Goldman Sachs Tried to Bank Everyone. It Forgot It Was Built to Bank No One.

In April 2016, Goldman Sachs — the firm that had spent a century arranging money for governments, corporations, and the very rich — quietly bought GE Capital Bank's US online deposit platform and went looking for the savings of ordinary people. The deposit acquisition came first. The lending brand, Marcus, came later that same year, offering no-fee personal loans to anyone with a decent credit score.¹ By the end of 2017, Marcus had originated $2 billion in loans.¹ Seven years on, Goldman was selling almost the entire thing for parts. The retreat is usually told as a graceful strategic refocus. It was nothing of the sort.

The official story is that Goldman tried something bold, learned a lesson, and wisely went back to what it does best. The truer story is that the firm walked into a business it was structurally unequipped to run, lost money at scale doing it, and retreated only when the bleeding became impossible to hide. The exit wasn't wisdom arriving. It was reality catching up.

A firm built to bank no one tried to bank everyone

Here is the thesis, plainly. Goldman is an institution optimized to serve a few hundred enormous clients with bespoke, high-margin advice. Mass-market consumer lending is the opposite discipline: millions of small accounts, razor-thin spreads, relentless cost control, and an underwriting machine that has to be right about strangers at industrial volume. Goldman had none of the muscles that game requires — no branch network feeding cheap sticky deposits, no decades of consumer credit data, no culture that tolerates a business losing money for years before it compounds. It tried to buy its way past all three at once. The losses were not bad luck on timing; they were the predictable cost of competing without the things competing requires.

Notice who actually started it. The legend credits David Solomon, but Marcus was conceived and launched in 2016 under Lloyd Blankfein, and Goldman's own history lists it among the Blankfein-era initiatives.² One of Solomon's eventual rivals for the top job, Harvey Schwartz, was among the original backers.⁸ Solomon inherited the bet, then doubled it — which matters, because it means the consumer dream wasn't one man's vanity. It was an institutional conviction. The whole firm wanted to believe it could bank everyone. That conviction is exactly what made the unraveling so expensive.

It's a tidy line, and that is the problem with it. 'Too much too quickly' frames the failure as a pacing error — as if the same strategy run slower would have worked. But pace wasn't the flaw. The flaw was building a mass-lending business without the foundations a mass lender stands on. Slower wouldn't have grown Goldman a branch network or a consumer credit history. It would only have stretched the same losses over more years.

The number Goldman could not keep off the page

When Goldman recast its segments, the math finally surfaced in a filing. Its Platform Solutions unit — the home for Apple Card, GreenSky, and transaction banking — lost $783 million in 2020, then $1.05 billion in 2021, then $1.21 billion in just the first nine months of 2022: more than $3.03 billion of pretax losses, accelerating year over year.³ And that figure is the understatement, not the headline. It deliberately excludes the Marcus consumer-banking losses, which sat in a separate segment and were never cleanly consolidated into one public number.³ The true bill was larger than the one anyone could point to.

Then there was the buying. In September 2021, Goldman agreed to acquire GreenSky — 'the largest fintech platform for home improvement consumer loan originations' — in an all-stock deal worth roughly $2.24 billion, closing the following March.⁶ On paper it looked like a discount: GreenSky had been valued near $4 billion at its 2018 IPO. But cheap is not the same as good. Goldman was buying a distressed asset, and it still managed to lose money on it — agreeing to sell GreenSky to a Sixth Street-led consortium in 2024, less than two years after the ink dried.⁶ Buying lending at the bottom did not save a firm that didn't know how to run it.

By 2023 the dismantling was systematic. Goldman completed the sale of substantially all of its Marcus unsecured personal loan portfolio, entered an agreement to sell GreenSky, and arranged to hand the General Motors credit-card program to another issuer.⁵ The firm had already eliminated the very segment that once housed Marcus, folding into three business lines.⁴ You don't reorganize the org chart around a strategy you still believe in. You reorganize to make the exits look like architecture.

But didn't Marcus pull in $50 billion in deposits?

The strongest case for the defense is that Marcus genuinely worked at one thing: gathering money. Within three years of its 2016 launch it had attracted some $50 billion in deposits — a real, low-cost funding base that any bank would envy.⁸ So the retreat, a defender argues, was a half-success abandoned too early: Goldman proved it could win consumers' savings and simply lost its nerve on the lending side. That's the fair objection, and it's partly right. The deposit gathering was the one piece that fit Goldman's strengths — a single product, sold online, to millions who never needed a relationship manager.

But it cuts the other way too. Gathering deposits is the easy half; deposits are a liability you have to profitably deploy. The hard half is turning that cheap money into loans that don't default — and that is precisely the half that lost more than $3 billion.³ CNBC's post-mortem put it bluntly: the business 'rapidly expanded and ultimately buckled under the weight of Solomon's ambitions,' with internal partner leaks to the press accelerating the unwind.⁸ A firm that can raise funding but can't deploy it safely hasn't built a bank. It has built an expensive way to pay interest. The $50 billion isn't evidence the strategy worked. It's evidence Goldman could do the part anyone can do, and not the part that mattered.

Goldman didn't retreat to what it does best out of strategic clarity. It retreated because the alternative was to keep funding losses it could no longer explain, in a business it never had the right bones for. The deposit beachhead, the bought fintech, the celebrity card — each was an attempt to rent a capability the firm couldn't grow, and rented capabilities don't compound; they just bill you. The real lesson isn't 'too much too quickly.' It's that a firm built to bank no one spent seven years and billions discovering it could not bank everyone. The most expensive thing Goldman learned is the cheapest thing to say: be great at what you're built for, and suspicious of every market that merely looks adjacent.