase. Burn is investment; the labs will grow into profitability; the borrower of today is the cash machine of tomorrow. The article says this is half right. There is no single frontier-lab borrower. There are at least two, and they are not the same credit.

Anthropic is a risky, but well-insulated, borrower. Its revenue is roughly 80% enterprise — sticky, recurring, contracted seats — and its unit economics are firmly above water, generating $1.70 of revenue for every dollar of computing power. Cash burn converges to a manageable ~9% of revenue by 2027 as profit margins normalize.

More importantly, Anthropic’s liabilities are protected by a classic 2008-style “co-signer” maneuver. In the ~$35 billion Apollo/Blackstone chip facility, Anthropic’s paper borrows the credit rating of its stronger backers: Google guarantees lease shortfalls, and Broadcom guarantees the residual value of the chips on the ~$31 billion senior portion. The co-signers are standing behind the bills and protecting Anthropic’s counterparties.

OpenAI is the “naked” borrower, making it the weakest and most volatile credit in the ecosystem. Its revenue mix is fragile — roughly 60% consumer. With cash burn hovering at a crushing ~57% of revenue through 2027 and cumulative cash destruction marching toward $115 billion by 2029, OpenAI has no path to positive cash flow this decade.

Worse, OpenAI has no real co-signer. While the market long assumed an implied Microsoft backstop, Microsoft stripped away every structural support in April 2026 — ending the revenue share, dropping exclusivity, and surrendering its right of first refusal to supply computing power. Microsoft kept its 27% equity upside but walked away from OpenAI’s bills. SoftBank, the other big OpenAI backer, is itself now trying to raise a $10 billion margin loan against its OpenAI stake — offering a personal guarantee after lenders balked at the collateral. Even the co-signer has no co-signer.

Microsoft’s own behavior is the signal. The most informed counterparty in the whole system — the one that saw OpenAI’s books from the inside for years — has recognized the credit risk. Rather than building its own data centers on fifteen-year leases that outlast the chips, it foresaw the commoditization of frontier models and committed over $60 billion to AI cloud providers through shorter, five-year capacity agreements: renting at the peak to avoid owning through the trough. That is not a bet on OpenAI’s durability. It is a lender shortening the loan term on a borrower it has decided not to underwrite — the same risky credit this section describes, priced by the party that knows it best.

The critical divergence here is counterparty risk. When an investor or lessor underwrites Anthropic, they are ultimately looking through the structure to underwrite the pristine balance sheets of Google and Broadcom. The credit risk is synthetically lifted to investment-grade. When underwriting OpenAI, there is no look-through. Counterparties are exposed to a naked, standalone startup sitting on an underwater unit economic model. Without a parental balance sheet, OpenAI is entirely dependent on a continuous refinancing treadmill — paying each old obligation with the proceeds of the next, larger equity raise. That is the counterparty risk hiding inside roughly half of the $2.1 trillion backlog the market has priced as bankable.

VII. The Reflexive Flip

The must-pay contracts are the structural foundation — the collateral that makes the borrowing possible. But they are not a passive constraint. The self-reinforcing arms race is what pushes the big cloud companies to sign those contracts in the first place, and to sign them at ever-larger scales. The race does not bypass the loan book; it writes the loan book. To justify the next gigawatt of spending, a big cloud company needs the next gigawatt of backlog — so it pushes its tenants to commit further forward. The $2.1 trillion backlog is not a pre-existing limit on the arms race; it is the arms race’s own paper trail. The contracts are the collateral; the race is the demand for more collateral.

Once that collateral is signed, it must be turned into infrastructure before the cash arrives. That conversion — turning a signed contract into a live data center — is what drives the capital spending machine. And that machine is now consuming cash faster than the backlog can validate it.

Aggregate capital spending as a share of operating cash flow ran near 30% in 2022, roughly 42% in 2023, about 50% in 2024, and approximately 60% in 2025; on consensus spending it reaches 100% in 2026. Above that line, by definition, every additional dollar of capacity is funded not from internal cash but from the balance sheet — debt or equity. The “strong balance sheet, self-funded” story is true only below 100%, and the consensus path crosses 100% this year. The article’s line: “The fortress is not being defended; it is being spent.”

Past 100% of cash flow, accelerating capital spending means borrowing more, faster, every quarter, against a credit rating that only has so many notches left. Big tech debt issuance has to climb steeply over the coming year — the bond market becomes the marginal funder of the entire build.

Why are big tech companies betting over 100% of operating cash flow on an uncertain return?

Because, until now, they have been paid to. Capital spending has gone vertical: roughly $150 billion in 2023, $226 billion in 2024, $410 billion in 2025, an estimated $725 billion in 2026, and approaching $1.1 trillion in 2027. As a level, it is the largest private capital-formation event in history. As a speed, the growth rates read +51%, +81%, +77%, +52%. And as an acceleration, the growth peaked at roughly +30 percentage points into 2025 and has turned negative: about −4 points, then about −25. The level is at records. The velocity is still high. The acceleration has already rolled over.

There is a recursion here that the headline numbers obscure. Big tech capital spending in this cycle is not primarily a response to AI demand. To a substantial degree, it is the demand. The labs’ revenue is, in large part, big tech spending recycled — cloud credits, compute commitments, equity-funded consumption. Nvidia’s revenue is big tech capital spending. The smaller AI cloud providers’ revenue is big tech capital spending, levered.

Strip out the spending and the demand it manufactures, and the organic, spending-independent demand is a fraction of the headline figure. Which means the single most important growth rate in the system is the acceleration of big tech capital spending — and it has already gone negative while every level chart still points to the sky.

The capital spending arms race is a self-reinforcing equilibrium, but a conditional one: it holds only while the market rewards the next dollar of spending as a bet on future growth. In that regime — the boom regime — the dominant strategy for every big tech company is to spend, because the alternative is to be the one player who blinked and ceded the future. Mutual escalation is stable precisely because the market applauds it. Each CFO spends because every other CFO is spending and the stock valuation rewards the spender.

Morgan Stanley caught the psychology exactly when it described 2027 capital spending estimates leaping thirty percent in a single quarter, toward $1.1 trillion, as the dynamics of an auction. An auction is the right frame, because in an auction the price is set by the most optimistic bidder and the act of bidding is itself the signal — the applause, the proof of seriousness. Keynes’s beauty contest, with chips: you are not spending on what you think the computing power is worth; you are spending on what you think the market will reward you for being seen to spend.

That equilibrium is not anchored to anything physical. It is anchored to a belief — the market’s reading of what the next dollar of capital spending means — and beliefs reprice. The flip comes the first time a big tech company announces a capital spending cut and its stock valuation rises on the news rather than falling. The instant discipline is rewarded instead of punished, every payoff on the board rewrites. Spending, formerly the dominant strategy, becomes the move that gets you punished alone; holding, formerly surrender, becomes the move that gets you re-rated. The equilibrium flips from “everyone spends” to “everyone cuts” — and because it is a coordination game, the flip is not gradual. The first mover rewarded for cutting gives every other CFO both the cover and the incentive to follow, and discipline cascades as fast as the spending it replaces. The day the market cheers a cut is the day the arms race ends.

Goldman’s head of Delta One trading put it plainly:

“The first hyperscaler to signal that it can slow the pace of spending will likely see its share price rewarded (and will crush semiconductor stocks). If that happens, others will take notice. That is the reflexivity that ultimately stalls the capex cycle — not a lack of demand, but investors deciding that incremental returns on the next dollar of spend are no longer attractive.”

The cruelty of the flip is what it does to the contracts. In the boom regime, a signed must-pay commitment is an asset to everyone who touches it: future demand for the big cloud company, bankable backlog for the AI cloud provider, collateral for the lender. In the repriced regime, the identical contract is a liability for all of them simultaneously.

The lab cannot fund the payments it locked in; the big cloud company holds a receivable from a visibly distressed counterparty; the AI cloud provider is left servicing debt against data centers it financed on a contract now worth less than the debt. This is negative convexity wired directly into the demand side: the same instrument is an asset on the way up and a liability on the way down, and the transition between the two states is a repricing of belief, not a change in the underlying hardware. Nothing physical has to break. The market only has to change its mind.

It lands hardest on the frontier labs, who can carry these contracts only by raising more capital — and the flip closes that window. What follows is not a clean default but a negotiation — volumes cut, schedules stretched, contracts restructured. The contracts do not vanish; they reprice — beginning with the borrower who needs the next round most.

VIII. Who Blinks First

Every reflexive cascade needs a first mover. So which big tech company cuts first? Who blinks?

The instinct is to say the weakest balance sheet, and the article says the instinct is wrong. The first to cut will be the one with the best information, the credibility to reframe the cut as strength, and the balance-sheet room to be rewarded rather than punished for it.

Zuckerberg holds dual-class control. He has run this exact playbook before and was rewarded with a tripling of the stock; and of all the big tech companies Meta has the weakest direct monetization of its AI capital spending — no public cloud to sell the capacity into — which makes its spend the hardest to defend and the easiest to cut. The only reason it has not cut yet is the self-reinforcing equilibrium — Zuckerberg is waiting for the market to tell him it is safe to stop spending.

The others array predictably. Google will not blink — it builds its own chips at a structural cost advantage and reports a cloud backlog north of $460 billion, so it benefits if rivals retrench. Oracle cannot blink: at roughly 86% of sales going to capital spending, with a balance sheet stretched around Stargate, its stress will surface as a credit event. Amazon may have its hand forced from the other direction — free cash flow already turning negative under the build. Negative free cash flow is the kind of thing capital markets eventually vote on, whether management calls the election or not.

The numbers tell the same story. Morgan Stanley pegs big tech investment-grade leverage at roughly 1.8 turns of gross debt — double what it was a year ago and now higher than the entire energy sector. That figure does not count the hundred-billion-plus parked off the balance sheet in the vehicles. What stands in its place is a leveraged, hard-asset, refinance-dependent balance sheet — and the marginal gigawatt, the thing cut first, is the most discretionary line on it.

IX. The Blast Radius

Let’s say OpenAI is subprime, the regime shifts, belief reprices, the capital window slams shut, and a big tech company cuts that marginal gigawatt to protect its own leverage. Who is exposed?

OpenAI is the single largest customer — by direct contract or one counterparty removed — of very nearly every name that sells into the AI build. Oracle’s contracted backlog is more than half OpenAI; CoreWeave’s book — once its Microsoft-routed capacity is traced through to the underlying tenant — runs to roughly two-thirds OpenAI; SoftBank’s commitments, through Stargate, are almost entirely OpenAI.

This is precisely the structure that made 2008’s senior tranches lethal: thousands of individual mortgages, geographically dispersed, statistically independent — until the one macro variable they all depended on, national home prices, turned, and the correlation the models had assumed away revealed itself to be one. Here the single variable is not home prices. It is whether OpenAI can clear its next mark. That is why chip stocks fell when OpenAI signaled it may delay its IPO from 2026 to 2027.

A correlation of one is invisible until it is tested. Then it is a transmission line. When the borrower at the center cannot clear its next mark, the loss does not stay put — it runs the length of the chain, into every counterparty that booked its commitment as demand. The naked borrower is not merely the weakest credit in the complex. It is the credit the complex is wired to.

That correlated exposure is now being packaged and sold. In May 2026, CoreWeave closed its DDTL 5.0 facility — $3.1 billion, issued through a bankruptcy-remote financing subsidiary. CoreWeave disclosed that the underlying capacity serves two large, non-investment-grade customers: OpenAI and Cohere. But the distinction that matters is structural: DDTL 5.0 was the first publicly syndicated GPU-backed facility, built to trade in the secondary market. The paper has left the originator’s balance sheet and entered the broad credit system — the distribution step, the moment originate-to-distribute stops being a metaphor.

The DDTL isn’t serviced by OpenAI’s earnings; it’s serviced by OpenAI’s ability to keep raising, which is underwritten by the AI capital spending narrative continuing to compound.

X. The Refinance of Last Resort

Trace the refinancing chain to its end and you arrive at the public market. Private capital is deep but finite: SoftBank, the sovereign funds, the big tech companies, the megafunds — each can absorb a round or two, but the labs’ cash burn is measured in tens of billions a year and compounding, and at some point the only pool of capital large enough to keep refinancing it is the one the index funds and the retail bid sit in.

The IPO is not an exit in this structure. It is the refinancing of last resort — the final, deepest teaser into which the whole edifice expects to roll once the private rounds can no longer carry the burn. Which is why news of OpenAI’s delayed IPO matters far more than the market initially understood.

The terminal refinance carries a trap the private rounds did not. To reach the public pool the borrower must file an S-1 — and the S-1 discloses exactly the fragility that made the refinance necessary: audited losses, customer concentration, the full $600 billion-plus of must-pay obligations laid out for any reader. The document that unlocks the capital is the same document that prices the risk.

OpenAI needs the market’s money and cannot fully afford the market’s scrutiny — the bind of a company whose story is better than its statements.

Now do the arithmetic the delay is hiding. The step-up from roughly $852 billion to the reported >$1 trillion target is the next hurdle — barely 1.23×, the lowest step-up in the entire sequence, and far below the 1.7×–1.9× multiples that funded the prior burns. It must do two incompatible things at once: clear at a level the public market will actually pay, and raise enough to retire a cumulative burn approaching $115 billion. The implied step-up cannot do both: the price that clears the market does not retire the burn, and the price that retires the burn does not clear the market. The refinance of last resort is failing quietly — pricing below the mark the structure requires, and waiting.

XI. How It Breaks

The trigger is narrow and specific: the next valuation fails to clear at the required step-up — not a collapse, merely a slowdown below the threshold the structure needs. This is the 2006 dynamic replayed: the velocity rolled over while the level was still climbing.

From there the sequence runs in order:

1. The terminal refinance prices below the required mark — the step from about $852 billion to more than $1 trillion does not clear, or clears at a level that cannot retire the burn; the delay is the signal.
2. The borrower pulls back on compute commitments to conserve cash — and a pull-back on a must-pay obligation is a covenant breach against the provider whose debt is collateralized by that commitment.
3. The breach lands first and hardest on the smaller AI cloud providers — CoreWeave, Lambda, Crusoe — whose entire business is the spread between borrowed money and resold computing power. A smaller AI cloud provider is not a business so much as a spread trade with no balance sheet to warehouse the risk: when the spread inverts, it is insolvent by definition, not by choice. Oracle, corporate-funded but dangerously concentrated, takes the next blow — its write-down deeper than the big tech companies’, but it does not seize; it bleeds. A big tech company can fund a missed payment out of Search, or Windows, or Retail; the smaller AI cloud provider has no second cash flow.
4. Credit freezes across the complex: backlog reprices from future demand to counterparty risk, chip-backed notes cannot roll, the originate-to-distribute machine seizes.
5. Stock valuations crater — negative convexity in reverse, capital spending repriced from option to cost, valuations compressing across every name in the chain.
6. The strong survive: the best-capitalized actors with the least exposure buy stranded data centers for pennies and backstop the leases that must endure.

A necessary concession: the author does not know when. The trigger could be quarters away or further; the borrowed-time window between the acceleration rolling over and the actual growth crossing zero can stretch further than any short-seller’s patience. There are three stretches that can extend it: a larger-than-expected private round, a sovereign or strategic backstop that postpones the terminal refinance (like an Intel-style federal equity stake), and the big tech companies’ continued ability to lever up — borrowing against the very backlog this article has described.

The last of these is the most powerful near-term stabilizer, because the big tech companies have real balance sheets, real cash flows, and real access to debt markets. But it is not infinite. Investment-grade leverage across the group has already doubled in a year and the rating agencies have only so many notches left. The sequence above is not a calendar; it is a mechanism, conditioned on a single variable — whether growth decelerates below the rate the refinance requires. But with OpenAI’s IPO already delayed, the clock is ticking.

XII. The Strongest Case Against This

Grant the bulls their strongest case: demand is real, backlogs are exploding, AI inference is in its infancy, and the risk of underbuilding a generational platform is acute. Supply is locked years out, and even skeptics see paths to $1.4 trillion in annual capital spending. The author takes this case seriously — but says it does not save the structure.

Every bull claim is about the level or the speed: backlogs, inference ramping, supply growth. Not one speaks to the acceleration. The author does not need demand to fail. He needs the rate of capital spending growth to flatten — and a structure this levered and dependent on perpetual acceleration breaks on the flattening alone. Grant every level argument. Housing demand was real in 2006 — and the financing detonated on deceleration, not the level.

There are three bull cases to address.

First, the fortress balance sheet. Big tech companies generate enormous cash flow; a tenant write-down is absorbable. This misses the wound, the article says. The write-down is accounting; margin collapse is structural. AI capacity carries a massive fixed-cost base — depreciation, power, interest — that does not flex when a tenant defaults. Utilization drops, but operating expenses do not. Revenue falls, yet costs remain anchored to the peak build. The same operating leverage that supercharged profits now destroys margins on the way down.

Second, the cross-subsidization defense. If AI margins crater, Search and Windows cash flows carry the division. The rebuttal is the conglomerate discount. Investors buy big tech companies for growth, not to subsidize perpetual losses. If AI consumes tens of billions without profitability, consolidated returns on invested capital decline. A high-return growth compounder that becomes a low-return capital-intensive operator loses its growth premium and trades down to a utility multiple. Worse, legacy cash cows are not infinite engines. Search faces structural erosion; Retail operates on thin margins; Windows is mature. Using shrinking profits from declining units to fill vacancies is not patient capital — it is value destruction. The conglomerate trades as a utility with a venture capital problem, commanding a lower multiple.

Third, the physical rebuttal: if OpenAI defaults, the provider re-leases the capacity. This is the “housing never loses value” argument of 2006. An OpenAI default will not occur in isolation — it will coincide with a broader deceleration, meaning big tech companies bring gigawatts online into a softening environment. You do not re-lease into a glut; you compete on price, and the clearing price falls below the debt-service coverage ratio. The replacement tenant, facing the same decelerating demand, will demand a 30–50% discount and a shorter commitment, turning a long-duration, high-yield asset into a distressed instrument. Re-leasing merely transforms a clean default into a prolonged vacancy crisis — the same mechanism that turned 2007’s subprime “re-performance” hopes into a five-year grind.

The bulls and the author do not disagree about AI. They disagree about which derivative the structure is written on. The bulls are watching the level. The author is watching its acceleration. That is not a difference about technology. It is a difference about arithmetic — and arithmetic, eventually, does not take opinions.

XIII. The Number Nobody Watches

Three errors, stacked, recreate 2008. The market is pricing AI as a technology cycle when its financing is the machinery of a credit-and-real-estate cycle. It is watching the level and the speed while the structure breaks on the acceleration. And it is treating a concentrated, single-borrower loan book as though it were diversified future demand. Each error alone might be survivable. Together they reconstruct, feature for feature, the conditions of the last great credit event — the same negatively convex structures, the same originate-to-distribute plumbing, the same correlation-of-one hiding inside the appearance of diversification, the same blindness to the one derivative that matters.

The law from the opening sections holds, unchanged: any structure whose serviceability depends on refinancing into growth does not need a decline. It needs only a deceleration. That deceleration is already happening.

The market remembers 2008 backwards. The defaults didn’t come when prices fell. They came when prices stopped rising faster — and this build-out is engineered, with exquisite precision, to break on the one number nobody watches.

The Second Derivative.

Vocabulary from the Article