Buy Now, Pay Later Is Booming. So Is the Debt Nobody Can See.

---

Begin with the phrase that should make any forensic reader's ears prick up: "phantom debt." It is the term that
has emerged to describe the central, structural peculiarity of the buy-now-pay-later industry — that the loans it
originates, by the tens of billions of dollars, are largely not reported to the credit bureaus, which means they
do not appear in the data that lenders, regulators, and economists use to gauge the health of the American
consumer. A borrower can take out a BNPL loan at Klarna, another at Afterpay, a third at Affirm, a fourth at Zip,
all in the same week, and no single one of those lenders — nor any bank evaluating that borrower for a credit
card or a car loan — can see the full picture, because the picture was never written down anywhere they can read
it. This is not hypothetical: one study found that nearly 63% of BNPL borrowers had more than one such loan
outstanding at the same time — the precise loan-stacking the invisibility enables and the official data cannot
detect.

This is not a quirk. It is the foundation of the business, and it cuts two ways at once. It is what allows BNPL
to extend credit so frictionlessly — no hard credit pull, no waiting, just "four interest-free payments" at the
checkout button. And it is what makes the resulting pile of debt impossible to measure, which means that every
reassuring statistic about BNPL's low delinquency rates and the American consumer's resilience is being computed
on data that has a multi-tens-of-billions-of-dollars hole in it, shaped exactly like the debt most likely to go
bad.

This essay is about that hole — about who is borrowing in it, why the comforting numbers are unreliable, why the
industry's best operator is not the same as the industry, and why a category sold as a harmless modern
convenience has quietly become a major, unmeasured source of leverage for the most stretched consumers in the
country.

The numbers that look fine are computed on incomplete data

Start with the reassurance, because dismantling it is the key move. The bulls on BNPL — and on consumer credit
generally — point to a genuinely comforting figure: reported BNPL delinquency rates run under 2%, lower than the
roughly 3.5% delinquency rate on overall consumer debt. If the people using BNPL were defaulting in droves, the
argument goes, it would show up in the delinquency data, and it doesn't, so the panic is overblown.

The problem is that the reported delinquency rate is calculated on the loans the lenders choose to measure and
disclose, in an industry whose defining feature is that the debt is not comprehensively tracked. When most BNPL
activity is not reported to credit bureaus, the "delinquency rate" is not a reading of the whole population's
behavior; it is a reading of each lender's own book, in isolation, blind to what the same borrower is doing at
four other BNPL apps simultaneously. A consumer who is current at Affirm because they are using a new Klarna loan
to stay afloat does not register as distressed anywhere — until the whole stack collapses at once. The system
cannot see the layering, and layering is precisely the behavior that turns manageable individual loans into an
unmanageable aggregate.

Set the clean sub-2% figure against a very different number from the same period: a LendingTree survey found that
41% of BNPL users reported experiencing a loan delinquency in the past year — up from 34% the year before. Those
two figures cannot both be the whole truth, and the gap between them is the measure of the blindness. The
official rate says everything is fine; the survey of actual users says four in ten missed a payment. The
reconciliation is that the official rate sees only a slice, and the slice it sees is the part the industry has
chosen to illuminate. When you compute a reassuring average over a dataset built to exclude the worst of the
behavior, you get a reassuring average. You do not get the truth.

Who is actually borrowing

Now to the question the convenient framing of BNPL — "a smart, interest-free way for responsible people to
manage cash flow" — most wants to avoid: who is actually using these loans, and for what?

The data is unambiguous and uncomfortable. Roughly two-thirds of BNPL originations go to borrowers with subprime
or deep-subprime credit scores. More than 60% of U.S. BNPL borrowers fall into subprime or near-subprime
categories. And critically, these borrowers are not using BNPL instead of other credit — they are using it on
top of it: BNPL users carry average credit-card utilization of around 60–66%, nearly double the ~34% of
non-BNPL users. In plain terms, the typical BNPL borrower is someone who has already drawn down most of their
available credit-card limit and is now reaching for a different, newer, less-visible form of credit to keep
spending. BNPL is not, for this large cohort, a clever convenience. It is the marginal dollar of credit for a
consumer who has run low on the conventional kind.

And the things being financed have drifted downmarket in a way that tells its own story. BNPL began as a way to
spread the cost of meaningful purchases — a mattress, a Peloton, a laptop. It has migrated to financing
groceries, everyday retail, and, in the partnership that became an instant symbol of the whole phenomenon,
DoorDash food-delivery orders. When consumers are using installment credit to pay for a restaurant meal in four
biweekly payments — "collateralized DoorDash obligations," as one wry observer dubbed it — you are no longer
looking at a tool for managing large purchases. You are looking at credit being used to bridge the gap between
income and the cost of daily life, which is the signature of consumer stress, not consumer convenience. A person
who can comfortably afford a burrito does not finance it in four installments; a person who does is telling you,
in the most quotidian way imaginable, that the money runs out before the month does — and they are doing it on a
platform that will never tell anyone else.

The early tremor: Klarna

If you want to see where the unmeasured risk surfaces first, watch the lenders most exposed to the lower-quality,
subprime-heavy end of the market — and watch two things: their credit losses, and the market's verdict on their
stock. Klarna, the Swedish BNPL giant, provides the clearest reading on both. On the operating side, Klarna's
consumer credit losses have been climbing — rising 17% in a recent quarter — even as the company scraped its way
back from a roughly $99 million net loss a year earlier to a razor-thin profit of about $1 million on revenue up
44%. Profitability returned, in other words, but it returned as a sliver, achieved while the credit losses
underneath it were still rising. And the market's judgment has been brutal: since its IPO, Klarna's stock has
fallen roughly 62%.

That combination — credit losses climbing while the bottom line balances on a knife-edge, and the stock down
nearly two-thirds from its debut — is the early signature of a lending business whose growth is outrunning the
quality of the credit it extends, and whose public investors have stopped believing the headline story. A
razor-thin profit built on top of rising loss rates is not stability; it is a margin of safety measured in
basis points, in a business whose borrower base is deteriorating. Klarna is not a fringe player; it is one of the
largest BNPL companies in the world, recently public, a bellwether for the entire model, and its shareholders
have already repudiated it. Crucially, a credit loss reported today reflects loans made months ago, in better
conditions; the losses on the loans being written now, into a more stretched consumer, will not surface until
later still. Reported credit losses are a rear-view mirror, and the road behind already looks worse than the
dashboard admits. When the bellwether lender to the subprime consumer is balancing on a one-million-dollar profit
while its loss rates rise and its stock collapses, the prudent assumption is not that the problem has peaked but
that it has barely begun to be measured.

Affirm is not the industry

Here is where intellectual honesty requires a sharp distinction, because the BNPL industry is not monolithic and
the best operator in it is genuinely good. Affirm — the largest U.S.-based player — is, on the evidence, the
high-quality end of this market, and lumping it together with the subprime-pay-in-four crowd would be analytically
lazy and unfair.

Affirm's recent results describe a real, profitable, comparatively prudent business. In its fiscal third quarter
of 2026 it grew gross merchandise volume 35% to $11.6 billion, grew revenue 33% to about $1.04 billion, and —
crucially — generated genuine profit: operating income of $88.4 million at an 8.5% margin, net income of about
$103 million, and adjusted operating income of $281 million. Its credit metrics held up, with 30-plus-day
delinquencies on its core monthly-installment loans (excluding the Peloton legacy book) around 2.8%, and it
maintains a substantial allowance for credit losses. Most tellingly, while Klarna and Afterpay have declined to
participate in the new credit-scoring models that would incorporate BNPL data, Affirm has been working with FICO
to build exactly such models — choosing, in other words, to help make its lending visible rather than to keep
it phantom. Affirm also does more genuine underwriting and offers longer, interest-bearing installment products
that look more like conventional lending than the frictionless pay-in-four model that defines the riskier end.

So the thesis here is emphatically not "BNPL companies are all doomed" or "Affirm is a fraud." Affirm is a
legitimately well-run lender that happens to operate in a category with a systemic transparency problem. The
risk this essay describes is about the aggregate — the industry-wide, system-wide blindness created by tens of
billions of dollars of mostly-unreported subprime-skewed credit — and it is concentrated at the lower-quality
end. Affirm's relative quality is real and is precisely why the distinction matters: when the cycle turns, the
difference between the prudent lender and the volume-chasing one will be the difference between a manageable
loss rate and a serious one. The category's danger does not mean every name in it is equally dangerous.

The funding side is fragile too

There is a second vulnerability in the BNPL model that the phantom-debt discussion can overshadow, and it sits on
the lender's side of the ledger rather than the borrower's: these companies do not lend their own deposits the
way a bank does. They fund their loans through warehouse credit lines, forward-flow agreements, and
securitizations — selling or pledging the loans to capital-markets investors to raise the cash to make the next
batch of loans. It is the same originate-to-distribute machine documented elsewhere in this series, and it
carries the same fragility: it runs beautifully as long as the capital markets are happy to buy the paper, and
it seizes the instant they hesitate.

That dependence ties BNPL's fate to the credit cycle in a doubly dangerous way. In a downturn, two things happen
at once. First, the subprime borrowers at the core of the BNPL book begin to miss payments at higher rates,
degrading the quality of the loans. Second — and often faster — the investors who buy securitized BNPL loans
demand higher yields or simply stop buying, exactly as they did to subprime auto and other consumer-credit
originators in past stress episodes. A lender that depends on selling its loans to keep operating is only ever as
solvent as the capital market's current appetite for its loans, and that appetite is most likely to vanish at the
precise moment the loan quality is deteriorating. The phantom debt makes the borrower risk invisible; the funding
structure makes the lender risk acute. Together they describe a business that can look healthy and well-capitalized
right up until a credit-market door closes, at which point the model's two hidden dependencies — on continued
borrower payment and continued investor appetite — fail in the same quarter.

A regulatory limbo that suits the industry

It is worth asking why, if BNPL debt is so consequential, it remains largely unreported and lightly supervised —
and the answer is a regulatory limbo that has, conveniently, served the industry's growth. BNPL has occupied an
ambiguous space between a true loan and a mere payment plan, and that ambiguity has let it sidestep much of the
disclosure, reporting, and underwriting apparatus that governs credit cards and installment loans. Efforts to
bring it fully under consumer-lending rules — to mandate credit-bureau reporting, ability-to-pay checks, and
standardized disclosures — have advanced and retreated with the shifting priorities of regulators, and the
current posture has been more permissive than restrictive.

For the companies, that limbo is a feature: light-touch oversight is what permits the frictionless, no-hard-pull,
not-reported-anywhere model that drives the growth. But for the system, it means the measurement gap is not an
oversight that is about to be fixed; it is a structural condition that the industry has incentives to preserve and
that regulators have not consistently moved to close. Layer that on top of a macro backdrop of record-high
aggregate U.S. consumer debt, and the picture sharpens: at the very moment household balance sheets are most
stretched, the fastest-growing category of new consumer credit is the one the system has agreed to keep off its
books. The reforms that would make the debt visible — universal bureau reporting chief among them — are exactly
the reforms that would slow the growth, which is why the better operators' voluntary moves toward transparency
are the exception rather than the rule, and why the phantom is likely to stay a phantom until a downturn forces
it into the light.

What the bulls get right

In fairness, the case for BNPL as a genuine and largely benign innovation is real, and the skeptic must grant it.
Used as designed — a short, interest-free way to split a legitimate purchase into a few payments and pay it off
on time — BNPL is a sensible product that many consumers use responsibly and to their benefit, avoiding
revolving credit-card interest entirely. The majority of BNPL loans are repaid on time. The industry has grown
~20% a year because it solves a real friction at checkout and merchants love the conversion lift it provides. And
the better operators, Affirm chief among them, have demonstrated that the model can be run profitably and with
disciplined underwriting. A blanket condemnation of BNPL as predatory would be both unfair and empirically wrong.

The honest synthesis is narrower and therefore more durable: BNPL is a useful product with a dangerous
externality. The product is fine; the invisibility is the problem. Because the debt is largely unreported, the
financial system has lost the ability to measure a fast-growing, subprime-concentrated category of consumer
leverage — which means the true health of the stretched American consumer is worse-documented than it has been in
a generation, and the losses, when they come, will emerge from a blind spot the industry built on purpose. You
do not have to believe BNPL is evil to see that an unmeasurable pile of subprime credit, growing 20% a year on
top of maxed-out credit cards, is the kind of thing that ends cycles rather than the kind of thing that is fine.

The kicker

Every credit crisis in history has shared one ingredient: leverage that the people watching the system could not
see until it was too late. The subprime mortgage crisis was, at its core, a failure of measurement — risk that
had been sliced, repackaged, and scattered until no one could total it up. Buy now, pay later has reinvented that
failure in miniature and in plain sight, not through complex securitization but through the simplest mechanism of
all: by mostly not reporting the loans at all. The debt is real, it is growing, it is concentrated among the
borrowers least able to bear it, and it is sitting in a place the official statistics do not reach. The reassuring
delinquency numbers are not lies. They are just measurements of the part of the iceberg above the water — and the
defining feature of an iceberg is that the part you can measure is the part that was never going to sink you.

A nation is told its consumers are holding up fine, and the data agrees, because the data was built not to see
the seventy-billion-dollar pile of installment loans stacked on maxed-out cards by subprime borrowers financing
their groceries four payments at a time. The phantom debt does not show up — until the day it does, all at once,
in a number no one was watching, owed by people no one was counting.