Goldman Sachs Just Delivered a Monster Quarter. I Want to Know How Much of It Is Repeatable.

A spectacular quarter has a dangerous second career. First it becomes an earnings release. Then it becomes a slide. By November, somebody has dragged it across four columns and called it next year’s plan.

Apparently, the market signed a service agreement guaranteeing that it will remain this helpful. I would like a copy. Legal probably would too.

Goldman’s latest results are good. I don’t need to dress up a strong performance as a problem to prove that I work in Finance. What interests me is the decision that follows: how much permanent spending would I authorize on the strength of a quarter like this?

That is where the celebration gets expensive. Revenue can have an exceptional afternoon. Headcount, leases and infrastructure prefer a longer relationship.

Goldman Sachs Q3 2026 earnings quality starts with what happened

Goldman Sachs reported third-quarter revenue of $18.68 billion, net earnings of $5.29 billion and diluted earnings per share of $17.01. Annualized return on average common equity was 18.7%. The October 13 release describes the revenue result as its second-highest quarter, rather than its highest ever. The company’s release is here.

Those figures deserve recognition. They don’t answer my planning question.

An annualized quarterly return is a rate calculated from one period. It isn’t a promise that the next three quarters will reproduce the economics. Multiplying quarterly revenue by four gives $74.72 billion. That calculation takes seconds. Establishing whether it is a defensible annual forecast takes rather longer.

I can admire the result and refuse to hire against that multiplication. Those positions fit together perfectly well. My calculator isn’t allowed to make commitments simply because it is faster than my judgment.

The company-specific facts above come from the release. The business examples below are hypothetical; they illustrate my analysis, not Goldman’s internal forecast.

The budget meeting I would interrupt

Picture a hypothetical business enjoying an unusually strong transaction quarter. Management wants to add permanent staff because demand is running ahead of capacity. Sales wants faster response times. Operations wants fewer bottlenecks. All sensible requests.

Then somebody says that the latest quarter proves the business can afford the additions.

I’d interrupt there. Affording a cost this quarter and supporting it through a normal cycle are different questions. Payroll will not politely decline when transactions do. It has direct debit.

I’d ask which part of the demand is contractual, which comes from an identifiable pipeline, and which arrived because conditions were unusually favorable. I’d also ask whether service delays are genuinely costing business or whether the request is mostly relief for an exhausted team.

Relief can be a valid investment. We should name it honestly. Otherwise we end up putting every worthwhile request under “growth,” like a household filing groceries, dental work and a speedboat under essentials.

I want to know what survives a quieter market

For a financial institution, client activity, transaction completion, market conditions, balances and fee rates can all influence revenue. They don’t move on the same calendar. A full pipeline can coexist with delayed closings. Higher assets can reflect market appreciation rather than new client money.

I’d map revenue to the economic event that produces it. An advisory mandate is different from a completed transaction. Assets under management are different from the fees recognized on those assets. High trading activity is different from a contracted subscription.

That classification matters because the forecast has to predict recognition, not enthusiasm. A prospective client who likes us has contributed exactly zero revenue until something happens that earns it.

Within each revenue stream, I’d separate retained business from new business and distinguish volume from rate or mix. If existing clients are doing more transactions, the sales story may be strong while the persistence assumption remains uncertain.

This is how I would examine repeatability. I wouldn’t assign a permanent label to a dollar merely because it appeared twice.

A worked example: the quarter everyone wants to adopt

Consider a hypothetical business unit with $100 million of quarterly revenue. Its expenses comprise $55 million of fixed and step-fixed costs and variable expenses equal to 20% of revenue. Quarterly operating profit is $25 million.

Now classify the revenue: $60 million comes from relatively stable relationships, $25 million from ordinary activity and $15 million from unusually favorable conditions. These are assumptions for the example, not disclosed Goldman figures.

The eager annual plan repeats the full quarter four times: $400 million of revenue, $80 million of variable expense and $220 million of fixed costs. Profit is $100 million.

My normalized case removes the $15 million exceptional component from each quarter. Annual revenue becomes $340 million. Variable expense becomes $68 million. With fixed costs unchanged, profit is $52 million.

We didn’t lose customers or discover an accounting scandal. We stopped assuming that exceptional conditions would renew automatically. Annual revenue is 15% lower than the enthusiastic plan; operating profit is 48% lower.

That’s why a revenue assumption deserves more attention than the font on the waterfall chart.

Now add the spending request

Suppose the unit proposes $8 million of additional annual fixed spending. Under the enthusiastic plan, profit falls from $100 million to $92 million. The spending looks comfortable.

Under the normalized case, profit falls from $52 million to $44 million. The same request consumes 15.4% of normalized profit. Against the enthusiastic case, it consumed 8%.

If the new capacity creates incremental revenue at an 80% contribution margin, it needs $10 million of annual additional revenue merely to cover the $8 million expense. That assumes no extra step costs, no implementation expense and no timing delay.

Those exclusions are exactly where I would keep digging. Hiring begins before revenue arrives. Training reduces initial productivity. Systems may need upgrading. A recruit is rarely delivered shrink-wrapped at full utilization.

I’d therefore make the sponsor show the ramp by month. A request that pays for itself in a full steady-state year may lose money during the first year. Both results belong in the decision.

The latest quarter cannot do that homework for us. It is busy being the latest quarter.

Revenue classifications need an expiration date

I wouldn’t freeze the $60 million, $25 million and $15 million categories forever. Evidence can move revenue between them.

A previously unusual transaction pattern might become routine as a market develops. A stable client relationship might weaken. An expected fee might fail to materialize. A classification is a judgment with a review date, not a tattoo.

For each material assumption, I’d record why it belongs in the base case and what would invalidate that treatment. Signed commitments, retention behavior, activity levels and realized pricing provide stronger evidence than a sponsor’s confidence.

I’d also avoid treating all uncertainty as downside. If the supposedly exceptional business continues for several periods and expands across clients, the model should recognize that improvement. Finance loses credibility when it clings to a cautious assumption after the evidence has changed.

My standard is disciplined updating. I won’t let optimism enter through a side door, but I’m equally unwilling to lock the front door when good evidence arrives.

Buying flexibility has a price

Permanent hiring is only one course of action. I’d compare it with temporary capacity, phased hiring and process improvements.

Suppose the $8 million permanent commitment could instead be met with a flexible arrangement costing $10 million annually when fully used. Flexibility carries a $2 million premium in the strong case. If demand falls and the arrangement can be reduced to $4 million, it costs $4 million less than the permanent alternative.

In a simplified one-year comparison, flexible capacity is preferable on expected cost when the probability of the weak case exceeds one-third: $2 million multiplied by the strong-case probability is offset by $4 million multiplied by the weak-case probability.

That threshold assumes equivalent quality, output and timing, and genuinely reducible costs. Real contracts may include minimum purchases or notice periods. “Flexible” is sometimes a word printed above a very inflexible termination clause.

I’d also price the cost of waiting. Missed client opportunities can make permanent capacity attractive. My point is to calculate the choice, rather than treat permanent expansion as the only respectable response to growth.

A phased plan can protect both service and cash

A third option is to authorize part of the investment immediately and release the rest after observable milestones.

For example, I might approve $3 million of initial capacity and reserve $5 million for a second phase. The release conditions could include sustained transaction volumes, demonstrated service constraints and booked contribution sufficient to cover the additional cost.

I would specify who measures those conditions and when. Otherwise a gate becomes an invitation for everyone to explain why their situation is special. Corporate stage gates can be remarkably easy to open when nobody owns the latch.

The phased alternative will not always win. Partial capacity might fail to solve the bottleneck. Recruiting in two waves can be more expensive. Some technology investment has indivisible minimum costs.

Those are reasons to test the design, not reasons to abandon it. I’d compare the operational effect of each phase with the actual constraint. Half a bridge is not useful. Half of a hiring plan might be.

Pipeline weighting can manufacture reassurance

A weighted pipeline looks scientific because it contains percentages. That doesn’t make its assumptions independent.

Imagine ten potential transactions, each carrying a 70% closing probability. If they all depend on the same financing window, adding their weighted revenues can conceal a shared exposure. They may all close or all slip.

I’d ask whether probabilities reflect demonstrated conversion by stage and customer type. I’d also examine correlated delays. A business can hit its eventual sales objective and still miss the quarter because every deal arrives one month late.

For planning, I would create a specific delay case alongside a demand case. The delay case shifts recognition and cash timing while preserving eventual volume. The demand case removes business. Management’s response should differ.

Cutting capacity for a timing slip may damage delivery when transactions resume. Holding costs unchanged through a genuine demand decline may destroy profit. A model that calls both outcomes “downside” has saved a label and lost a decision.

The strongest quarter can conceal a capacity debt

There is another possibility worth taking seriously: the business delivered the result by exhausting people or deferring work.

If strong output required excessive overtime, postponed maintenance or unusually heavy senior intervention, repeating it may require investment. In that case, refusing all spending because revenue is uncertain can be false economy.

I’d separate the cost of sustaining current service from the cost of expanding into new demand. The first may protect an existing franchise. The second needs an incremental growth case.

Suppose a $2 million annual systems improvement prevents $3 million of recurring processing cost at current activity. That is a different argument from an $8 million expansion depending on additional transactions. Combining them in one proposal can make the useful improvement hostage to a speculative growth forecast.

I want that distinction visible. Finance should be able to say yes to a well-supported investment without endorsing every assumption stapled to it.

A spending freeze is easy to administer. Thoughtful capital allocation requires more work, unfortunately for those of us who own the workbook.

Operating profit isn’t the whole permission slip

For a bank, capital and liquidity constrain growth alongside profitability. In other businesses, working capital and funding capacity can perform the same role.

I’d avoid transplanting an ordinary industrial cash model directly into a financial institution. Loans, deposits, risk-weighted assets and regulatory requirements change the analysis. Public quarterly profits alone don’t tell me how much growth the balance sheet can support.

For a hypothetical distributor, meanwhile, $10 million of additional annual revenue might require $1.2 million of incremental receivables and $800,000 of inventory, partly offset by $600,000 of payables. That’s $1.4 million of working capital before equipment or implementation costs.

The revenue may generate attractive contribution and still need cash upfront. For a bank, the analogous planning question includes capital consumption and funding economics rather than simply inventory days.

The lesson travels across sectors. The calculation does not travel unchanged. I’d rather admit that than force every business into a model because the template has already been approved.

My forecast would show what management can change

A useful downside case needs actions, dates and realistic cost behavior.

I’d show the revenue effect before actions and the profit result after actions. If management can defer $3 million of planned spending but only from July, the current-year benefit is not $3 million. Time has once again failed to respect the PowerPoint.

I’d distinguish expenses we can avoid, expenses we can defer and expenses we have already committed. A deferred bill still exists. Eliminating a future vacancy differs from terminating an occupied position. A variable compensation pool may respond to performance; technology contracts may not.

For each action, I’d capture service consequences and restart costs. Aggressive cuts can improve the quarter while making the recovery harder. The model should expose that tradeoff.

I would also run an upside case that includes the costs required to deliver the upside. Revenue growth with no capacity consequence is possible in some settings. In others, it is fiction with a particularly attractive margin.

How I would keep the review honest

I’d use a rolling review with a small number of decision-relevant signals: actual activity, realized pricing, retained business, pipeline timing and utilization of the proposed capacity.

The thresholds would reflect our economics. A minor revenue fluctuation doesn’t justify rewriting the organization. A sustained decline in contribution beneath the level funding a planned expansion does justify revisiting it.

I’d preserve the original forecast alongside updated versions. That lets us distinguish model error from new information. Otherwise a moving forecast can become so accommodating that management never technically misses anything.

I also want assumptions attached to owners who can explain them. Finance shouldn’t be expected to invent a transaction forecast after Operations and Sales have supplied only their aspirations.

There is no shame in saying that a forecast remains uncertain. There is considerable shame in presenting a precise answer whose confidence comes mainly from formatting. Decimal places are not a substitute for knowing how the money arrives.

What I would authorize after a quarter like this

I would approve spending backed by recurring economics, documented capacity constraints or a strong return under a range of plausible conditions. I would stage investment supported mainly by favorable activity. I would challenge commitments that require the best quarter to repeat indefinitely.

That is a positive investment policy. It gives useful projects a way forward and tells sponsors what evidence earns more funding.

I’d make the recommendation concrete: the initial amount, the remaining authorization, the release conditions and the response if conditions weaken. A board shouldn’t need to infer our decision from twelve slides of scenarios.

My wider argument about 2027 budget assumptions and scenario planning applies here: an assumption earns its place through evidence and decision usefulness.

The goal isn’t to win an argument with an optimistic executive. It’s to preserve the ability to invest when the next opportunity arrives. Permanent commitments made casually can spend that ability before anyone notices.

I want the next investment to deserve the last quarter

Goldman’s results are impressive. A strong franchise should use strong periods to build capabilities, support clients and improve its economics. I am entirely in favor of that.

What I oppose is the quiet conversion of favorable conditions into permanent obligations without anyone testing the dependency. That isn’t confidence. It is an expensive assumption enjoying diplomatic immunity.

My recommendation is to fund proven needs, buy flexibility where uncertainty warrants it and demand explicit milestones for the rest. If the evidence strengthens, invest more. If it weakens, keep enough room to respond.

A good quarter should improve our choices. It shouldn’t remove our judgment.

When somebody tells me that the latest result means we can afford anything on the wish list, I want to see what happens when the revenue normalizes. The payroll department will still be here. So will the lease. The spectacular quarter may have other plans.

Sources and calculation notes

Company figures and the quarter’s ranking come from Goldman’s October 13, 2026 results announcement. All unit models, spending choices, probabilities and working-capital figures are illustrative. They are not Goldman guidance or estimates of its segment economics.