The Goldman Sachs Group, Inc. GS

Key data

The business

Goldman makes money four ways. It advises companies on buying and selling each other and collects a fee when the deal closes. It underwrites, placing new shares and bonds for companies and governments and keeping a cut. It runs a markets business, dealing bonds, currencies, commodities and stocks for clients and its own account, earning the spread and the swings. And it manages money for the wealthy and for institutions, collecting a steady fee on the pile. The first three are lumpy: they gush in busy years and dry up in quiet ones. The fourth, asset and wealth management, is the annuity Goldman has spent years growing, because the market pays more for a recurring fee than for one good year at the desks. The consumer-lending experiment that cost billions is being wound down. The moat is the client roster and the balance sheet: the biggest companies call Goldman first, and few rivals can both advise on a deal and finance it.

The numbers

Profit has nearly doubled in a year, and the beat against expectations widened as it went.

The inflection is Q1 2026, when EPS stepped up to $17.55 from the low teens and stayed there. Q2 2026's $20.98 came against a $14.47 estimate, a 45% beat, the widest of the run.

You are looking at a recovery off a trough. The pandemic-era 2021 peak collapsed to the 2023 low, when a dead deal market met the consumer-banking losses. Net income has since doubled, 2023 to 2025, and the first half of 2026 already carries $38.53 of EPS, a pace that annualizes near $77 and lifts return on equity back to about 17%. Here is the tell in the pricing: FY2027 consensus is $73.34, below that 2026 run-rate. The market is not paying for the current pace to continue, it is paying for the pace to cool. The variant this memo holds is narrow: whether elevated deal and market activity persists a year longer than the Street's fade assumes, and the print that settles it is quarterly net revenues holding their year-ago level.

Management

Insiders sold and did not buy: 69 sales worth $59.2 million over the year, zero purchases. The largest came from General Counsel Kathryn Ruemmler ($7.8 million plus $2.5 million) and Controller Sheara Fredman ($7.6 million) in early May 2026, and the Form 4 footnotes do not state a 10b5-1 plan, so scheduled cannot be told from discretionary. Capital return is heavy and paid up: $12.4 billion of buyback in 2025 and $5.0 billion in Q1 2026 alone, retiring stock at 2.5 times book. Book value sits near $409 a share and the stock at 2.5 times it, unchanged from a read a day earlier, so the multiple has not drifted while earnings ran. Pay leans on return on equity and relative performance, which at least points the incentive at the number that matters.

How it fails or surprises you

Deal and market freeze. If M&A and the markets desks mean-revert the way they did in 2022 and 2023, when net income fell about 60%, then 16 times peak earnings quietly becomes 25 times normalized ones. The first tell is quarterly net revenues rolling under their prior-year level and the advisory backlog thinning.

The estimate that undercuts the story. FY2027 consensus of $73.34 sits below the 2026 run-rate near $77. Either the Street is too cautious, or 2026 is a peak it already sees fading. This is the fact the bull read explains least, and a single in-line-to-soft quarter that ends the beat streak would prove it.

The supercycle holds (right tail). If activity stays elevated and asset-and-wealth fees keep compounding, ROE holds above 15% through a soft quarter and the multiple re-rates toward the roughly 20x peers carry. Watch fee revenue in wealth management and whether ROE stays in the teens when the desks have an off quarter.

Closing thoughts

Goldman's earnings are set by how busy Wall Street is, and no single quarterly print tells you whether the current boom lasts or fades, what matters is that the balance sheet and franchise stay strong when it eventually cools. The fatter tail is the downside, because the base is elevated and consensus already flags the cooling: what is at risk if dealmaking stalls is a fast halving of earnings power, as 2022 to 2023 showed in real time, while what is on offer if the boom runs another year is a re-rate toward peers on top of a book that keeps compounding. The judgment, not a number, is that some normalization is more likely than none, but the balance sheet and the franchise come through any freeze intact, which is why the left tail is a drawdown and not a wipeout.

The bet is still that Goldman keeps earning fat fees from takeovers, underwriting and its markets desks, and that Wall Street stays busy. It breaks when net revenues roll over and the run of beats ends. The one pair that tells you first is quarterly net revenues against the year before, and whether EPS keeps landing above consensus or finally misses. When those two turn down together, the peak is behind you.