Goldman Sachs (GS) — Franchise or Beta?

Goldman Sachs earned 22.4% on its tangible equity in the first nine months of 2026, a return beaten only by 2021 in the fifteen years of the series. At $892 the shares trade at 2.6 times tangible book, a price that needs that return, or close to it, forever. On the firm's segment tables and an even split of its costs, this year's excess over its cost of equity comes from market-making and deal fees that turn with the market. Fair value $555 a share, 38% below the last trade; the proponent's own best case $796.

TON618 Capital Research · As of 13 October 2026 · Download PDF · Download model

On Tuesday 13 October Goldman Sachs reported net revenues of $18.7 billion, its second-highest quarter, earnings of $17.01 a share against a consensus of $13.05, and an annualized return on tangible common equity of 20.5%. Advisory fees, equities financing and asset-management fees all set records, and the firm raised its full-year alternatives fundraising expectation to about $150 billion from a plan of $75–100 billion. The shares opened 3% higher and closed slightly down, at $891.91. Asked whether the returns were cyclical or structural, the chief executive said both: "we're in a top quartile environment. And so does that flatter what will be the average return through the cycle? Absolutely. But structurally, the base returns of the firm have been meaningfully uplifted over the last five years." This note is about how much of each.

Terms used throughout, defined once. ROTCE / ROTE: return on tangible common equity, net income to common shareholders over equity less preferred stock, goodwill and intangibles; the return the firm earns on the capital it holds. Goldman reports it on monthly-average equity ("ROTE"); the fifteen-year history here uses year-end averages. TBV: tangible book value, that same capital; $341.43 a share at 30 September. P/TBV: price over tangible book. Cost of equity (CoE): the return shareholders require; built here as the ten-year Treasury plus the stock's beta times an equity risk premium. Excess return: ROTCE minus the cost of equity, earned on tangible book; a firm earning exactly its cost of equity is worth book. Terminal: the level a quantity settles at after the explicit forecast; the fade or glide is its path there. FICC: Fixed Income, Currency and Commodities — the trading business in interest-rate products, credit, mortgages, currencies and commodities; with Equities, it makes up the markets business inside Global Banking & Markets. Intermediation: market-making in FICC and equities — buying and selling for clients, earning spreads and commissions and managing the risk held in between. Financing: lending to markets clients against collateral — repo, structured and mortgage lending in FICC; prime brokerage and portfolio financing in Equities. Durable revenues: Goldman's own term — FICC financing and Equities financing in Global Banking & Markets, Management and other fees and Private banking and lending in Asset & Wealth Management. GBM / AWM / PS: the three segments — Global Banking & Markets, Asset & Wealth Management, Platform Solutions (the Apple Card, being transferred to another issuer). CET1: common equity tier 1 ratio, core capital over risk-weighted assets; Goldman's requirement is 11.4% and management runs 50–100 basis points above it. SLR: supplementary leverage ratio, capital over total exposure without risk weights. G-SIB surcharge: the extra capital a global systemically important bank must hold, 3.5% for Goldman now.

From partnership to bank. Founded in 1869 and a partnership until its 1999 listing, Goldman became a bank holding company on Sunday 21 September 2008, approved together with Morgan Stanley so that both could reach the Federal Reserve's window. It has spent the years since deciding what kind of bank to be. It built a consumer deposit base under the Marcus name, then retreated from consumer lending: the GreenSky and Marcus loan books were sold from 2023, the General Motors card in 2025, and in December 2025 the Apple Card was agreed for transfer to another issuer, a two-year transition the firm says "will substantially complete the narrowing" of that focus. Assets went from $1.1 trillion at the end of 2008, as later recast, to $1.8 trillion at the end of 2025 and $2.2 trillion now. The 1MDB settlements of 2020 left an asset-recovery guarantee to Malaysia that is still in arbitration.

The scorecard. The 2020 Investor Day set the targets the firm measures itself against, and the 2025 shareholder letter keeps the score: earnings per share from $21.03 in 2019 to $51.32 in 2025, book value per share from $218.52 to $357.60, return on equity from 10.0% to 15.0%, total shareholder return of 341% since 2019; durable revenues doubled, historical principal investments cut from about $64 billion to $6 billion, the stress capital buffer down a cumulative 320 basis points. The letter restates the through-the-cycle targets and the intention "in the near term, [to] exceed them — though I know our journey will not be a straight line."

The landscape, as five forces.

• Rivalry. Advice is the one place where share behaves like a moat: #1 in M&A for 23 consecutive years on the firm's count, $1.6 trillion advised in 2025, more than $250 billion ahead of the next adviser, and, by its own account, $5 billion more advisory revenue than the #2 since 2020. In markets the rivals are JPMorgan, Morgan Stanley, Citigroup and Bank of America; on June tangible book, Goldman trades at 2.6 times and Morgan Stanley at 3.5. In wealth, Morgan Stanley is the mirror image, with wealth revenues of $31.8 billion at a 29% margin in 2025 on $10 trillion of client assets, against Goldman's $2.1 trillion in an ultra-high-net-worth market it calls "very, very fragmented" with "no big leader". The listed alternative managers — Blackstone, KKR, Apollo, BlackRock — compete with it in private credit and are its clients in prime financing.
• Buyers. Hedge funds and sponsors on one side, corporates on the other. Equities financing demand in Asia "continues to exceed available industry capacity", in a market the chief financial officer calls "supply-of-financing constrained", so pricing power sits with the firm for now; sponsors, "slow to monetize" their 2020–21 vintages, hold the option to wait.
• Suppliers. Talent, at a 30% compensation ratio and "a very, very strong bid for the very, very best people"; and capital, priced by the Federal Reserve — a requirement that rises from 11.4% to 11.9% in 2028 and possibly 12.4% from 2029, and a leverage ratio that caps how much financing the balance sheet can carry.
• Substitutes. Private credit for bank lending, which Goldman answers by raising it itself (about $150 billion of alternatives this year); the private investment-grade market and the hyperscalers' own cash for public debt in AI financing; electronification, which it absorbed in equities.
• Entrants. The new rails — tokenization, agentic trading, decentralized-finance vaults, prediction markets — which management treats as an expansion of its market rather than a threat (§8).

Where management says it is going. Durable revenues at about 45% of the total as the base. Alternatives fundraising of $75–100 billion a year "on a sustainable basis" (about $150 billion in 2026), $750 billion of alternatives under supervision by 2030 and $300 billion of credit alternatives by the end of 2028. Asset & Wealth Management to a 17–19% return and a 30% pre-tax margin within three to five years of 2025, on fee and private-banking revenue that has compounded 12% a year since 2021, with wealth inflows of 5% a year (12% now). AI-infrastructure financing — "north of $10 trillion of investment through 2032" — as the next cycle, which the firm intends to advise, originate, distribute and fund from public and private pools. Asia, now 20% of revenue from 13% a year ago. One Goldman Sachs 3.0, "a new operating model propelled by AI". Consumer finance out. And the growth it does not promise, as transcribed: "I do not believe we're going to grow our revenues by 11% compounded on the top line for the next five years." The house reads the plan as a firm whose fee and financing lines grow steadily while its largest earnings line remains the market's; §3 to §6 price it on that reading.

Nine things the quarter said.

1. Equities financing (a record, +96%) and advisory (a record, +18%) carried a 30% earnings beat. FICC intermediation fell 7% and the backlog ended the quarter lower than it began, though "our backlog remains very robust" and "there's an enormous amount of dialogue that continues".
2. Financing revenue grew about 3% sequentially (equities financing +96% year on year), which management put down to clients de-levering in the quarter.
3. The $368 million provision reflected portfolio growth and impairments on "a couple of different names", which the chief financial officer called "highly idiosyncratic, the non-correlated set of names … we don't think it's necessarily the signal of something else to come". Every loan under the allowance is now wholesale (the Apple Card book is held for sale), and 85% of gross loans are secured.
4. The compensation ratio is 30% and management intends to hold it — "our best estimate at the moment for what we expect for the year" — with no "diminution in terms of the market for talent. Frankly, quite the opposite."
5. CET1 fell to 12.4%, the top of the 50–100 basis-point buffer, as risk-weighted assets grew $26 billion; the buyback slowed for a second quarter ($5.0, $4.0, $3.0 billion) while the dividend rose 25%.
6. Alternatives fundraising guidance rose from $75–100 billion to about $150 billion for the year: "there's nothing opportunistic about it".
7. Incentive fees of $587 million came from clients' "early access to a high-growth investment opportunity", recognized "following the company's IPO", and are expected to "remain elevated in the fourth quarter". A $450 million charitable contribution cost about $1 a share, with "a meaningful contribution in the fourth quarter" to follow — together at least five years of giving pulled forward.
8. The chief executive, on succession, twice: "there is no succession, transition or timeline underway".
9. On rates: "a 5% 10-year treasury doesn't scare me … What creates headwinds for markets is when you have very, very abrupt moves." The environment has gone from "top decile" in May and June to "a top quartile environment" now — "that's still really, really good".

Where management puts the risk, in its own words. The shareholder letter: "we know from history that nothing is linear over time … it is not hard to come up with scenarios where risks become a lot more pronounced. In recent weeks, for example, concerns about private credit, including underwriting quality or exposure to software companies that may be adversely affected by AI, are a reminder that the credit cycle has not been repealed. Higher levels of market volatility across various risk assets, elevated geopolitical uncertainty, and greater capital deployment, especially into AI, require diligent risk management." The 10-K's risk factors run from market conditions and asset values to liquidity, credit and concentration, then operations, including AI and cyber. The call put geopolitics first: the Middle East, energy flows, inflation, bond yields, and a Federal Reserve that hiked for the first time in three years, with markets "pricing in approximately 75 basis points of additional tightening in this cycle across the U.S., U.K., and Europe". On AI financing, the cycle the firm puts first, the chief executive gave the most hedged answer of the morning — "I don't know and nobody knows … There is risk to pressure on … crowding out" — then affirmed the demand: "there's a lot of demand for compute at the moment. The supply is not coming to balance. More is going to get built, more financing is going to be necessary. There will be a point in time when this recalibrates, resets for sure."

What went unraised. Neither management nor the analysts took up the securities book and its marks, deposit betas, lending to nonbank financial institutions (the "other collateralized" category is $113 billion of the $276 billion of loans), the timing of the Apple Card transfer, or the run-down of the historical principal investments ($6.0 billion at year-end, from $64 billion).

The insiders. Officers and directors sold $149 million of stock in the open market over the last twelve months and bought none, mostly between January and May, at about $920–1,150 a share and an average near $950 — roughly 2.8 times tangible book, inside the range this note puts above fair value. The chief executive kept 137,000 shares after selling $3.5 million; the largest seller, a director and former chief financial officer, kept about 550,000 directly (more in trusts) after selling $66 million; four of the eight officers who sold kept fewer shares directly than they sold (Form 4 holdings exclude unvested awards). It is stated here as a fact rather than a signal.

The identity. The quarter showed what Goldman earns now; the price is a statement about what it earns forever. Price to tangible book is a statement about return: P/TBV = (ROTCE − g) / (CoE − g). At $891.91 and $341.43 of tangible book a share the multiple is 2.61×. With growth at 3% and an 11.0% cost of equity (built below), that is a perpetual return of 23.9%; at 10% it is 21.3%, at 12% 26.5%. Let the return start at 19.5%, below this year's 22.4%, and glide down (the start is explained below), and the resting level the price requires is 24.9% (21.3% / 28.7% on the grid).

The record. Fifteen fiscal years: 3.9, 11.3, 11.6, 11.7, 7.9, 9.9, 5.3, 14.0, 10.6, 11.6, 24.3, 11.2, 8.0, 13.5, 15.9 percent (net income to common over average tangible common equity, from the SEC filings), then 22.4% for the first nine months of 2026 on the company's basis. Mean 11.4%, median 11.3%; 9.8% in the decade before the Investor Day, 14.6% in the five completed years since (15.9% with the 2026 run-rate). The 10-K target is "through-the-cycle … ROE within a range of 14% to 16% and ROTE within a range of 15% to 17%". The price needs about seven points more than the top of that range.

The cost of equity, in the open. Ten-year Treasury 5.22% (8 October); beta 1.21 on five years of weekly returns against the S&P 500 (262 weeks), 1.14 Blume-adjusted, 1.41 over the last two years; an equity risk premium of 4.8%, the house assumption for financials, shared with the JPMorgan valuation published alongside this one. CoE = 5.22% + 1.21 × 4.8% = 11.0%; 10.7% on the Blume beta, 12.0% on the two-year. Two cross-checks: the line the market draws through twenty-one large banks (§5) values a bank at book when it earns 10.3%, the market's own cost of equity for the group (8.2% on the line through the banks alone); and the FDIC's all-bank return on equity has run 11–13% for five years. JPMorgan's five-year beta on the same method is 1.0, a 10.0% cost of equity, and its valuation uses 10.5%, rounded up from 10.3% on an adjusted two-year beta. Goldman's raw beta runs 0.2–0.3 above JPMorgan's in both windows, worth a point or more; the house takes half of that, because the two-year window that widens the gap most is the one most coloured by 2026. Every result is shown on a 10 / 11 / 12% grid.

The fade. The model values tangible book plus the present value of the excess return — (ROTCE − CoE) × tangible book — each year for twelve years, discounted mid-year, plus a terminal value on the excess that remains. Book grows at the terminal growth rate each year, so the payout is whatever earnings exceed that growth: about 84% next year and 79% at the terminal in the base case. The house sets the start at 19.5%, below this year's 22.4%, because the base case has intermediation and advisory normalizing from a top-quartile 2026 and the compensation ratio settling at 31–32% (the base story in §5); from there the return glides to a resting level on a half-life: 19.5% to 14.5% on a 4.5-year half-life in the base case (18.8% in 2027, 18.2% in 2028, 17.6% in 2029), 18.5% to 11.5% in the bear, 20.5% to 16.5% in the bull. The resting level is where the franchise question is answered, and §4 is how the house set it.

Management's definition. "More durable revenues" are, by the shareholder letter's footnote, FICC financing and Equities financing in Global Banking & Markets and Management and other fees and Private banking and lending in Asset & Wealth Management. On the call they have "more than doubled since 2020 Investor Day and now representing approximately 45% of revenues for the year to date". On the segment tables: $24.9 billion of $56.2 billion, 44.3%; 48.1% in the third quarter; 45.2% in 2025; 43.1% in 2023.

Stability. The durable lines rose from $6.1 billion to $9.0 billion over the eight quarters, higher every quarter, with a quarter-to-quarter standard deviation of 4% and a peak-to-trough ratio of 1.48. The beta lines ran from $5.8 billion to $10.9 billion and back to $8.6 billion: standard deviation 20%, peak to trough 1.90, and 2.03 for intermediation alone. Financing on its own is 1.82 peak to trough, a balance-sheet business whose balances are the clients' leverage. The chief executive: "when equity markets come down, there's a deleveraging … there's going to be some cyclicality to that." The durable lines are durable as a trend, not as a level.

What they contribute. Goldman publishes no margin or equity by line, so the house spreads the firm's nine-month costs evenly across revenue: 60.8% of revenue in compensation, non-compensation and provisions, a 20.4% tax rate, preferred dividends pro rata, on $100.5 billion of average tangible equity. On that basis the durable lines contribute 9.9 points of the firm's 22.4% return, intermediation and investment-banking fees 11.2, the remaining lines 1.3. Put plainly: without the beta lines and their costs, the durable lines would earn 9.9% on the equity the firm holds today, a point below its cost.

Two caveats cut the proponent's way. A stand-alone fee-and-financing business would need less equity than the whole firm, so its own return would be higher. The firm does not disclose the split, but Asset & Wealth Management, mostly durable lines, earned 14.4% on its attributed equity in the nine months, above the cost of equity, while financing's equity sits undisclosed inside Global Banking & Markets' $81.5 billion. And the even split understates financing's margin, since it uses balance sheet rather than headcount. Run the other way — the fee lines at Asset & Wealth Management's disclosed 24% pre-tax margin and financing at Global Banking & Markets' 45% — the durable lines contribute 8.8 points; at a 50% margin on financing, 9.5; at 60%, 10.8; at 70%, 12.2; at 80%, 13.5. Financing would need about a 61% pre-tax margin, against the 45% its segment reports in a record year, for the durable lines to reach the 11% cost of equity, and about 84% to clear it by the three points §9 asks for.

The conclusion is narrower than "the franchise earns its cost of equity and no more": The excess return over the cost of equity in 2026 is contributed by the beta lines in a top-quartile year; the durable lines' contribution, on the firm's whole equity base, is about the cost of equity. The segment returns say the same from the other side. Global Banking & Markets, which holds $81.5 billion of the $109.4 billion of common equity, earned 11.3% in 2023, 13.8% in 2024, 16.4% in 2025 and 23.5% for the nine months of 2026 — a return that tracks the cycle. Asset & Wealth Management earned 4.5%, 14.6%, 12.5% and 14.4% on $24.4 billion — above the cost of equity in two of the last three years and below its own 17–19% target.

Why the base case rests at 14.5%. The five completed years since the Investor Day average 14.6% (15.9% with the 2026 run-rate), the fifteen-year mean is 11.4%, the band 15–17%. A resting return of 14.5% credits the uplift — three points above the long mean and above the durable lines' contribution — and sits half a point below the band's floor for the cycle that management itself says flatters the present. The bull case's 16.5% is the band hit at the top in perpetuity; the bear's 11.5% is the fifteen-year mean rounded up, the franchise covering its cost base and no more. The proponent's 19% says the uplift is the whole of the last five years and permanent; it is run below, unweighted.

The stories. Bear: intermediation and investment-banking fees return to their 2023–24 run-rate, financing stalls on a deleveraging, the G-SIB surcharge steps to 4.5% and the buyback slows; the return lands on the fifteen-year mean. Base: the durable lines compound in the high single digits, intermediation and advisory normalize from top-quartile to mid-cycle, the compensation ratio settles at 31–32% and CET1 sits inside the buffer; the return rests half a point below the band. Bull: the Investor Day re-rating holds — AWM reaches its 17–19% return and 30% margin, financing stays above 40% of markets revenue with Asia pricing, the Basel and G-SIB rules land as proposed, and the count shrinks 3% a year; the return rests at the top of the band. The proponent's case takes the chief executive's "structurally … meaningfully uplifted" at its strongest, about 19% in perpetuity at the Blume beta, and reaches $796. Weighted fair value $555, 38% below the last trade; the weights are a judgment, stated.

Three choices bind every case. Terminal growth is capped at 4%, because a bank's book cannot outgrow nominal GDP in perpetuity, and no case uses more than 3.5%. The share count is held at 295.6 million, because buybacks are one form of the payout the model already values: a buyback above fair value transfers value from the remaining holders to the sellers rather than creating it, so a shrinking count at today's price is not a source of per-share value the model omits. And the half-lives shorten from bull to bear (5.5, 4.5, 3.5 years), because a return that is mostly cycle reverts faster than one that is mostly franchise, which is what each case asserts.

Sensitivities. Each point of resting return is worth about $32 a share ($523 / $587 at ±1 point); each point of cost of equity about $65 ($629 at 10%, $497 at 12%). Among the operating inputs, revenue in Global Banking & Markets — 74% of revenue, 74% of attributed equity, 85% of nine-month pre-tax earnings — is where value moves: a 10% permanent shortfall is about 3.1 points of return through the fade, roughly $100 a share. A point on the compensation ratio is about 0.6 points of return, $20; a CET1 run level 50 basis points higher defers $4 billion of buyback, about $14. Rates barely register: the firm's own table puts ±100 basis points at +$56 million and −$226 million of twelve-month net revenue, so a 25 basis-point cut is $57 million.

The line. Part 1 of the bank thematic found that the market prices the large banks on one line, price to tangible book against return on tangible equity. Re-priced to 13 October across the twenty-one banks the JPMorgan valuation uses — Part 1's twelve less BlackRock, plus ten large regionals and custodians — it is P/TBV = −0.73 + 0.167 × ROTCE (R² 0.88), with a bank worth book at a 10.3% return. Through the nineteen without Schwab and BNY, whose tangible books are a tenth of their balance sheets, it is 0.02 + 0.119 × ROTCE (R² 0.53), book at 8.2%.

At this quarter's 20.5% the line pays $920 a share, 3% above the price: a residual of −0.08× of tangible book, $28 a share. At the nine-month 22.4% it pays $1,028, a residual of −0.40× (−0.49× on the second-quarter basis; −0.07× and −0.13× on the bank-only line). At the base case's resting 14.5% it pays $578, and at the top of management's band, 17%, $720 (on the bank-only line, $840, $917, $597 and $698). Against the nine-month return the market applies the markets-house discount it has applied for a decade, about $135–165 a share on the full line and almost nothing on the bank-only line; against the quarter's return the discount is $28. Two claims, kept apart: the relative one — Goldman is a little cheap to its peers — is true and small; the absolute one — the whole group is priced on one year's returns — is where the house disagrees, and it carries the number.

Goldman's segments are different businesses, so each is priced on its own terms and the sum is reconciled to the whole. Global Banking & Markets is valued on excess return on its $81.5 billion of attributed equity at a 12.0% cost of equity — a point above the firm's, the rate on the two-year beta of 1.41 — with its return fading from about 20% to 12.5 / 15.0 / 17.0% (a point above its 2023 level of 11.3%, a point and a half below its 2025 level of 16.4%, and above it): $100 / $124 / $146 billion, 1.23 / 1.52 / 1.79 times its equity.

Asset & Wealth Management is valued on fee earnings — management fees at the third-quarter run-rate and private banking at the nine-month run-rate, plus $0.6 billion of normalized incentive fees, at the segment's 24% nine-month pre-tax margin (about 27% in the bull case, halfway to the 30% target) and the firm's tax rate — times 13 / 16 / 19, a band from the traditional asset managers to BlackRock's 25, plus the $2.8 billion of attributed equity behind the historical principal investments ($6.0 billion of investments at year-end): $48 / $58 / $76 billion. Each multiple carries an implied perpetual growth rate, at the firm's cost of equity and an 80% payout, of 4.6 / 5.7 / 6.5%, above the 4% ceiling the consolidated model obeys; at the 10% cost of equity a proponent would apply to a fee stream it is 3.6 / 4.8 / 5.6%, and the bear multiple fits under the ceiling. The segment carries the firm's cost of equity rather than Global Banking & Markets' extra point because its management fees are the least volatile line the firm has (Exhibit 2). Platform Solutions is carried at its $3.5 billion of attributed equity: the Apple Card is held for sale and the transfer completes in about two years; the segment lost $95 million before tax in the quarter.

The reconciliation. Weighted, the parts sum to $187 billion ($632 a share) against the consolidated $164 billion ($555): a spread of $23 billion, 12% of the parts and 14% of the whole, with three named sources. The largest is the AWM multiple: an asset-manager price-to-earnings embeds 4–6% perpetual growth, which the consolidated model caps at 4%. Valued instead on excess return on its own equity (its return fading from 16% to 13 / 16 / 18%), AWM is worth $34 / $40 / $45 billion and the parts sum to $137 / $167 / $194 billion — within 1% of the consolidated value in the base and bull cases and 6% above it in the bear. The whole spread, in other words, is the asset-manager multiple on a segment that is 15% of nine-month pre-tax earnings. The second source is the base: the segments start from attributed common equity, which includes $9.2 billion of goodwill and intangibles, while the consolidated model starts from tangible equity; at equal earnings the two converge over time. The third is Platform Solutions at book against its negative return inside the consolidated fade, worth about a billion. The spread to the market is in one place. Market capitalization of $264 billion, less AWM at 16 times ($58 billion) and Platform Solutions ($3.5 billion), leaves $202 billion for Global Banking & Markets — 2.5 times its equity. At a 12% cost of equity and 3% growth that is a perpetual return of about 25% on that equity, against 16.4% in 2025, 13.8% in 2024, 11.3% in 2023 and 23.5% in the top-quartile first nine months of 2026.

The map. Common equity tier 1 of $101.4 billion on $817 billion of standardized risk-weighted assets is 12.4% (12.9% at June, 14.3% at the end of 2025). The requirement is 11.4%: the 4.5% minimum, a 3.4% stress capital buffer the Federal Reserve froze through September 2027, and a 3.5% G-SIB surcharge. The surcharge steps to 4.0% in 2028, taking the requirement to 11.9%, and the firm says it is "within the 4.5% G-SIB surcharge range" on first-half data, which would take it to 12.4% from 2029 at the earliest, pending the Fed's recalibration proposal; management expects the Basel and G-SIB proposals together, if adopted as proposed, would not "materially change" its requirements. The supplementary leverage ratio is 4.3% against a 3.75% requirement, and the chief financial officer says "the largest … balance sheet extension is behind us." Global core liquid assets are $574 billion, 26% of assets. At the insured bank, GS Bank USA, CET1 is 13.5% and the securities marks were under 1% of equity at June, the smallest of the markets-house and money-center banks Part 1 measured; the holding company's September marks arrive with the 10-Q in November, so the marks here are second-quarter and bank-level.

What the buyback is doing. Capital returned to common shareholders was $6.4 billion in the first quarter, $5.4 billion in the second and $4.5 billion in the third — $16.2 billion against $16.9 billion of net income to common, a 96% payout — with the buyback itself slowing from about $5.0 billion to $4.0 billion to $3.0 billion as risk-weighted assets grew $90 billion in nine months, 12%. In the third quarter, retained earnings after the dividend were $3.6 billion against a $3.0 billion buyback, so capital was flat ($101.4 billion from $101.7 billion); the ratio fell about 45 basis points because risk-weighted assets grew $26 billion, worth about 40 of them. The dividend is $5.00 a share, up 25% on the year and from $2.00 five years ago.

What the model needs. The base case pays out about $16 billion in 2027 and 79% of earnings at the terminal, with book growing 3% a year — less than the firm is paying now. That fits management's order of use, "Clients first. Then when we can't deploy back to shareholders", within the buffer only if risk-weighted assets grow at the pace of retained capital rather than the 17% annualized of 2026's first nine months. A valuation that assumed this year's payout and this year's balance-sheet growth together would be inconsistent; this one assumes neither.

Deposits: underpaid, and does it matter here? The sector argument is real, and the house makes it in its JPMorgan valuation: $2.08 trillion of interest-bearing deposits paid 2.18% against a 4.23% three-month bill, with $655 billion paying nothing, a gap worth about $70 billion a year to that one bank. It does not transfer to Goldman. Its $568 billion of deposits are consumer savings ($230 billion of Marcus and Apple balances, priced to the market and the first to reprice — private banking and lending revenue rose 20% in the quarter "driven by higher net interest from consumer deposits"), private-bank balances ($105 billion), transaction-banking operating balances ($93 billion), brokered certificates ($52 billion), sweep balances ($42 billion) and other ($46 billion). Substantially all of it is interest-bearing, and the insured bank's all-in cost of funds was 3.59% at June against a 3.50–3.75% funds target. Goldman's deposit risk is a funding cost that rises with the funds rate, not flight: the money is already paid the market rate. The run question is the other kind — 43% of the bank's deposits are uninsured — and Part 1's run test put Goldman's cover at 1.18 times those deposits from cash, reverse repo and available-for-sale securities, third-highest of eleven. The capital-flight argument belongs to banks that pay 0.07% on checking; it is not a lever in this valuation.

What the firm has built. Asked on the call whether tokenization, agentic trading, decentralized-finance vaults and prediction markets "nip at the core" of the markets business, the chief executive laid out the strategy. "We were publicly supportive of clarity" — on the house's reading, the market-structure bill that failed cloture in September — because it "would be good to give the market more clarity … and ensure investor protection", and "in the absence of that, we're already seeing agencies take steps to provide clearer rules". The firm has "been investing in companies and the digital asset, broader market infrastructure space for years", built "our own rails for digital assets" (the GS DAP platform), "tokenized bonds", "a number of money market funds that are tokenized" and "real estate for certain clients", and "participated in the launch of DTCC's tokenization service" in July, where it is "actively assessing … scalable use cases". The gain it expects is operational: "greater efficiency, automation, streamline processes, reduce costs, take away friction, improve liquidity across asset classes." Its own sizing, twice: "very small at this point" and "small and at the margin". Goldman is also one of the twenty-one institutions that announced a jointly issued dollar stablecoin for the first half of 2027, and it has announced no coin of its own.

What it means for the business model. Management's frame is the electronification of equities, a change "people were concerned about" that left "the opportunity set for us … greater than what anyone could have contemplated": new products, "greater activity, potentially new clients, broader distribution channels … expand the TAM for our business." The house's reading follows from the deposits in §7. The stablecoin threat to a bank is the loss of underpaid deposits; Goldman has none to lose, so the funding leg of the stablecoin case Part 1 tested does not apply. What tokenized collateral and always-on settlement would change is the plumbing of the firm's largest growth engine — financing, $12.7 billion of revenue this year, $9.2 billion of it equities financing, where prime and portfolio financing sit, on balances that settle and re-margin daily — toward lower cost and higher velocity. That cuts both ways: it is where the financing franchise could widen, and where a new market structure could price it down. The model takes no view on either. The observables are the Treasury rule on affiliate yield, the consortium coin's terms in 2027, and the first time the firm discloses a tokenized-financing revenue line.

What the house will score, and when. Seven dated tests, each with a rule for silence. This is a valuation, not an event pair; Part 2 of the bank thematic scores Part 1's tripwires on this print.

1. The first year of the fade: fiscal 2027 return on tangible equity, company basis, from the January 2028 release — at or above 20% scores the bull path, at or below 17.2% the bear, between them the base, each read with the cycle control above and provisional until fiscal 2028; always printed.
2. The durable share: fiscal 2026 durable revenues at or above 45% of net revenues and fiscal 2027 durable growth of 8% or more, both together, fire for the franchise; a share below 42% fires for beta; the lines are always printed.
3. The market-beta segment: Global Banking & Markets' fiscal 2027 return on attributed equity at or above 17% (bull) or at or below 12.5% (bear); always printed.
4. The buffer: a quarter-end standardized CET1 below 11.9% while repurchases exceed $2 billion in the quarter; both print every quarter, so there is no silent branch.
5. The requirement's path: the G-SIB recalibration finalized by 31 December 2026 with Goldman's 2028 surcharge at or below 3.5% reverses the path; above 3.5% is as modeled; not finalized is scored "slipped" and re-read at the 10-K.
6. The asset-manager multiple: Asset & Wealth Management's fiscal 2027 pre-tax margin and return (27% and 15% raise the band; 24% or below lowers it); always printed.
7. The exit: the Apple Card transfer closed by 31 December 2027 with Platform Solutions' attributed equity released below $1 billion; if not, it is scored "slipped" and the segment stays at book with its losses.

The dates. Part 2 of the bank thematic, which follows this note, reads the third-quarter prints of the twelve, Morgan Stanley's among them. Goldman's third-quarter 10-Q in early November carries the marks, Level-3 assets and the segment equity detail; the 13F deadline is 16 November; the FDIC's third-quarter profile and the Fed's Financial Stability Report arrive in late November. The dividend is paid 30 December. Management is "hopeful" of the Basel and G-SIB finalizations by year-end. Fourth-quarter and full-year results are due 19 January 2027, confirmed by the company. The current stress capital buffer runs to 30 September 2027 and a new one follows the 2027 stress test; the G-SIB surcharge steps on 1 January 2028; the Apple Card transfer completes around the end of 2027.

What the note cannot settle. Whether 2026 is the top quartile or the new normal, which only the next several years' returns can show; the durable lines' true margin, which the firm does not publish; and whether the market's cost of equity for the group is the 11.0% built here or nearer the 9.9% at which the proponent's case reaches the price.