$420.75
$889.82
bear $420.75
base $685.07
bull $889.82
margin of safety $320.00
market price* $881.26
Goldman Sachs makes money in two ways. Global Banking & Markets advises on mergers, underwrites stock and bond issues, and makes markets in equities, bonds, currencies and commodities for institutions; this is the larger and far more cyclical half. Asset & Wealth Management runs money for institutions and wealthy families and earns recurring management fees that have grown steadily, from $9.5 billion in 2023 to $11.5 billion in 2025, alongside lumpier gains from the firm's own private investments. The consumer experiment is ending: the Marcus loan book and the GM card were sold earlier, and in December 2025 Goldman agreed to hand the Apple Card program to another issuer over roughly 24 months, which cost about $2.3 billion of markdowns and contract termination charges in 2025, offset by a $2.5 billion release of credit reserves. In January 2026 it closed the purchase of Industry Ventures, a venture capital fund manager, for $665 million plus up to $300 million contingent on results through 2030. For a bank and dealer, deposits and borrowings are the raw material of the business, so we measure owner earnings as net income left for common shareholders after preferred dividends, value it directly and subtract no debt. Operating cash flow is not usable here because it swings by tens of billions with trading inventory. On that measure, per diluted share, earnings were $16.29 in 2016, $9.01 in 2017 (cut by a one-time tax-law charge), $25.27 in 2018, $21.03 in 2019, $24.74 in 2020, $59.45 in the 2021 boom, $30.06 in 2022, $22.87 in the 2023 trough, $40.54 in 2024 and $51.32 in 2025. Because the share count has fallen from 435 million to about 305 million, we value total dollar earnings and divide by today's share count rather than counting the buybacks twice. In dollars, common earnings over 2018 to 2025 ranged from $7.9 billion to $21.2 billion with a median of about $10.3 billion, while the business is now carrying far more capital: common equity is about $110 billion at mid-2026 (roughly $360 a share of book value and about $329 of tangible book). We therefore anchor on return on common equity across a full cycle rather than raw dollars: 2016 to 2025 returns ran from about 4% (2017) and 7.5% (2023) to 23% (2021), with a median near 11%. Our normal earning power is 11% on today's $110 billion, about $12.1 billion a year. The first half of 2026 was very strong: $11.8 billion of common earnings, up from $8.1 billion a year earlier, after 2025 earnings grew about 20%. In the base case the second half is softer and 2026 lands near $18 billion, a 10% gain, below last year's growth; 2027 fades to $14.5 billion as deal and trading activity normalizes, and from 2028 the firm earns its normal $12.1 billion indefinitely. We credit no further growth: the fee business has grown, but adding assets depends on markets and share gains, so it is left out, and the private-investment book is being run down. The perpetuity is supported by named structural features: a top-two position in global M&A advice and equity underwriting for decades, a prime brokerage and market-making franchise with institutional clients, and over $3 trillion of client assets under supervision. That gives about $685 a share. The bull case assumes the firm sustains what it earned in 2024 to 2026 rather than its long-run median: 2026 at $20.5 billion, 2027 at $18 billion, then 14.5% on equity, about $15.9 billion a year, consistent with the 15% return of 2025 and management's demonstrated ability to run with a larger fee base. That gives about $890 a share. The bear case puts a full capital-markets and credit downturn into the story, sized from Goldman's own record. In 2008 common earnings fell to $2.0 billion and the return on equity to under 5%; from 2011 to 2016 returns stayed in a 4% to 11% band as trading volumes and deal fees stayed depressed. We assume 2026 still closes at $10 billion (the first half is already banked), then 2027 falls to $1.5 billion with large trading losses, writedowns on remaining principal investments and a rise in credit provisions on corporate and real-estate loans, recovering to $4 billion in 2028 and $7 billion in 2029, after which the firm earns 7.5% on equity, about $8.2 billion a year, the weak end of its post-crisis decade. The firm stayed profitable even in 2008 and today runs a common equity tier 1 ratio of 14.3% against a 10.9% requirement, so we do not model forced dilution; buybacks simply stop. That gives about $421 a share. This is above tangible book of about $329 because, on the firm's own history, even the weakest multi-year stretch produced positive single-digit returns on equity, and the recurring management fees of $11.5 billion a year do not depend on deal activity. What could make us wrong: a loss far deeper than 2008 from a trading or counterparty failure, regulatory changes that raise capital requirements again, legal penalties of the size of the 1MDB settlement in 2020, or the loss of talent that carries the franchise. The firm's dependence on wholesale funding and on confidence is the risk no case models: in September 2008 it needed a $10 billion government capital injection and a $5 billion investment from Berkshire Hathaway to stay funded, and a repeat could dilute shareholders heavily. Margin of safety price: we are reasonably confident in an 11% normal return because it is the median of ten years that include both a boom and a trough, but the year-to-year range is enormous and the firm's balance sheet is around 12 times its common equity, so the worst plausible outcome is a funding crisis rather than a bad year. If we are wrong about the normal return (9% instead of 11%) the shares are worth about $580; if a 2008-style crisis forces a 10% share issuance at trough prices, the bear value falls to about $380. The price assumes a four-to-five-year holding period, with a dividend of $18 a share a year, about 5.6% of the margin-of-safety price. At $320, slightly below tangible book value, a return to the base case within four years implies better than 20% a year, and even the bear case implies value roughly a third above that price within a year or two of the trough. The margin-of-safety price is $320. Previously we valued the shares on a different basis, and those figures are not comparable with these.