Goldman Built a Consumer Bank to Win the Crowd. The Crowd Wasn't the Problem.

In October 2016, the most prestigious investment bank on earth started taking $5,000 personal loans from people it would never have let through the lobby. It called the platform Marcus, after Marcus Goldman, the co-founder who started the firm in 1869.¹ The pitch was elegant: take the freshness of a digital offering and bolt it onto the strength and heritage of Goldman Sachs. The brand that closed billion-dollar deals would now help a dentist in Ohio consolidate her credit-card debt. It sounded like the future. It became one of the costliest detours in the firm's history.

The official story is that Goldman tried consumer banking, found the margins thin and the competition brutal, and wisely walked away. That's the comfortable version. The real one is sharper: the part of Marcus that actually worked never got the chance, because Goldman bolted two structurally weak partnerships onto a business it had never built the plumbing to run — and the bill for exiting came in higher than the venture ever returned.

The deposit machine was fine. The bank around it wasn't.

Start with what Goldman got right. In April 2016 it bought GE Capital Bank's US online deposit platform, then launched Marcus with no-fee unsecured personal loans later that year.² Gathering deposits online is the one piece of consumer banking that maps neatly onto what Goldman already was: a firm trusted with money, paying a competitive rate, with no branches to fund. Deposits flowed in. That much worked the way the strategy deck promised.

Lending is the other half, and lending is a different sport. A consumer lender lives or dies on the unglamorous machinery: underwriting models tuned over years of loss data, servicing operations, collections, fraud systems, dispute handling at retail scale. Goldman had spent a century being brilliant at the opposite — bespoke, high-margin, relationship deals where one mistake is one client, not one million accounts. It entered a volume business with the instincts of a boutique. The consumer push lost roughly $1.3 billion from inception in 2016 through mid-2019, by data Goldman itself shared with the SEC, even as the firm told itself the business would break even by the end of 2022.⁴ The break-even date was a forecast. The losses were the receipt.

Two partnerships that handed away the very thing that printed money

Rather than slow down and build the missing infrastructure, Goldman accelerated by partnering. In March 2019 it announced it was the issuer behind Apple Card, and by that October David Solomon — who had inherited Marcus, not built it — was calling it 'the most successful credit card launch ever.'⁵ In September 2021 Goldman agreed to buy GreenSky, the largest fintech platform for home-improvement consumer loans, in an all-stock deal worth about $2.24 billion.⁶ On paper this looked like a shortcut to scale. In practice it was the wrong half of the value chain.

Here is the mechanism almost everyone misses. In a card partnership like Apple's, the partner owns the customer, the interface, the brand love, and the design of the product. The bank owns the balance sheet, the credit risk, and the regulatory liability. Goldman took the seat with all the downside and almost none of the relationship — for a customer base it didn't control, on terms it didn't set, in a product whose loyalty belonged to a phone. It had become, in effect, a rented balance sheet. The franchise that was actually working — cheap, sticky deposits — got buried under the franchise that wasn't.

Once Goldman reorganized these pieces into a 'Platform Solutions' segment, the bleeding became impossible to hide. By the bank's own SEC filing, the segment lost $783 million in 2020, $1.05 billion in 2021, and $1.21 billion through just the first three quarters of 2022 — more than $3 billion in under three years, on top of the earlier consumer losses.³ No 'most successful credit card launch ever' survives that arithmetic intact.

The retreat cost more than the adventure

Unwinding turned out to be its own expensive education. GreenSky, bought in 2021 for about $2.24 billion in stock, was sold to a Sixth Street-led consortium in a deal that closed March 15, 2024 — a round trip measured in years and write-downs.⁶⁷ Then came the headline exit. On January 7, 2026, Goldman announced it would transition the Apple Card program and its $20 billion-plus portfolio to Chase, a move expected to take roughly 24 months and still requiring regulatory approval.⁸ Read the mechanics of that disclosure and you see the true cost of leaving: Goldman expected a $0.46 EPS bump in the fourth quarter of 2025, built from a $2.48 billion reserve release partly offset by $2.26 billion in revenue reductions from portfolio markdowns and contract-termination obligations.⁸ You pay to get in. Then you pay to get out.

Wasn't this just a reasonable bet that didn't pay off?

The honest counter is that diversifying away from volatile trading and advisory revenue toward stable consumer fees was a defensible thesis — plenty of sober people backed it, and the deposit base really did work. Fair. But the lesson isn't 'don't diversify.' It's that Goldman diversified into the half of consumer banking it was worst equipped to run, and outsourced the customer relationship that would have made the risk worth carrying. A patient firm builds the underwriting and servicing muscle first and grows the lending book slowly behind its own brand. Goldman did the reverse: it kept the credit risk and rented out the customer. The bet wasn't doomed by markets. It was doomed by which seat the firm kept choosing to sit in.

Marcus was supposed to prove that the heritage of the firm could be wrapped around the future of banking. It proved something quieter and more useful instead: a great brand can buy you a seat at a table where you don't know the game, and the table will happily take your money to teach you. Goldman gathered the deposits it was built to gather, then handed the customers it was never going to own to a bank that already knew how to keep them. The detour ended where it should have started — with Goldman doing the one thing it was actually good at, and letting Chase do the rest.