Teardown

Goldman Sachs just printed $17.55 diluted EPS on $5.63 billion in net income, and the Street is cheering, but I want you to slow down and look at what this filing is actually built on — because the composition of that beat matters more than the beat itself.

Here is the first thing you need to understand about Goldman's income statement. They do not report revenue and cost of revenue the way an industrial company does. There is no gross margin line sitting neatly in the data. That is not an accident. Goldman's accounting structure reflects a trading and advisory business where net revenues are the operative concept — you net financing costs, you net trading losses against gains, and what you report is what remains. So when you see nulls across the gross margin and operating margin fields, that is not a data problem. That is Goldman's architecture. And that architecture is load-bearing for how you interpret any quarter they report.

Now let's talk about that $5.63 billion. On a trailing basis, this is a number that reflects an extraordinarily favorable operating environment for the businesses Goldman does best. Equities trading, in particular, has been printing across the street this cycle. Volatility is the product. When markets are dislocated — rate uncertainty, geopolitical noise, positioning unwinds — Goldman's market-making franchise generates spread and flow that a calmer quarter simply cannot replicate. That is not a criticism. That is the business model. But it means you cannot extrapolate Q1 2026 in a straight line.

The number I want professionals in the room to think about is the relationship between that EPS print and diluted share count. $5.63 billion in net income divided by $17.55 per share implies roughly 321 million diluted shares. Goldman has been a consistent and aggressive repurchaser. The buyback is a real driver of EPS growth independent of earnings growth. This matters because if net income plateaus but shares outstanding decline, EPS can still climb. That is not manipulation — it is capital allocation. But when you are building a normalized earnings power framework, you need to separate operating leverage from financial engineering on the share count, or you will overestimate the organic earnings trajectory.

The piece most people walk past: Goldman's return on equity. At a $309 billion market cap and $5.63 billion in a single quarter, you are looking at a quarterly ROE that, annualized, would be exceptional relative to their own historical range and relative to where large cap banks typically earn through a cycle. The question is whether the fee wallet in investment banking is durable here, or whether this quarter captured advisory and underwriting revenue that was deferred from softer prior periods and is now pulling forward. DCM and ECM pipelines have been reopening. M&A backlog has been converting. That creates a timing dynamic — you get a fat quarter as the backlog clears, and then you need new mandates to refill it.

What the Street tends to misprice on Goldman specifically is the operating leverage at the compensation line. Goldman accrues compensation as a ratio of net revenues. When revenues spike, comp accruals spike with them. The question in a strong quarter is always whether non-comp expenses held discipline. If non-comp costs are growing ahead of revenue on a percentage basis, the operating leverage story starts to erode quietly before anyone talks about it publicly.

The filing itself is a 10-Q, period ending March 31st. What you want in the actual document is the segment breakdown — Global Banking and Markets versus Asset and Wealth Management — because those two businesses have fundamentally different margin profiles and different cyclicality. One is transactional and volatile. The other is sticky, fee-based, and Goldman has spent years deliberately growing it. The ratio between them tells you whether Goldman is structurally less cyclical than it was a decade ago. The answer, slowly, is yes — but Q1 probably flattered the trading side of that ratio.

That is the teardown. The numbers were always there — most people just do not look. See you at the next filing.