Goldman Sachs Spent a Decade Learning One Lesson: It Doesn't Know How to Lose Money to Strangers.

In the third quarter of 2022, Goldman Sachs set aside $515 million in provisions for credit losses — against a consumer-banking unit that had been generating hundreds of millions in net revenue.⁴ Read that twice. The most prestigious risk-pricer on Wall Street - the firm that earns its keep telling other people what their risk is worth - was bleeding two-thirds of its consumer revenue straight back out the door to cover the consumers it had lent to. This was the firm that prices sovereign debt and structures derivatives, discovering that a person with a personal loan is a different kind of animal entirely.

The official story is that Marcus, Goldman's consumer brand, was a savings-account experiment that flopped. Almost every part of that is wrong. The savings account never flopped - Goldman kept it, because cheap deposits are exactly the kind of thing an institution loves to own. What flopped was the part nobody talks about: the part where Goldman owned the credit risk.

The brand everyone blames wasn't the thing that broke

Start with what Marcus actually was, because the popular memory has the chronology backwards. The deposit platform came first: in April 2016 Goldman bought GE Capital Bank's US online deposit book, the low-cost savings engine. Marcus as a brand launched that October - and it launched offering no-fee unsecured personal loans, not savings accounts.¹ The two halves got merged under one friendly first name, and that naming choice did Goldman a quiet disservice for years, because it let the public file everything under "Marcus" and assume the savings account was the disaster. It wasn't. A deposit you hold cheaply and lend against is the oldest profitable trick in banking. The loan you make to a stranger and pray gets repaid is a different business, and it was that business - personal loans, partner cards, buy-now-pay-later - that did the damage.

Here is the thesis in one line: Goldman didn't fail at consumer banking because it couldn't run a savings account. It failed because it confused a fee-based, institutional-client model it has mastered for two centuries with a balance-sheet business where it personally eats the losses - and it does not know how to eat the losses of millions of strangers. The Marcus story isn't a product flop. It's the cost of buying the risk you don't know how to price.

Three adjacency bets, each one closer to the edge

What makes this instructive is that it wasn't one bad decision. It was a sequence, each step pulling Goldman further from the thing it's good at and deeper into the thing it isn't. The unsecured personal loans put its own capital behind borrowers it had never met. Then came the partner cards - the Apple Card and the GM card - where Goldman supplied the balance sheet and ate the credit risk while a glamorous brand owned the customer relationship. That is a structurally lopsided trade: you carry the defaults, someone else carries the love.

Then came the acquisition that crystallized everything. In September 2021 Goldman agreed to buy GreenSky, a buy-now-pay-later lender, in an all-stock deal worth about $2.24 billion.⁵ The detail that should have been the warning was right there: at the time of Goldman's announcement, GreenSky's stock had already fallen roughly 70% from its peak.¹⁰ Goldman was buying more consumer credit exposure - a near-prime, cyclical lending book - at the very moment the cycle was about to turn against it. Two years later, Goldman announced the sale of GreenSky to a Sixth Street Partners-led consortium in October 2023 — a transaction that closed in March 2024 — at a material loss.¹¹ The whole arc took twenty-five months: buy high into the risk, sell low out of it.

Why an institutional genius can be a consumer amateur

The causal mechanism is the whole point, so work it down. Goldman's core business is fee-based and institutional: it advises, underwrites, and trades for a small number of sophisticated counterparties, and the firm is paid for its judgment rather than for warehousing the outcome. Mass-market consumer lending inverts every one of those properties. The counterparties are millions, not dozens. They are unsophisticated, which means adverse selection is brutal - the people most eager for an unsecured personal loan are disproportionately the people you least want to lend to. And the risk doesn't get sold off into a fee; it sits on your balance sheet as a provision, growing quietly until a downturn detonates it. That is exactly the shape of the losses: $783 million in 2020, $1 billion in 2021, and $1.2 billion in just the first three quarters of 2022.³ The line doesn't drift up. It accelerates - which is the signature of a portfolio whose true risk you underestimated going in.

And notice what Goldman did and didn't keep when it retreated. In Q1 2023 it began winding down the personal-loan book - a partial sale plus a transfer of the rest to held-for-sale, which let it release roughly $440 million of reserves it no longer needed against loans it no longer held.⁸ By October 2023 Goldman had announced the GreenSky sale, with the transaction closing in March 2024.¹¹ But the savings and CD deposits? Those it pulled back inside, into Asset & Wealth Management, because cheap, sticky funding is precisely the thing an institution treasures. Goldman didn't exit consumer banking. It exited the half where it owned the loss, and kept the half where it owned the funding. That selective amputation tells you exactly where the disease was.

Wasn't this just bad timing - a downturn anyone would have caught?

The fair objection is that the cycle did it. Goldman launched into a benign credit environment, the rate shock of 2022 spiked defaults across the whole consumer-lending industry, and any new entrant carrying a young, untested loan book would have looked terrible at exactly that moment. There's truth in it - timing genuinely hurt, and the GreenSky purchase price was at least partly a casualty of buying late in a frothy market. But the steelman concedes the real point rather than rescuing it. A mature consumer lender survives a downturn because it has decades of loss data, a seasoned book, and a model calibrated through prior cycles. Goldman had none of those - it had a new portfolio, an institutional risk culture poorly matched to mass-market underwriting, and a regulatory apparatus (the Fed reportedly pressed it on Marcus risk management and governance) noticing it had moved faster than its controls.³ "Bad timing" is what you call it when a seasoned operator gets unlucky. When the operator is new to the business and the loss line accelerates, the more honest word is mispriced.

Goldman went looking for a consumer franchise and found a mirror. It already knew how to gather deposits and lend them - that part it kept. What it learned, at a cost of roughly $3 billion and a great deal of institutional pride, was that the skill it sells to the world's most sophisticated clients does not transfer to a stranger asking for an unsecured loan. The firm that built its fortune pricing other people's risk discovered, expensively, that the hardest risk to price is the one you can't sell to anyone else. So it sold the loans, sold GreenSky, kept the savings - and went back to charging for its judgment instead of betting its balance sheet on it.