Goldman Sachs nearly doubled its profit, led the SpaceX flotation and advised on $1 trillion of deals — with 2% fewer people than three months earlier

Every figure in this article is drawn from Goldman Sachs' second-quarter 2026 results and Form 10-Q, or from CNBC as attributed. Sections marked as analysis are identified as such.

Goldman Sachs earned $6.63 billion in the three months to June, against $3.72 billion in the same quarter a year earlier. Diluted earnings per share went from $10.91 to $20.98. Net revenues were $20.34 billion.

Headcount over the same period went down, to 46,200, 2 per cent lower than at the end of March.

Where it came from

Equities trading produced record net revenues of $7.42 billion, up 72 per cent. Fixed income, currencies and commodities added $4.59 billion, up 32 per cent. Between them the trading businesses generated roughly $12 billion in a quarter — a scale of market activity that does not occur in calm conditions.

Investment banking fees reached $3.40 billion, up 55 per cent, and the composition matters more than the total. Equity underwriting rose 130 per cent and debt underwriting 75 per cent. Goldman led the initial public offering of SpaceX in late June — the largest flotation ever completed — and advised on more than $1 trillion of announced mergers and acquisitions in the first half of the year.

The cycle turned

For three years the investment banking business was widely described as structurally impaired. Interest rates had risen, the initial public offering window had closed, private equity could not sell its holdings, and the fee pools that support several thousand bankers simply were not there. Goldman spent that period cutting, retrenching from consumer banking, and being told its business model was a relic of a cheaper-money era.

Equity underwriting up 130 per cent in a single year is what the end of that period looks like. When companies can float again, the entire chain reactivates: private-equity owners exit, sponsors deploy fresh capital, corporates use their own newly valued shares as acquisition currency, and the mergers that were postponed in 2023 arrive at once. A trillion dollars of announced deals in six months is not a recovery. It is a backlog clearing.

The composition of that backlog says something about the wider economy as well. Equity underwriting rising twice as fast as debt underwriting means companies are choosing to sell shares rather than borrow — a preference that appears when management teams think their stock is well valued and financing is expensive. It is the signature of a late-stage bull market rather than an early one.

Analysis: the headcount line is the story

The most quietly remarkable number in the release is 46,200 people, 2 per cent fewer than three months earlier, producing nearly double the profit. Investment banking has historically been the most operationally leveraged business in the economy in the wrong direction — revenue arrived in cycles, compensation did not fall as fast as it rose, and headcount ratcheted upward in every boom.

Goldman appears to have broken that pattern, at least for now. The firm has been unusually explicit about deploying artificial intelligence across document review, pitch preparation, coding and the enormous volume of routine analytical work that has traditionally been done by first- and second-year analysts working through the night. It has also been disciplined about not rehiring into the recovery.

It is worth being careful about how much credit to give the technology. Two per cent of headcount over one quarter is a small number, and some of it is ordinary attrition in a strong labour market for bankers. The claim worth testing over the next two years is not whether AI removed jobs this quarter, but whether the firm can carry a doubling of deal volume without the hiring wave that has accompanied every previous boom.

If that combination holds through a full cycle, it changes the economics of the industry rather than just this quarter's earnings. The historic knock on investment banks as investments was that their good years were given away in compensation and their bad years were absorbed by shareholders. A bank that can take a 55 per cent fee increase without adding staff is a different kind of asset.

The counter-pressure is competitive rather than technological. Bankers are mobile, compensation is the main thing that moves them, and a rival prepared to pay up during a boom can take a team out of a firm in an afternoon. Restraint of this kind is easy to maintain in a downturn and extremely difficult in a market where everyone is making money — which is why the interesting test arrives at the end of this year, not now.

Analysis: what to be careful about

Trading revenue of $12 billion in a quarter is exceptional and it is also, by nature, the least predictable line in the firm. Record equities results reflect volatility, client repositioning and financing balances — all of which are conditions rather than franchises. They can reverse in a quarter, and historically have.

The banking pipeline is more durable but it is not immune. A deal cycle powered by a reopened IPO market depends on the market staying open, and the single largest transaction of the period — the SpaceX flotation — is precisely the kind of event that is easy to attribute to institutional strength when it may be an unrepeatable mandate.

Concentration in the fee line is a familiar feature of good years in this business. A handful of very large mandates can lift a quarter, and the same league-table position that wins one of them does not guarantee the next. That is why the trillion dollars of announced M&A is arguably the more meaningful figure than the SpaceX listing: advisory work spread across dozens of transactions is a better indicator of franchise strength than any single deal, however large.

There is also the ordinary risk in a very good quarter: it becomes the comparison. Goldman's second half will be judged against a period in which equities set a record and fees rose 55 per cent, and a firm reporting flat revenue next year will be reported as slowing while doing nothing wrong.

None of that changes what happened. Goldman Sachs went into this cycle having been told its core business was in secular decline, cut its way through the trough, declined to staff back up, and has come out of it earning twenty-one dollars a share in a quarter. Whether or not the deal boom lasts, the firm that emerged from the downturn is measurably more efficient than the one that went in — and that part does not depend on the market staying friendly.

The broader read-across is about the deal cycle itself. Investment banking fees are among the earliest and most honest indicators of corporate confidence, because they measure decisions that chief executives and boards make only when they believe the future is legible. Two years of near-silence followed by a trillion dollars of announced transactions in six months is a clearer statement about how the corporate world currently feels than any confidence survey.