Goldman's Best Quarter Ever: Downgrading to Hold as 3.1x Book Capitalizes a Peak-Cycle ROE

Results vs. Consensus

Q2 2026 Scorecard

A 45% EPS beat is not a normal event, and it deserves an honest framing rather than a victory lap. Sell-side models do not forecast trading windfalls; they extrapolate. The consensus process failed to model a quarter in which equities intermediation rose 60% year over year, and the resulting "surprise" measures the gap between a spreadsheet and a volatility regime, not a 45% improvement in franchise quality. What follows is an attempt to separate the two.

Year-Over-Year Comparisons

Quarter-Over-Quarter Comparisons

Revenue

The $4.0B revenue beat is almost entirely one line: equities intermediation at $4,157M against roughly $2.7B in each of the prior two quarters. That is the volatility-linked business, and it is the piece least likely to repeat at this level. Underneath it, though, the mix improved in ways that matter beyond one quarter. Equities financing of $3,259M (+91% YoY) and FICC financing of $1,216M (+14%, a record) combine to $4,475M, or 37% of the FICC-and-equities pool, up 62% year over year. Financing is a spread business built on prime balances and structured lending. It does not evaporate when the VIX does. Geographic mix tells the same story: Asia revenue doubled to $3,614M and rose from 12% to 18% of the firm, which is a deliberate, funded share gain rather than a windfall.

Margins and Expenses

The Q2 efficiency ratio of 57.4% compares with 60.5% in Q1 and 63.4% a year ago, and the first-half ratio of 58.8% sits below the firm's 60% target. Operating expenses rose 12% sequentially against 18% revenue growth, and compensation grew 30% year over year against 39% revenue growth. The honest caveat, which management volunteered rather than hid, is that this is operating leverage on a revenue surge and not a re-based cost structure. Transaction-based expenses of $3,052M (+56% YoY) are the tell: a large slice of the non-comp increase is a direct variable cost of the very trading activity that drove the beat. If revenue normalizes, most of that expense normalizes with it, but so does the flattering ratio.

EPS

The bridge from a 73% pre-tax gain to a 92% EPS gain is worth walking. Pre-tax earnings rose 73% to $8,563M. Net earnings rose 78% to $6,628M, helped by a modest year-over-year tax tailwind (22.6% versus 24.9%). Net earnings to common rose 84% to $6,399M as preferred dividends fell to $229M. Diluted EPS then rose 92% to $20.98 as the average diluted share count fell 4% to 304.9M. So roughly two-thirds of the EPS growth is pre-tax operating performance, with the balance split between the tax line, the preferred stack, and the buyback. One disclosure deserves a flag for the full-year model: share-based award tax benefits reduced first-half taxes by about $965M, worth roughly $3.15 of first-half EPS and 1.7 points of ROE. That benefit is concentrated in Q1 and does not recur at the same scale, which is exactly why the full-year tax rate guide of ~20% sits above the 18.5% first-half rate.

Segment Performance

Geographic Mix: The Asia Story

Asia is the most underdiscussed number in the release. Revenue doubled year over year and the region moved six points of firm mix in twelve months, from 12% to 18%. On a first-half basis Asia produced $6,658M against $3,495M, up 90%. This was not luck. Management described a deliberate decision, taken after receiving regulatory capital relief at the turn of the year, to deploy people, technology, and balance sheet against what it viewed as a competitive shortcoming in Asian equities.

Assessment: A doubling of regional revenue on a funded, pre-announced investment is the strongest evidence in this print that Goldman took share rather than merely rented volatility. It is also the concentration risk: 6 points of firm mix now depends on a region whose activity is tied to the AI capital formation cycle, and Asian trading booms have historically cooled as abruptly as they arrive. Management was notably unwilling to size the client concentration behind it.

Investment Banking

IB fees of $3,395M (+55% YoY, +20% QoQ) were carried entirely by underwriting. Equity underwriting of $985M rose 130% on marquee mandates including the SpaceX IPO and the Alphabet equity raise, and debt underwriting of $1,032M rose 75% to the best quarter on record on leveraged finance and asset-backed activity. Advisory, by contrast, was the one revenue line in the entire firm that declined sequentially, falling 8% to $1,378M even though it rose 17% year over year.

That advisory dip looks like a blemish and is actually the most bullish datapoint in the release, because of how the revenue is recognized. Advisory fees book at deal completion, not announcement. Goldman advised on $1.2 trillion of announced volume in the first half, became the first bank ever to cross $1 trillion of announced volume in a six-month period, and holds roughly a $425B lead over its closest competitor. Large-cap M&A volumes rose 90% in the first half. Almost none of that announced volume has hit the advisory line yet.

"Even with very strong investment banking revenues this quarter, our backlog increased to its highest level in 5 years and its second highest level on record, underpinned by a record advisory backlog and reflecting the strength and breadth of our client engagement." — David Solomon, Chairman and CEO

Assessment: The advisory engine has not yet fired. A record advisory backlog plus a five-year-high total backlog, built after a quarter that already produced $3.4B of fees, points to advisory acceleration in the second half and into 2027 as announced deals close. This is the single strongest argument for the bulls and the reason our full-year estimate moves up sharply. It is also, unfortunately, the most visible fact in the release and the one most thoroughly reflected in a 9% rally.

Equities

Equities produced $7,416M (+72% YoY, +39% QoQ), an all-time record for any bank in a single quarter and the third consecutive quarter in which Goldman's equities unit has set that record. The composition matters more than the headline. Intermediation of $4,157M (+60%) reflects derivatives and cash activity in a market defined by high single-stock dispersion against a rising index, a combination management correctly identified as unusually supportive of client hedging and repositioning. Financing of $3,259M (+91%) reflects record average prime balances, concentrated in Asia.

"There's a dynamic in the marketplace right now where we're seeing single name equity dispersion relative to index, and we're seeing that dispersion relative to index while the market's going up. That sort of concoction is very supportive for the activity that clients are undertaking to manage their own portfolios and their own returns." — Denis Coleman, Chief Financial Officer

Assessment: Split the franchise in two and the investment case gets clearer. The financing half is an annuity: it grows with prime balances, it is priced on spread, and management sees pricing leverage because competitors are capacity-constrained. Underwrite it near current levels. The intermediation half is rent collected on a specific market regime, and Coleman's own description of the "concoction" driving it is a description of conditions, not of a moat. Q1 intermediation was $2,718M and the year-ago quarter was $2,595M. A normalization toward $3.0B to $3.5B would cost roughly $700M to $1.2B of quarterly revenue with no change whatsoever in Goldman's competitive position.

FICC

FICC delivered $4,592M (+32% YoY, +14% QoQ), with intermediation up 39% on interest rate products, commodities, and mortgages, partially offset by weaker credit products, and financing up 14% to a record $1,216M. Last quarter FICC was the problem: it came in roughly $860M below consensus, and we argued at the time that the shortfall reflected client caution in a specific macro environment rather than any structural impairment, and that the line would recover toward $4.5B to $5.0B once rate clarity improved.

Assessment: That call resolved correctly and quickly. FICC at $4,592M landed inside the recovery range we set one quarter ago, and it did so through intermediation, which is exactly where the Q1 weakness sat. The bear point that FICC is structurally challenged should be retired; the bear point that FICC intermediation is cyclical stands, and it is the same caveat that now applies with greater force to equities.

Asset & Wealth Management

AWM revenue of $4,597M (+20% YoY, +13% QoQ) came with a record management and other fees line of $3,355M (+20%) and total assets under supervision crossing $4 trillion for the first time, at $4,041B. Wealth management client assets reached roughly $2 trillion. Alternatives fundraising was a record $59B in the quarter and $85B in the first half, prompting management to raise its full-year fundraising expectation to more than $125B, including $31B of private credit raised in the quarter alone. The firm also won Verizon and Lockheed Martin retirement mandates representing $70B of AUS.

Three things temper the enthusiasm. First, the headline inflow number is not what it appears. Management celebrated a 34th consecutive quarter of long-term net inflows at $91B, but a footnote in the release discloses that this figure includes $31B of inflows from the Innovator Capital Management acquisition, partially offset by $15B of outflows from the disposition of Goldman Sachs TFI. Organic long-term inflows were therefore closer to $75B. That is still an excellent number, and it is not the number that was said out loud. Second, of the $4.04T AUS headline, $1,065B is liquidity products, which earn minimal fees; the fee-generating long-term base is $2,976B. Third, the first-half pre-tax margin of 24% remains six points short of the 30% target set at Q4 2025, and the segment ROE of 13.5% is barely half the firm's 23.5%.

"Across asset and wealth management, we are relentlessly driving our growth strategy forward with quarterly management and other fees up 20% year-over-year. We delivered our 34th consecutive quarter of long-term net inflows, including $19 billion in wealth management." — David Solomon, Chairman and CEO

Assessment: AWM is the segment the bull case needs most and the one delivering least. It is the durable, fee-based counterweight that is supposed to justify a higher multiple across the cycle, yet it earns a 13.5% ROE while GBM earns 25%, and its margin sits six points below target three quarters after that target was set. The fee growth is genuine and the record AUS provides a real base. But a quarter in which the segment's revenue was flattered by $441M of private equity marks and its inflow headline was flattered by an acquisition is not the quarter that proves the AWM re-rating thesis.

Platform Solutions

Platform Solutions produced $221M, down 64% year over year and 46% sequentially, driven by net markdowns on the Apple Card loan portfolio that was moved to held-for-sale in Q4 2025. Management guided quarterly revenues for the remainder of the year to be broadly consistent with this level.

Assessment: The wind-down continues to be noisy but immaterial, and it is now roughly 1% of firm revenue. The guide to a flat ~$221M run-rate is the first explicit forward marker management has given on this segment, which modestly improves model clarity. The markdowns are unwelcome but small enough that they no longer move the thesis.

Key KPIs

Key Topics & Management Commentary

Overall Management Tone: The most assured posture in at least two years, and notably, assurance without triumphalism. Management delivered records across every major line as the mechanical output of investments made years ago rather than as a surprise, and it repeatedly volunteered the limits of its own story: that the cycle will not move in a straight line, that recalibrations are coming, that there has been no structural change in the expense base. Where the call was less convincing was on quantification. The three questions that would most sharply test the durability of the quarter, concerning client concentration in equities, the size of the cross-sell multiplier, and how low the leverage ratio can run, each drew a framework rather than a number.

1. The Equities Record: Financing Is the Story, Intermediation Is the Rent

The $7.42B equities result is the third consecutive all-time record for any bank, and the temptation is to treat it as a single event. The more useful decomposition is that Goldman now runs two equities businesses with entirely different economics. Financing at $3,259M (+91%) is a balance-sheet spread business tied to record average prime balances. Intermediation at $4,157M (+60%) is a flow business tied to a market regime.

Management was explicit that financing growth was manufactured, not received: a multi-year program of hiring, risk systems, and technology investment, accelerated after regulatory capital relief at the turn of the year, aimed specifically at an identified share gap in Asia.

"Obviously, the performance of our equities business in the last quarter is the result of a number of multi-year investments. We obviously identified a long time ago, particularly on the equity financing piece of the equation, a commitment to grow that as a component of our GBM public business, and we've been putting in place the talent, the risk management capabilities. We're making multiple years of technological investments to underpin our capabilities, particularly in international jurisdictions such as Asia." — Denis Coleman, Chief Financial Officer

Assessment: Underwrite the financing half near current levels and haircut the intermediation half hard. Across FICC and equities, financing now runs at $4,475M per quarter, 37% of that combined pool and up 62% year over year. That is a genuinely higher floor than the franchise carried three years ago and it is the best argument for a structurally higher multiple. But it is a floor of roughly $18B annualized, not $30B, and the difference between the two is the volatility rent that the current price appears to treat as permanent.

2. The Binding Constraint Migrates from CET1 to SLR

This is the most consequential disclosure in the release and it received the least airtime. Goldman's SLR fell 40bps to 4.3%, now the lowest among its peer group, while CET1 improved to 12.9%. The two moved in opposite directions for a reason: prime financing consumes balance sheet and leverage capacity but carries low risk weights. Grow it aggressively and the risk-based ratio looks better while the leverage ratio erodes.

The exchange this drew in Q&A, examined in detail below, produced a confirmation and a refusal in the same breath: management acknowledged it runs multiple "oscillating" binding constraints and manages to all of them, then conceded the decisive point without quantifying it.

"You're right that we have facilitated some balance sheet expansion to facilitate client activity. Ultimately, there will be a limit to our appetite to expand that." — Denis Coleman, Chief Financial Officer

Assessment: "Ultimately, there will be a limit to our appetite to expand that" is management confirming a ceiling on the growth engine without saying where it sits. The financing franchise is the durable half of the equities story and the half that requires leverage capacity to compound. At 4.3% and falling 40bps a quarter, the runway is finite and the firm declined to size it. Basel III finalization and a friendlier leverage framework, both of which management is openly lobbying for, would relieve this. Until then, the most durable part of the bull case is capped by the least discussed number in the release.

3. Advisory Fell While M&A Boomed, and That Is Bullish

Advisory revenue of $1,378M declined 8% sequentially, the only revenue line in the firm to fall, in a half during which Goldman advised on $1.2 trillion of announced volume and became the first bank to cross $1 trillion in a six-month period. The reconciliation is recognition timing: advisory books at completion. The CFO attributed the 17% year-over-year gain to "higher completed volumes," which is the same mechanism working with a lag.

The forward marker is the backlog, which rose to a five-year high and the second-highest on record, underpinned by a record advisory backlog, even after a quarter that already produced $3.4B of IB fees. Management also flagged that sponsor volumes remain subdued versus historical averages, which it framed as unrealized upside rather than weakness.

Assessment: The advisory line is a coiled spring, and this is the cleanest forward-looking signal in the print. A record advisory backlog after a record fee quarter means the announced-volume boom has not yet reached the income statement. We expect advisory to accelerate through the second half and into 2027 as deals close, and our full-year IB estimate moves up accordingly. The uncomfortable part for the rating: this is the most legible fact in the release, and legible facts get priced.

4. The AI Capex Supercycle and the Limits Management Volunteered

Solomon's framing of the opportunity is expansive. The AI investment cycle is pushing capital needs beyond core technology into infrastructure, energy, and data centers, generating demand for structuring, financing, risk management, and capital markets execution across both public and private markets. Goldman sits at the intersection of all of it, and the marquee mandates cited (lead left on the SpaceX IPO, the Alphabet equity raise, Dominion's sale to NextEra, the Comcast spin of NBCUniversal) support the claim.

What is notable is how much downside management volunteered without being pushed hard.

"All the indicators we have is that we are in the relative early innings of a very, very significant, when you're talking about the AI build-out cycle, of an AI build-out cycle. We all know, because we've all been around for a long time, that these things don't go in a straight line, and they can ebb and flow. I'm not smart enough to tell you whether or not there can be recalibrations, in the short term, sometime in the next six months, the next 18 months." — David Solomon, Chairman and CEO

Assessment: This is the correct posture and it is more candid than the price action suggests investors heard. Management is telling you the cycle is early, real, and non-linear, and that it cannot underwrite the path over the next six to eighteen months. An investor paying 3.1x book is underwriting exactly that path. The asymmetry between what management is willing to promise and what the multiple now requires is the core of our downgrade.

5. Capital Allocation: Clients First, Buybacks Last, and a 25% Dividend Raise

Goldman returned $5.36B in the quarter, comprising $4.00B of buybacks (4.1 million shares at an average $984.57) and $1.36B of dividends, and the Board raised the quarterly dividend to $5.00 from $4.50, a 25% increase year over year and 150% over five years. The buyback stepped down from Q1's $5.00B, which is directionally what we predicted last quarter, though our forecast of a $1.5B to $2.5B quarterly pace proved far too conservative. We underestimated how much capital the RWA reduction would free.

Solomon's articulation of the priority stack was unusually blunt, and it is a mild negative for anyone modeling buybacks as the primary return vehicle.

"It is always our desire to take our capital and allocate it toward supporting our clients and their activities when we can do so in a way that produces benefit for our clients and accretive returns for our shareholders. That would always be our preference. In fact, if we could do that and have no excess capital, we would, obviously with a buffer, but no excess capital for buybacks if it's generating accretive returns." — David Solomon, Chairman and CEO

Assessment: Read that quote carefully and it says buybacks are the residual, not the plan. In an environment where management sees financing demand outstripping what it is willing to supply, the capital goes to the client book first. That is the right call for long-run value and it caps the pace of share retirement precisely when the stock is expensive. The dividend raise to $5.00 is the more meaningful signal: it is the commitment management cannot quietly reverse, and at $20 annualized it yields 1.75% on the current price. Note also that Q2 repurchases at an average $984.57 are already 16% in the money against the $1,140 close, which is good execution and a reminder that the buyback got cheaper stock than today's buyer.

6. Operating Leverage Without Structural Change

The first-half efficiency ratio of 58.8% improved 320bps year over year and sits under the firm's 60% target, with the compensation ratio net of provisions at 31%. Revenue grew roughly 40% while comp grew 30% and non-comp grew 22%. Management then declined to claim credit for it as a structural achievement.

The mechanism management described is real but conditional: the OneGS 3.0 initiative has let the firm absorb a surge in client activity without growing headcount proportionally, which is a genuine change in how marginal revenue is produced. Headcount actually fell 2% sequentially to 46,200 while revenue rose 18%.

Assessment: The candor here is the useful part. When the CFO says there has been no structural change in the expense base, he is telling you the 57.4% quarterly efficiency ratio is a function of the numerator. Transaction-based expenses of $3,052M (+56% YoY) are variable and will fall with activity, but so will the revenue they support, and the fixed base has not been re-set. Model the efficiency ratio back toward 60% in a normalized revenue environment, not 57%.

7. The Funding Mix Quietly Deteriorated

Total assets grew $68B to $2,128B, but deposits fell $3B to $558B. The balance sheet growth was funded substantially by unsecured long-term borrowings, which rose $33B to $348B. Loans grew 3% sequentially to $261B, and global core liquid assets averaged $555B against $494B in Q1.

Assessment: This got no airtime and it deserves some. Deposit growth stalling while the balance sheet expands means the marginal funding for the financing engine is wholesale and more expensive than deposits. Net interest income of $3,954M (+27% YoY) is not yet showing strain, and one quarter of flat deposits is not a trend. But if prime and FICC financing are the durable growth story, the cost and stability of the funding underneath them becomes a thesis variable rather than a footnote. Watch the deposit line next quarter.

8. Why This Beat Was Not Tax-Aided

The effective tax rate was 22.6% in Q2 against 13.2% in Q1 and a full-year guide of approximately 20%. In other words, Q2 carried a tax rate above the full-year expectation, and the tax line was a sequential headwind that suppressed reported EPS growth relative to pre-tax growth.

The reason Q1 was so low is disclosed in a footnote: share-based award tax benefits reduced first-half taxes by roughly $965M, adding about $3.15 to first-half EPS and 1.7 points to first-half ROE. Those benefits arise on the settlement of employee awards, which cluster in the first quarter.

Assessment: Two implications. For the quarter, this is a mark of quality: a 45% EPS beat with the tax line working against you is as clean as a beat gets. For the model, it is a caution: the first-half EPS of $38.51 embeds roughly $3.15 of a benefit that does not recur at the same scale, and the second-half tax rate must run near 21.5% to average the guided 20%. Anyone annualizing first-half EPS is overstating the run-rate by about 4%.

9. The One GS Flywheel That Will Not Be Quantified

The strategic centerpiece of Solomon's remarks is the multiplier: an advisory mandate in the boardroom becomes financing, risk management, capital markets execution, distribution, and eventually a wealth management relationship with the founders whose stock just became liquid. The firm cites roughly 900 referrals to wealth management from investment banking since the start of 2025 as evidence.

Asked directly to size it, management declined, offering a framework instead and one revealing phrase about competitive position.

"We're finding ourselves in these discussions earlier and alone and without other banks, that allows us to command more of the wallet structurally." — David Solomon, Chairman and CEO

Assessment: "Earlier and alone" is the most economically loaded phrase on the call, and if it is true it is worth more than any single quarter's records. The problem is that it is unfalsifiable as presented. A firm that genuinely commands structurally more wallet per advisory dollar should be able to show the attach rate, and 900 referrals over six quarters is an activity metric, not a revenue one. We accept the direction and decline to pay for the magnitude until it is measured.

10. Headcount Down 2% While AI "Will Not Replace What Matters Most"

Solomon addressed the AI-and-workforce debate directly, arguing that AI expands the capabilities of the firm's people rather than substituting for them. In the same release, headcount fell 2% sequentially to 46,200 while revenue rose 18% sequentially, and the CFO attributed part of the efficiency gain to producing marginal revenue without growing the human capital footprint proportionally.

"There has been much debate around the broader implications of AI on the workforce. It will change how work gets done, it will not replace what matters most in driving our business, our extraordinary people." — David Solomon, Chairman and CEO

Assessment: The rhetoric and the disclosure point in different directions, and for shareholders the disclosure is the better news. A firm growing revenue 18% sequentially on 2% fewer people is demonstrating operating leverage of exactly the kind the efficiency target requires. The tension is worth naming rather than resolving: management is unwilling to frame AI as a headcount story publicly, which is understandable, but if the productivity gain is real it will eventually show up as the structural expense change the CFO says has not yet happened.

11. Regulatory Posture: Lobbying in Plain Sight

Both executives spent meaningful airtime on regulation. Goldman was pleased with its CCAR result, its stress capital buffer of 3.4% is unchanged and effective through September 2027, and management is openly pushing for Basel III finalization, better stress-test calibration, and "better alignment of regulatory outcomes with underlying risk."

Assessment: This is not boilerplate given the SLR disclosure. The firm running the lowest leverage ratio among peers, while growing the leverage-intensive business hardest, has the most to gain from a recalibrated leverage framework. Regulatory relief is a genuine call option on the thesis and management is working it. It is also outside management's control and outside our ability to underwrite, so we assign it no value in the base case while acknowledging it as upside.

Guidance & Outlook

Goldman gives no quarterly or annual EPS guidance. What it provides is a set of structural targets plus a handful of specific forward markers, and this quarter produced an unusually rich set of the latter.

The forward markers that matter most are the backlog, the incentive fee guide, and the alternatives fundraising raise, because all three point to second-half revenue that is not in the second-quarter run-rate. Incentive fees at $112M were a trough and are guided materially higher. Alternatives fundraising was raised to more than $125B for the year against $85B already booked in the first half. And the advisory backlog is at a record after a record fee quarter.

Implied second-half setup: First-half EPS of $38.51 does not annualize cleanly to $77, because roughly $3.15 of it is a first-quarter tax benefit and because the second-half tax rate must run near 21.5% to average the guided 20%. Working the other way, the advisory backlog, the incentive fee step-up, and a full quarter of the Verizon and Lockheed OCIO mandates all add revenue that Q2 did not contain. The swing factor is equities intermediation: at $4,157M it sits roughly $1.4B above its own trailing two-quarter average, and no part of the guide suggests management expects that to persist.

Guidance style: Consistent with Goldman's practice of structural targets over point estimates, but this call carried more explicit forward markers than any in recent memory (a Platform Solutions run-rate, an alternatives fundraising raise, an incentive fee direction, and a tax rate). That is a management team comfortable enough with its position to be pinned down on specifics. It is also, read cynically, a management team that knows the second half needs help offsetting an equities line that cannot repeat.

Analyst Q&A Highlights

Whether the Equities Quarter Can Repeat

The opening question of the call went straight to the number everyone wanted explained, and framed it as a puzzle rather than a triumph: market volumes rose 9% while Goldman's equities revenue rose far more, and risk-weighted assets somehow went down. Management answered the "how" (multi-year investment in financing capability and a deliberate Asia push) but conspicuously did not answer the "will it repeat," which the questioner had pre-emptively acknowledged was hard to address.

Q: "I'm going to try to pick on one here in equities. I know it's hard to comment on sustainability. The environment's good. A lot went right this quarter. Maybe we could talk about wallet share gains and any concentration we should think about, because I saw overall market volume's up 9%, but margin was up 50%, and your revenue was up like 86% or whatever. Yet somehow RWA is down."
— Glenn Schorr, Evercore ISI

A: "Obviously, the performance of our equities business in the last quarter is the result of a number of multi-year investments. We obviously identified a long time ago, particularly on the equity financing piece of the equation, a commitment to grow that as a component of our GBM public business, and we've been putting in place the talent, the risk management capabilities. We're making multiple years of technological investments to underpin our capabilities, particularly in international jurisdictions such as Asia."
— Denis Coleman, Chief Financial Officer

Assessment: The answer is true and incomplete. Multi-year investment explains the financing line, which grew 91% and is the durable half. It does not explain intermediation growing 60%, which is a regime, not an investment. Management chose to answer the question about the durable half and let it stand for the whole. The RWA observation embedded in the question went unaddressed on the call and is, as it turns out, the quarter's most important structural fact.

Client Concentration Inside the Equities Book

The follow-up was the sharpest question of the call and drew the least satisfying answer. The premise was historically grounded: Asian trading businesses have produced bursts of growth before that later cooled, and such bursts are often driven by a handful of very large clients. The response listed client types and asserted diversification without offering a single figure.

Q: "Can you talk a little bit about the client composition? I ask because parts of the business have had bursts of growth across Asia, and then things cool down. Maybe you could talk about what kind of clients we're talking about, and then if we should be thinking about any concentration risk, because at times there's a handful of big clients that drive that."
— Glenn Schorr, Evercore ISI

A: "We have very big clients in a lot of our businesses, they are very influential, this is a highly diversified suite of clients across the world. We are intermediating equity asset flows in cash, derivative format, financing format for a multitude of clients all over the world."
— Denis Coleman, Chief Financial Officer

Assessment: "We have very big clients... this is a highly diversified suite of clients" is an answer that concedes the premise and denies the conclusion in the same breath, without evidence for either. Asia went from 12% to 18% of firm revenue in twelve months, largely on prime balances. If a handful of large clients drive a meaningful share of that, the earnings base is more fragile than 23.5% ROE suggests, and the question is entirely fair. Not answering it is a choice. This is the clearest unresolved risk on the call.

Where Capital Goes, and Whether Buybacks Are the Plan or the Residual

A question about the relative attractiveness of deploying capital across regions and businesses versus returning it produced the most strategically revealing answer of the call, and one with a real implication for anyone modeling share count.

Q: "Talk to us in terms of this capital allocation when you're looking through the businesses and through markets, where are the best opportunities? ... how you're thinking about capital allocation across your markets business in Asia versus Europe versus the United States, market financing versus all the CapEx AI activity that's going on, and then how you think about all of that relative to the buybacks that we did this quarter."
— Ebrahim Poonawala, Bank of America

A: "It is always our desire to take our capital and allocate it toward supporting our clients and their activities when we can do so in a way that produces benefit for our clients and accretive returns for our shareholders. That would always be our preference. In fact, if we could do that and have no excess capital, we would, obviously with a buffer, but no excess capital for buybacks if it's generating accretive returns."
— David Solomon, Chairman and CEO

Assessment: This is management telling you plainly that buybacks are the residual claimant on capital, not a commitment. In a quarter where financing demand is described as exceeding what the firm is willing to supply, the implication is that share retirement decelerates when opportunity is richest. That is correct capital allocation and a genuine negative for the mechanical EPS-accretion leg of the bull case. The dividend, raised 25% year over year, is the harder commitment and the better signal.

Durability of the AI Capex Cycle

The follow-up pressed on bubble risk, citing management's own widely reported remark that there is more greed than fear in the market. The answer was more balanced than the record results might have licensed, and it declined to underwrite the near term.

Q: "Talk to us about the durability of the AI CapEx cycle. ... I think, David, you were quoted in the press during the quarter about there's more greed than fear in the market. As investors think about the risk of maybe an investment bubble around AI ... maybe give us a sense of comfort that you have around how durable this cycle could be when we think about the financing business and what that means for Goldman and Goldman's earnings outlook."
— Ebrahim Poonawala, Bank of America

A: "All the indicators we have is that we are in the relative early innings of a very, very significant, when you're talking about the AI build-out cycle, of an AI build-out cycle. We all know, because we've all been around for a long time, that these things don't go in a straight line, and they can ebb and flow. I'm not smart enough to tell you whether or not there can be recalibrations, in the short term, sometime in the next six months, the next 18 months."
— David Solomon, Chairman and CEO

Assessment: Management explicitly refused to underwrite the next six to eighteen months while affirming the three-to-five year direction. That is intellectually honest and it is also the exact window over which a 12-month rating must be judged. The stock rallied 9% into a view that management declined to endorse at that horizon. Note too the specific mechanism flagged for the eventual recalibration: uncertainty about how enterprises will buy the infrastructure once built, how it gets priced, and how chip efficiency changes the economics. Those are demand-side questions that no amount of Goldman franchise quality answers.

The Asia Trade and Pricing Power in Prime

A two-part question asked how much of the equities result is the Asia hyperscale trade, whether a recent correction there is concerning, and whether constrained industry prime capacity is translating into pricing power. Management answered the pricing half affirmatively and simply did not engage the sizing or the correction.

Q: "How much is the Asia hyperscale trade driving the equities revenues at Goldman? Does the recent correction, is that any cause for concern? Secondarily, we've also heard that given the balance sheet demand, in Asia, that prime capacity has been more limited. I'm wondering, given your financing number, how that played out for Goldman Sachs and whether or not in prime you're starting to see a little bit of pricing power, given the global demand for balance sheet."
— Erika Najarian, UBS

A: "We do see opportunities for pricing leverage. There are some participants in the market that are also, I'll say, being more disciplined, and there's a lot of desire from our clients to engage. We're going to remain selective and careful about how we grow the business. There's a big demand from clients. There's a secular growth opportunity that, given the multi-year set of investments we've made, it's something that we can capture."
— Denis Coleman, Chief Financial Officer

Assessment: Confirmed pricing power in prime is a real and underappreciated positive, and it supports underwriting the financing line near current levels rather than haircutting it with intermediation. But the question about the Asia correction was left on the table, for the second time on this call. Two separate analysts asked, in different ways, how concentrated and how durable the Asia-driven surge is. Neither got a number. That silence is the single largest gap between this print and the price it produced.

How Low the Leverage Ratio Can Go

The question that most directly tests the growth engine asked how much the leverage ratio governs the financing build and how low management is willing to run it. The answer confirmed a ceiling exists and declined to locate it.

Q: "Hear you on the CET1 rising, SLR did fall 40 basis points to 4.3%. Clearly there's a lot of growth in sort of low RWA leverage-intensive financing like prime. Your SLR ratio is now the lowest among peers. How much does SLR govern constraints on growing the financing business? Then maybe give us a sense of how low you're willing to run that ratio."
— Christian Bolu, Autonomous Research

A: "We have a number of different, often oscillating, binding constraints as a firm. We manage to all of them. We're managing the CET1, we're managing the SLR. ... Ultimately, there will be a limit to our appetite to expand that. ... We will look at the opportunity set that our clients are presenting to us and then make choices about the relative resource allocation."
— Denis Coleman, Chief Financial Officer

Assessment: An acknowledged but unquantified ceiling on the most durable growth driver in the story. "Oscillating binding constraints" is a sophisticated way of saying the firm optimizes across CET1 and leverage dynamically, which is true and also means the leverage ratio is now the one doing the binding. A 40bps quarterly decline from an already peer-low base cannot continue for many quarters without either a regulatory change or a deliberate slowdown in the financing build. Management is lobbying for the former and staying silent on the latter.

Whether the Efficiency Gain Is Structural

A question distinguishing cyclical operating leverage from genuine cost takeout produced the most useful admission on the call, and management made it without prompting or hedging.

Q: "On expenses, pretty impressive. Your, I think, efficiency ratio for the first half was under your 60% target. I think you're still investing in kind of cloud and AI under OneGS 3.0. Just curious about how much of the expense or efficiency gains you've had so far is a function of just a strength in the markets business versus what you've already taken out from a structural cost perspective."
— Christian Bolu, Autonomous Research

A: "We've been able to grow our revs at about 40%, revs net of PCL slightly higher. Comp expense is only growing at 30% and non-comp at 22%. There's been disciplined growth on the expense side relative to top line. ... We're developing marginal levels of revenue production, and sort of not growing our human capital footprint quite the same way. ... I would not say that there's been any structural change in our expense base at this point."
— Denis Coleman, Chief Financial Officer

Assessment: "I would not say that there's been any structural change in our expense base at this point" is the most valuable sentence on the call, and it is management arguing against its own best-looking metric. The 57.4% quarterly efficiency ratio is a numerator story. It should be modeled back toward 60% as revenue normalizes, which is precisely what our second-half estimate does. Credit management for saying so plainly rather than letting the ratio speak for itself.

The Multiplier Effect the Firm Will Not Size

The most economically pointed question of the call asked the firm to put a number on the cross-sell flywheel that is the strategic centerpiece of its narrative: for every dollar of merger fee, how much additional revenue follows across financing, risk management, execution, and wealth. Management declined, twice, then offered a framework and one unusually revealing claim.

Q: "I guess you said your merger lead in advising on announced deals is 50% higher than the next largest, and that you have record advisory backlogs. What is the multiplier effect? For every $1 of merger fee that you generate, how much more do you get in all those other activities from the financing and risk management and execution and all those other things that you say? What's the multiplier effect? Is it like 10% elsewhere, 20%, 50%?"
— Mike Mayo, Wells Fargo Securities

A: "I'm not going to give you a specific percentage, and to be honest, I don't know if I could give you an exact percentage that I could underwrite on a public call. ... We're finding ourselves in these discussions earlier and alone and without other banks, that allows us to command more of the wallet structurally. ... I can't give you a percentage, but it's a meaningful part of the integrated performance of the firm and our One GS operating ethos."
— David Solomon, Chairman and CEO

Assessment: The candor of "I don't know if I could give you an exact percentage that I could underwrite on a public call" is worth more than a fabricated number would have been. But the flywheel is the load-bearing claim in the equity story, invoked to justify why Goldman's franchise deserves a premium to its trading-house history, and after several quarters of telling it the firm still cannot measure it. "Earlier and alone" is either the most valuable competitive statement made this quarter or an unfalsifiable assertion, and the firm has not given the market the data to tell which. We are not paying 3.1x book for an unmeasured multiplier.

What They're NOT Saying

1. That $16B of the AWM inflows were bought, not won: Management celebrated a 34th consecutive quarter of long-term net inflows at $91B on the call. The release footnote discloses that the figure includes $31B of inflows from the Innovator acquisition against $15B of outflows from the TFI disposition, putting organic inflows near $75B. The organic number is excellent and would have survived being said out loud. Presenting an acquisition-inflated figure as an unbroken organic streak is the least candid moment in an otherwise candid disclosure.
2. How concentrated the Asia franchise is: Two analysts asked, in different forms, what client concentration sits behind a region that doubled to 18% of firm revenue. Both received qualitative reassurance and no data. Given that Asian prime and trading booms have historically cooled quickly, and that this is now the marginal driver of the firm's marginal earnings, the absence of any sizing is the most material omission in the quarter.
3. Where the SLR floor is: Management confirmed "there will be a limit to our appetite" for balance sheet expansion but declined to say where. With SLR at 4.3%, peer-low, and falling 40bps in a single quarter, the runway for the financing engine is finite and unquantified. This is a ceiling on the most durable part of the bull case and it went undiscussed except in response to one analyst's question.
4. Any framing of equities intermediation normalization: Intermediation of $4,157M sits roughly $1.4B above its own trailing two-quarter average. Management characterized the market "concoction" driving it without ever suggesting what the line looks like when that concoction ends. No range, no normalized level, no acknowledgment that the prior two quarters ran at $2.6B to $2.7B. Investors are left to do that arithmetic themselves, and the 9% rally suggests many did not.
5. That deposits fell while the balance sheet grew: Total assets rose $68B; deposits fell $3B; unsecured long-term borrowings rose $33B. The funding mix for the financing build is shifting toward wholesale, which is more expensive and less stable than deposits. It received zero airtime in prepared remarks and drew no questions.
6. The tension between the AI rhetoric and the headcount: Solomon stated that AI "will not replace what matters most in driving our business, our extraordinary people," in a quarter where headcount fell 2% sequentially and the CFO credited the efficiency gain partly to producing marginal revenue without growing the human capital footprint. Both statements can be reconciled, but management chose not to reconcile them, and the productivity story is the more valuable one for shareholders.
7. Why advisory fell sequentially: The one revenue line that declined got no explicit explanation in prepared remarks beyond attributing the year-over-year gain to "higher completed volumes." The recognition-timing answer is favorable to Goldman and easy to give. Leaving it implicit in a quarter of record announced volumes means the most bullish structural point in the release was left for readers to assemble.

Market Reaction

• Pre-print setup: Closed at $1,045.91 on July 13. Up 19.0% year to date against the S&P 500's 9.8%, up 46.6% over the trailing twelve months, but down 1.6% over the trailing 30 days. The 52-week closing range entering the print was $700.41 to $1,106.37, so the stock came in roughly 5% below its own 52-week closing high after a month of consolidation.
• Reaction session (July 14, print day, before the open): Opened at $1,085.00, a 3.7% gap. Traded a range of $1,082.00 to $1,143.84. Closed at $1,140.00, up 9.0% (+$94.09), a new 52-week closing high and effectively the high of the day.
• Volume: 3.8 million shares against a 2.3 million 30-day average, or 1.7x normal.
• Benchmark: The S&P 500 rose 0.4% on the session, so essentially all of the move was idiosyncratic.

The shape of the session is as informative as its size. The stock gapped 3.7% on the print, then spent the day grinding higher to close within $4 of the intraday high, which is the signature of a market re-rating a franchise rather than covering shorts into a headline. The gap captured the beat; the drift captured the call. Between the open and the close, management delivered the backlog disclosure, the confirmation of pricing power in prime, and the dividend raise, and buyers kept paying up for all of it.

Consider what had to break for this reaction to happen. Goldman entered the print down 1.6% over 30 days and 5% below its 52-week closing high, having consolidated a large twelve-month gain. That is a setup with room to run on good news, and a 45% EPS beat qualifies. The move cleared the prior 52-week closing high decisively, on 1.7x volume, with the index flat.

Our reading is that the market did not re-rate the whole franchise so much as extrapolate the equities line. The beat was concentrated in equities intermediation, the one component with the weakest claim to persistence, and the session priced it at close to full value. Meanwhile the two facts that most complicate the story, the peer-low leverage ratio and the acquisition-inflated inflow headline, sat in a footnote and a single Q&A exchange. A 9% single-day move that takes a mega-cap bank to 3.10x book, on a quarter management itself declined to characterize as structurally repeatable, is the market pricing the good news and skipping the fine print. That gap is the entire basis of our rating change.

Street Perspective

Debate: Is the Equities Record a Franchise Re-Rating or Volatility Rent?

Bull view: The bull case on the Street holds that this is a permanent step-up. Financing revenue rose 91% on record prime balances built through years of deliberate investment, financing now represents 37% of the combined FICC-and-equities pool, and management confirmed pricing power because competitors are capacity-constrained. Three consecutive record quarters is a pattern, not an accident, and the Asia share gain was funded and pre-announced rather than stumbled into.

Bear view: The bear camp contends that a 60% year-over-year jump in intermediation is what a specific dispersion-plus-rising-index regime pays, and regimes end. Intermediation ran at $2.6B to $2.7B in each of the two prior quarters and printed $4.2B in this one. Nothing in Goldman's competitive position changed in ninety days, so roughly $1.4B of quarterly revenue is a rent on market conditions, and capitalizing it at 15x is a category error.

Our take: Both are right about different halves, and the split is the whole analysis. Financing is an annuity worth underwriting near current levels; intermediation is rent worth haircutting hard. The mistake the bulls make is treating a 37% financing mix as though it de-risks the other 63%. The mistake the bears make is dismissing a genuine, funded share gain in Asia that doubled a region's revenue. Our estimate splits them, holding financing and normalizing intermediation toward $3.0B to $3.5B, and the resulting earnings power still supports a much higher estimate than we carried in April. That is why this is a Hold and not a downgrade to Underperform.

Debate: Does the Record Backlog Guarantee Second-Half Advisory Acceleration?

Bull view: The optimistic camp argues the advisory line is the next leg and it is nearly mechanical. A record advisory backlog and a five-year-high total backlog, built after a quarter that already delivered $3.4B of IB fees, plus $1.2 trillion of first-half announced volume and a $425B league-table lead, means completions and therefore fees accelerate through the second half and into 2027. Sponsor volumes remain subdued, which is upside not yet in any number.

Bear view: Skeptics note that backlog is not revenue and that the conversion depends on deals actually closing. The company's own risk language flags exactly this: transactions may be modified or not completed at all, and financing conditions, regulatory approval, and geopolitical disruption all sit between announcement and fee. A record backlog in a market where management concedes recalibrations are possible within six to eighteen months is a promise with a financing condition attached.

Our take: The bulls have the better of this one, and it is the strongest pillar in the entire story. Advisory fees are the most predictable large revenue line Goldman has, because the pipeline is contracted before it is earned, and a record backlog after a record quarter is about as clean a leading indicator as this industry produces. We model advisory acceleration in the second half. Our disagreement with the bulls is not about the earnings; it is about whether an 8-point-visible earnings driver is worth paying 3.1x book to own, when the market has already seen the same backlog disclosure we have.

Debate: Has the Binding Constraint Moved from Capital to Leverage?

Bull view: The constructive view is that this is a manageable and temporary optimization problem. Goldman manages multiple oscillating constraints by design, CET1 improved to 12.9% with a 150bps buffer, the stress capital buffer is locked at 3.4% through September 2027, the CCAR result was strong, and Basel III finalization plus a recalibrated leverage framework would relieve the pressure entirely. The firm has demonstrated it can shrink RWA $25B in a quarter when it wants to.

Bear view: The bear argument is that the mechanism is structural, not cyclical. Prime financing consumes leverage and not risk weight, so the harder Goldman grows its most durable business, the faster SLR erodes. At 4.3%, peer-low, falling 40bps in one quarter, and with management confirming an unquantified ceiling, the growth engine has a governor on it. Regulatory relief is a hope, not a plan.

Our take: The bears are closer to right and the Street is underweighting this. The elegance of the quarter, CET1 rising while the balance sheet grew $68B, is precisely what creates the problem: this is a business model that converts leverage capacity into revenue, and the leverage capacity is the scarcer resource. The mirror image of "CET1 rebuilt to 12.9%" is "SLR fell to 4.3%," and only one of those made the headlines. We assign no value to regulatory relief in the base case while acknowledging it as the single largest upside option in the story.

Debate: Is 3.1x Book Defensible on a 23.5% ROE?

Bull view: The bull framing is that the multiple simply follows the returns. A 23.5% ROE and 25.5% ROTE justify a premium that Goldman's trading-house history never did, because the mix has shifted toward durable revenue: record management fees, a $4T AUS base, a financing franchise that is 37% of the markets pool, and a backlog that underwrites the next several quarters. On earnings rather than book, 13.6x annualized Q2 EPS is not demanding for a franchise compounding book value and retiring 4% of its shares a year.

Bear view: The bear case is arithmetic. At 3.10x book with a 10% cost of equity and 4% growth, the market is capitalizing a 22.6% sustainable ROE. Goldman just printed 23.5% in the single best quarter of its 157-year history, on a tax rate that will rise, an efficiency ratio management says is not structural, and an equities line that cannot repeat. Paying 3.10x book requires believing the peak is the plateau.

Our take: The bears win, and the arithmetic is the reason this note carries a Hold. Run the justified P/B at a range of sustainable ROEs and the picture is stark: 18% ROE supports 2.33x book, or $858; 20% supports 2.67x, or $980; even a heroic permanent 23.5% supports 3.25x, or $1,195. At $1,140 the market has already migrated to the top of that range. We are not arguing Goldman's ROE collapses. We are arguing that a firm whose CFO just told you the expense base has not structurally changed and whose CEO just told you the cycle will not move in a straight line is not a firm whose peak-quarter ROE should be capitalized in perpetuity. On earnings, 15.2x our revised $75 is not obviously expensive, but $75 is itself a cycle-peak number, and paying a full multiple on peak earnings is how investors lose money in banks specifically.

Model Update Needed

Valuation impact: Our estimate rises materially and our rating falls, which is the correct response to a stock that rose faster than our numbers did. At the $75 midpoint of the revised FY2026 range, GS at $1,140 trades at 15.2x forward earnings and 3.10x book. Applying 15.3x to 16.7x to $75 produces a fair value range of $1,150 to $1,250, a midpoint near $1,200, or roughly 5% above the current price. Add the 1.75% dividend yield and the expected twelve-month total return is approximately 7%, which is at or slightly below what we would expect from the S&P 500. That is the definition of a Hold under our framework.

The valuation discipline worth stating explicitly: $75 is a cycle-peak earnings number, not a mid-cycle one. Applying a full multiple to peak bank earnings is the classic way to lose money in this sector, and the justified-P/B cross-check is unforgiving. A 20% sustainable ROE, which would still be excellent and above anything Goldman managed between 2010 and 2024, supports roughly 2.67x book, or $980. We are not forecasting that outcome, but the gap between $980 and $1,140 is the downside if the AI-driven capital formation cycle takes the pause management explicitly refused to rule out.

Thesis Scorecard Post-Earnings

Scored against the standing thesis carried since our Q1 2026 note, pillar for pillar.

Overall: Thesis strengthened, rating downgraded. These are not in conflict, and the distinction is the whole point of the note. Seven of the eight standing pillars moved in Goldman's favor this quarter. The franchise thesis we argued in April was not just confirmed, it was confirmed faster and larger than we modeled, which is why our FY2026 EPS estimate rises 15% from a $65 midpoint to $75. But a thesis and a rating answer different questions. The thesis asks whether the business is compounding; it is. The rating asks whether the next twelve months of ownership beat the index from today's price; at 3.10x book, capitalizing a peak-cycle ROE that management itself declines to underwrite past six months, that answer is no longer clearly yes.

Action: Take the win and stand aside. We initiated the Outperform in April at roughly $920 on the argument that a FICC miss and a CET1 scare had created a price dislocation in an executing franchise. Both of those bear points have now resolved in our favor, FICC recovered into the exact range we forecast, CET1 rebuilt to 12.9%, and the stock has returned roughly 24% while the S&P returned a fraction of that. That is the trade working. What remains is a superb business at a price that requires the best quarter in its 157-year history to be the new baseline. We would revisit Outperform on either a pullback toward the high $900s, which would restore a genuine margin of safety, or on evidence that the financing franchise can keep compounding without the leverage ratio binding. We would move to Underperform if equities intermediation normalizes toward $3B while the multiple holds above 3x book. Existing holders with a long horizon have no reason to sell a franchise this well positioned in an AI capital formation cycle that is genuinely early. New money has better places to go.