GS · Forward model · Bear case

The Bear case, 20 quarters out

Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.

READ THE FIELD NAMES WITH CARE: this is an equity-level DCF wearing an enterprise-value engine's labels, the same construction proved on the BAC and JPM pages. The page will say 'Enterprise value' and 'Net cash'. What it means is EQUITY value and a deliberate zero. Deposits of $558B - a disclosed figure - and the unsecured long-term borrowings alongside them are FUNDING at a bank, not financing to be netted (the research brief quotes borrowings at both $348B and $438B in different places and carries no sourced figure for them, so no borrowings number is asserted here), so netCash is 0, the engine's 'Equity' IS the equity value and fair value is a true per-share number. Seven substitutions make that reading correct, all checked against the June quarter. (1) The three verticals are the disclosed segment net-revenue lines - Global Banking & Markets, Asset & Wealth Management and Platform Solutions - which sum EXACTLY to reported total net revenues in all fourteen quarters of history, $20,338M in the basis quarter. Nothing is split below a reportable segment: Goldman publishes revenue sub-totals inside GBM but never pre-tax earnings by sub-line, so Banking cannot be separated from Markets and this model does not pretend otherwise. (2) 'EBITDA margin' is PRE-TAX EARNINGS margin, already net of segment operating expenses AND of each segment's own provision for credit losses. It is NOT pre-provision pre-tax income, which would strip the credit cycle out of the model and would lift GBM from 48.32% to 48.80%. (3) capexIntensity is 0.2% of revenue as a premises proxy on the two operating verticals and EXACTLY 0 on Platform Solutions, whose revenue printed MINUS $1,676M in 2025 Q4 and where a positive intensity would produce a negative capex, a phantom cash inflow. (4) corporate.taxRate is 22.6%, the basis quarter's own GAAP effective rate ($1,935M over $8,563M). THERE IS NO FULLY-TAXABLE-EQUIVALENT BASIS AT GOLDMAN and none is imported from the BAC or JPM precedents: the consolidated statement of earnings reports 'Total net revenues' = non-interest revenues plus net interest income, already net of $18,093M of interest expense, with no taxable-equivalent adjustment anywhere in the 8-K, the 10-Q or the 10-K. The rate and the pre-tax figures are therefore already on one basis, and no conversion is needed. The 18.5% in the release is a FIRST-HALF rate and the ~20% on the call is a FULL-YEAR guide depressed by roughly $965M of first-half tax benefits on share-based award settlements; quoting either against a single quarter's pre-tax earnings would overstate net earnings by $351M and $223M respectively. (5) overheadPctRevenue of 1.45% carries PREFERRED STOCK DIVIDENDS and nothing else: $229M grossed up at 22.6% is $295.9M of pre-tax equivalent, 1.4548% of net revenues. Segment operating expenses are already inside each vertical's margin, so a second overhead would double-count them; Goldman even allocates the preferred dividends by segment ($186M / $36M / $7M), which sums to the $229M carried here. Without this line free cash flow would land on net earnings of $6,628M rather than on net earnings applicable to COMMON of $6,399M, and dividing the former by common shares would overstate value by 3.6%. (6) discountRate is a COST OF EQUITY of 10.5%, derived by inverting P/TBV = (ROTE - g) / (CoE - g) on Goldman's own numbers, not reused from another bank. (7) exitEvEbitda of 9.29x is a 12x P/E grossed DOWN by tax, because the engine applies it to terminal PRE-TAX earnings. CALIBRATION, WHICH IS THE TEST OF ALL SEVEN: the first projected quarter's free cash flow is $6,371.5M against actual 2026 Q2 net earnings applicable to common shareholders of $6,399M - a -0.43% error. A static version with no growth and no glide - ($15,520 x 48.32%) + ($4,597 x 24.19%) + ($221 x -21.72%) = $8,563.3M of segment pre-tax, less 1.45% x $20,338M of preferred-dividend overhead, less 0.2% x $20,117M of capex, all times (1 - 0.226) - gives $6,368.6M, -0.48%. The whole residual is the capex proxy, and it is a deliberate conservatism rather than an estimate: unlike BAC, Goldman DISCLOSES capital expenditure, $502M against $509M of depreciation and amortisation in the same quarter, within 1.4% of each other, and depreciation already sits inside the pre-tax margins - so the economically correct incremental capex here is about zero. Setting capexIntensity to 0 moves the calibration to +0.06% and fair value from $1,049.13 to $1,050.96, $1.83 a share. The 0.2% is kept for comparability with the BAC and JPM pages. If a future revision breaks this calibration, the tax or the margin basis has drifted. CREDIT: provisions stay inside the vertical margins, which is where Goldman's own disclosure puts them - the 10-Q publishes provision for credit losses, operating expenses, pre-tax earnings, taxes, preferred dividends and average common equity for each of the three segments, and they sum exactly to the firm figures in all fourteen quarters with no unallocated corporate line and no invented allocation. The firm provision was $102M in the basis quarter: $75M in GBM on wholesale loans, $27M in AWM and NIL in Platform Solutions, whose portfolio moved to held-for-sale. That is a different credit shape from a consumer bank's - lumpy and single-name rather than cyclical - and it is why the smooth margin glide is a mid-cycle average and not a credit cycle. The cycle is carried as a margin step in the BEAR case. BANK ANCHORS, all for 2026 Q2, because a bank valued on a DCF alone is not credible: tangible book value per share $336.61 and book value per share $367.67; ROTE 25.5% for the quarter, 23.4% for the half and 19.96% trailing twelve months on ending tangible common equity, against a 15.0% ROE for the whole of 2025; Standardized CET1 12.9% on $101,657M of CET1 capital and $790,640M of RWAs against an 11.4% requirement (4.5% minimum plus a 3.5% Method 2 G-SIB surcharge plus a 3.4% stress capital buffer), and Advanced 13.6% against 10.5% - so STANDARDIZED IS BINDING at 150bp of headroom, about $11.9B, against 310bp under Advanced; supplementary leverage ratio 4.3% against 3.75%; efficiency ratio 57.4% for the quarter (the release text quotes only the 58.8% first-half figure, and the two are not interchangeable); compensation ratio 30.01% of net revenues for the quarter, where management's quoted 31% is first-half compensation over revenues NET OF PROVISIONS and the same half over gross net revenues is 30.66%; $9.0B repurchased in the first half at an average $949.67 against a $40B authorisation with $23.0B remaining at 30 June 2026, about 5.1 quarters at that pace, so the buyback beyond late 2027 depends on a board authorisation that does not yet exist; and the quarterly dividend raised 11% to $5.00 on 13 July 2026. THE PAYOUT, which was the expected honesty problem and turns out not to be one: Goldman returned $11.74B to common shareholders in the first half - $9.00B of buybacks and $2.74B of dividends - against $11,802M of net earnings applicable to common, a 99.5% payout. The engine's implicit full distribution therefore describes what Goldman actually did. What IS strained is the capital behind it. CET1 capital FELL $2,640M in the six months while Standardized RWAs grew $63,302M (+8.7%), which is how a ~100% payout and a growing balance sheet were financed simultaneously, and the ratio fell from 14.3% to 12.9% as the requirement rose on a G-SIB surcharge stepping from 3.0% to 3.5%. Holding 12.9% while RWAs keep growing at the first-half pace would absorb about $8.2B of CET1 a half-year and leave roughly a 31% payout, not a 99.5% one. So the engine's full distribution is right for the last six months and would overstate distributable cash by about $8.2B a half-year if the financing build continues at that rate. That is stated here in prose rather than plugged into capexIntensity, because programme spend is deducted pre-tax and would corrupt the calibration above. LIMITATION ONE, the labels: as stated, 'Enterprise value' is equity value and 'Net cash' is a chosen zero, not a measured balance. Read them that way or the page will mislead. LIMITATION TWO, the share count: the engine holds shares flat at 298.4M while Goldman retired 4.2% of them in the year to June 2026 (311.5M to 298.4M basic). That is CORRECT arithmetic, not a conservatism - the model discounts total firm earnings and divides by today's count, and buybacks at fair value are NPV-neutral, so shrinking the count would double-count them. What it does misstate is the per-share EARNINGS path: implied terminal net earnings of $28,760M give $96.38 of EPS on today's count and $119.44 on the 240.8M shares the same 4.2% shrink would reach by 2031, so the model's per-share earnings run 19.3% below Street figures built on a shrinking count. The buyback shows up here as price appreciation, not as EPS growth. SEGMENT BASIS: Goldman changed its segment composition commencing 4Q25 - transaction banking moved from Platform Solutions into GBM, some syndication and structured letter-of-credit results moved into FICC financing, Urban Investment Group was spread across all three segments and AWM's Equity and Debt Investments were combined - and recast prior periods back to fiscal 2021 in the 8-K filed 8 January 2026. EVERY quarter here is on the recast basis; mixing vintages would have been material for Platform Solutions, whose 2025 Q2 was reported as $685M at the time and is $619M on the current basis, a 9.6% difference. History starts at 2023 Q1 because the recast publishes conformed QUARTERLY segment results only that far back; 2025 Q4 is disclosed, not derived: Goldman published those quarterly segment net revenues directly in the 15 January 2026 8-K exhibit (GBM $10,411M, AWM $4,719M, Platform Solutions -$1,676M, total $13,454M), so all three lines are carried as reported. Capital ratios, RWAs and AUS in the 8-K are preliminary: the Advanced CET1 ratio was 13.7% in the release and 13.6% in the 10-Q, and the 10-Q figures are used throughout. Fiscal quarters are calendar; 2026 Q2 ended 30 June 2026, so seasonal index 0 is January-March. SEASONALITY is applied to Asset & Wealth Management ONLY, factors [0.9439, 0.9380, 0.9710, 1.1471], so its growthQoQ is a deseasonalised trend rate; Global Banking & Markets passes the spread gate at 2.34:1 but is left ASEASONAL because its Q4 factor is unstable across methods and its Q4 weakness is disappearing rather than repeating, and Platform Solutions cannot be tested because its revenue printed negative. THE LOCAL SERIES MUST NOT BE USED as a cross-check: data/companies/gs/series.json carries total net revenues LESS the provision for credit losses from 2024 Q1 to 2025 Q4 and then switches to gross net revenues for 2026 Q1 and Q2, and its fcf series is misaligned by one quarter before 2026 Q1. The history here comes from the filings. The reference price is the $1,040.87 close captured on 27 August 2026; data/companies/gs/profile.json still carries a stale $1,106.37.

Where each case comes from

From cash flow to fair value

The projected path

Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.

Model revisions

Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.