At today's price, Goldman Sachs's stock is priced as if it can sustain a return on equity of about 22.9% forever — 1.7 times its 5-year average of 13.7%. The catch: the most recent quarter's ROE was 23.5%, essentially matching what the price already requires, so the real question is whether that's a new normal or a temporary peak.
Why this isn't a standard free-cash-flow model
At its current price, GS is priced as if it will sustain a ~22.9% return on equity forever.
Actual ROE over the past 5 years ranged from 7.5% to 23.0%, averaging 13.7% — though the most recent quarter (Q2 2026) came in at 23.5%, almost exactly matching what the price requires.
Required ROE vs. historical ROE
| Required ROE (price-implied) | 22.9% |
| 5-year average | 13.7% |
| 5-year low (2023) | 7.5% |
| 5-year high (2021) | 23.0% |
| Q2 2026 (annualized) | 23.5% |
Sensitivity: what if the cost of equity moves?
| COE | Required ROE |
|---|---|
| 8% | 17.5% |
| 9% | 20.2% |
| 10% (base case) | 22.9% |
| 11% | 25.7% |
| 12% | 28.4% |
A volatile securities firm like Goldman is often assigned a somewhat higher cost of equity (10-11%) than a stable large-cap bank (8-9%) — and that single choice swings the required ROE from 17.5% to 25.7%.
What would move this number
- Lowering the long-term growth rate (g) from 2.5% to 1.5% actually raises the required ROE, from 22.9% to 24.7% (a smaller-denominator effect).
- Using year-end-2025 book value per share ($357.60) instead of the latest quarter's ($367.67) raises the P/B multiple to 2.80x, nudging the required ROE up to 23.4%.
- Switching to total shareholder returns (dividends + buybacks) as an FCF proxy and solving with the standard growth-rate reverse-DCF model gives a completely different picture: at COE 10%, the required growth rate is just 5.8% — well below the actual 3-year growth rate in shareholder returns (35.8% a year). In other words, "how much better does it need to get" (the ROE lens) looks demanding, while "how much more does it need to return to shareholders" (the cash-flow lens) looks modest — the fact that the two lenses disagree is itself a useful signal.
Show your work
Inputs, sources, model assumptions, and the calculation
- Share price$1,001.95 — TradingEconomics, Aug 20, 2026 close
- Shares outstanding291,171,408 — 10-Q Q2 FY2026 cover page, Jul 17, 2026
- Book value per share (BVPS)$367.67 — 10-Q Q2 FY2026 balance sheet, Jun 30, 2026
- Cost of equity (COE)10% base case (8-12% tested)
- Long-term growth rate (g)2.5% — roughly long-run real GDP growth
Why not standard FCF: the 10-K FY2025 consolidated cash flow statement shows net income +$17,176M vs. operating cash flow -$45,154M (2024: -$13,212M; 2023: -$12,587M). Balance-sheet items unrelated to underlying earnings power — trading assets (-$69,866M), trading liabilities (+$57,560M), collateralized transactions (-$12,891M) — dominate the reported cash flow. FCF is structurally negative and distorted here, so a bank-valuation alternative was used instead.
Formula: P/B = (ROE − g) / (COE − g) → ROE = g + P/B × (COE − g). P/B = $1,001.95 / $367.67 = 2.725x. With g=2.5%, COE=10%: required ROE = 2.5% + 2.725×(10%−2.5%) = 22.9%. This reformulates a dividend-discount model in terms of return on equity — a standard approach in bank and broker-dealer valuation.
Cross-check: Goldman returned an average of 87.5% of net income to shareholders (dividends + buybacks) over the past three years (60% in 2022, 110% in 2023, 83% in 2024, 98% in 2025). Using that total shareholder-return figure as an FCF proxy in the standard 10-year-plus-terminal two-stage model and solving by bisection: with CF₀ = 2025's total return ($16.78B) and market cap $291.7B (2026-07-17 basis), the implied g ≈ 5.8% at COE 10%. Using the trailing-3-year average CF₀ ($12.66B) instead shifts that to g≈9.6% — buybacks are discretionary and swing more year to year (2025's figure ran +32.6% above the 3-year average), so this alternate metric carries more noise than the ROE-based model.
Historical benchmarks: ROE 23.0% (2021), 10.2% (2022), 7.5% (2023), 12.7% (2024), 15.0% (2025); Q2 FY2026 annualized ROE 23.5%; 3-year net income CAGR (2022→2025, off the cyclical low) +15.1%; 4-year BVPS CAGR (2021→2025) +5.9%.
The fine print
- This isn't a fair-value price target — it only shows what the current price already assumes.
- Change the cost of equity, the growth rate, or the FCF proxy used for the cross-check, and the result moves substantially — see the sensitivity table and "what would move it" section above.
- Banks and broker-dealers don't fit the standard FCF reverse DCF, so a ROE-based model was substituted here — that substitution is itself a judgment call, and other analysts might choose differently.
- Whether the required ROE is realistic depends on whether the current run of record results — and its dependence on the AI infrastructure investment cycle, covered in the story piece above — is durable, not on this math alone.
- Any investment decision, and its outcome, is your own responsibility.
Frequently asked questions
Why does this use return on equity instead of free cash flow?
Goldman Sachs's balance sheet makes a standard free-cash-flow reverse DCF meaningless, since it's a securities firm, not an operating company — so this analysis uses the bank-valuation model investors actually rely on: reverse-engineering the return on equity the stock price already assumes.
What ROE does GS's stock price assume?
At today's price, Goldman Sachs's stock is priced as if it can sustain a return on equity of about 22.9% forever — 1.7 times its 5-year average of 13.7%.
Is Goldman Sachs's current ROE close to what the price requires?
Yes, unusually so — the most recent quarter's ROE was 23.5%, essentially matching the ~22.9% the price requires, so the real question is whether that's a new normal or a temporary peak.