The brand and scale are real, but investment-bank earnings swing with M&A, IPO, and FICC cycles in ways Buffett himself called 'far less predictable.' At 1.45x the 10-year average P/E, the price already prices in the good cycle.

Plain English

Goldman Sachs is the most prestigious investment bank in the world. It does three things: advises companies on big mergers, helps them sell stocks and bonds, and trades and manages money for rich clients. It makes a lot when markets are busy and much less when they are quiet. The brand is real and the people are talented, but the earnings swing violently with the cycle, and the bankers themselves take a third of every dollar in pay. Buffett himself said this kind of business is harder to predict than most. So we pass.

Thesis

Sector caveat first. Goldman Sachs is an investment bank. The deterministic scorer's IV math (DCF on owner earnings minus maintenance capex) does not model financials correctly — invested capital for a bank is dominated by deposits, repo, derivatives, and tier-1 capital, not PP&E, and reported owner earnings of $18.3B TTM include trading-flow noise that does not recur ratably. The scorecard's iv_low/iv_base/iv_high of $1,046 / $1,652 / $2,097 is mechanically derived and unreliable here. We substitute the bank framework: anchor on tangible book per share, ROTCE, and CET1, and price against through-the-cycle (TTC) earnings rather than peak-cycle owner earnings.

The business. GS earns from three engines: (1) Global Banking & Markets — M&A advisory, equity/debt underwriting, FICC and equities trading; (2) Asset & Wealth Management — fees on $3T+ of AUS plus alternatives carry; (3) Platform Solutions — transaction banking and remnants of consumer (still being shrunk). Engine (1) is the franchise — top-tier league-table position in M&A and ECM for two decades — but its revenue per banker swings violently with the capital-markets cycle. Engine (2) is the higher-quality compounder GS is trying to grow toward.

Why it might compound. A scaled, brand-anchored advisory franchise with durable client relationships, a global FICC platform that benefits when volatility returns, and an AWM business migrating toward fee-based recurring revenue. Share count is down 2.9% over a decade — modest but consistent buybacks at a wide range of prices.

Price/IV math. P/E TTM is 20.14 vs a 10-year average of 13.88 — a 45% premium to its own history. The 10-year average ROIC printed by the scorer is 0.0% (a financials-formula artifact), but reported ROTCE has cycled 10–20% across the last decade with peaks in advisory boom years. Buying near 1.3–1.5× tangible book per share has historically marked decent entry; trim above ~2.0× TBV. At $923.71 with reported TBV near $370–$390/share (segment disclosures), GS is trading near 2.4–2.5× TBV — the upper end of fair, not the lower end of cheap. Buffett's own warning [1] applies: the economics of investment banking are "far less predictable" than industries he commits to.

Moat

Pricing power. Goldman commands premium fees in M&A advisory and equity underwriting because the seller's board cares about certainty of close and price discovery more than 25 basis points of fee. League-table leadership in announced M&A deal value compounds: CEOs hire the bank that just advised the last comparable deal, which is usually GS or Morgan Stanley. Yet pricing power is episodic, not continuous — in flat trading and underwriting, fees collapse and the franchise still pays the comp pool. Verdict on pricing: real in advisory, weak in commoditized FICC and equities flow.

Switching costs. Within institutional client relationships these are moderate. A pension fund using GS prime brokerage, derivatives, custody, research, and execution has stickiness from operational integration. But the client can — and routinely does — diversify across two or three primes. In M&A advisory, switching cost is essentially zero per-deal: the next mandate is re-competed. Damodaran [3] notes that for financial firms book values reflect market-marked balances, so the "asset" the customer is buying isn't a proprietary product — it's people and brand. People walk; juniors get poached every cycle.

Network effects. Modest in market-making (more flow → tighter spreads → more flow), strong in research distribution (top-rated analysts attract trading commissions), and meaningful in capital introduction within the alternatives ecosystem. Damodaran [4] makes the structural argument that investment-bank indispensability rebuilt itself via complexity — the more arcane the instrument, the more clients need an intermediary with the quant desk and the placement network. GS sits at the heart of that. But each new electronic-trading wave compresses the moat in the simpler products.

Intangibles (brand). This is the single strongest GS moat. The Goldman name carries reputational value with corporate boards, sovereigns, and ultra-high-net-worth families. Buffett's 1987 letter [1] emphasized that what attracted him to Salomon was "the ability and integrity of John Gutfreund" — he was buying people, not the franchise itself, and even then he priced his exposure as a convertible preferred because he could not underwrite the business economics. The brand is a real moat against new entrants — no fintech is going to advise on a $50B cross-border merger this decade — but it is not a moat against the cycle, against compensation creep, or against reputational shocks (1MDB, the consumer-banking misadventure with Marcus, Apple Card frictions). Brands in capital markets get reset by single events.

Cost advantages. Scale matters in technology spend and balance-sheet financing. GS funds itself cheaper than any boutique. But the comp ratio (~33–35% of net revenues) is the structural problem: the talent captures most of the marginal economics. As Buffett wrote in 1987 [5], "plenty of unintelligent capital allocation takes place in corporate America" and investment bankers are themselves the example of expensive money-shufflers [6]. The cost advantage is real versus boutiques, neutral-to-negative versus JPMorgan's universal-bank cross-subsidy of investment banking with deposits.

Competitor stress test ($10B + 5 years). Could a competitor armed with $10B and five years dent GS? In M&A advisory, no — the brand and roster are not buyable. In FICC and equities, partially — Citadel Securities, Jane Street, and Susquehanna have already eaten meaningful share in market-making without anyone "spending $10B" in the conventional sense; they spent intellectual capital on systems. The trading moat is narrowing. In AWM, Blackstone, Apollo, KKR, and BlackRock are the threat — and they are all larger or growing faster in alternatives than GS.

Erosion risk. The 2008 crisis wiped Bear and Lehman, forced GS into a bank-holding company, and permanently raised the capital intensity of the business. The Volcker Rule and Basel III/IV trimmed prop trading and inflated risk-weighted assets. The next regime change — whether Basel IV finalization, a stress test failure, an AML/sanctions enforcement event, or a major trading loss — could re-rate the multiple downward by 20–30% overnight. Buffett's 2008 warning [1, latticework] about back-tested models applies in spades to GS's VaR and stress frameworks.

Moat verdict: NARROW.

Management

The five capital-allocation choices applied to GS under David Solomon (CEO since October 2018):

(1) Reinvest in the business. The strategic narrative since 2020 has been to pivot mix toward Asset & Wealth Management (more fee-based, less balance-sheet-intensive, higher multiple) and to wind down the direct-to-consumer Marcus / Apple Card experiment that ate billions in provisions and operating losses. The pivot is correct. Execution has been mixed: Marcus checking was wound down, GreenSky was sold in 2024 at a loss versus the 2022 acquisition price, and the Apple Card relationship is being exited. These are admissions of strategic error — to management's credit they are admissions, not denials, but the original capital deployment was poor.

(2) Acquisitions. GreenSky ($2.2B in 2022 → divested 2024 at material loss) is the clearest mistake of the Solomon era. NN Investment Partners (2022) and the United Capital Financial Advisors deal (2019) were modest and strategically aligned with the AWM pivot. The track record is uneven. Buffett warned in his 2014 letter [2] that capital is best moved by people who do not have entrenched interests in the source business — GS's M&A bankers internally advising on GS acquisitions is exactly the principal-agent issue he describes.

(3) Debt. Investment banks live on leverage; the question is quality of funding, not quantity. GS has lengthened liability duration meaningfully post-crisis, holds large pools of HQLA, and runs CET1 in the 14–15% range — above regulatory minimums. The deterministic scorer's net-debt-to-EBITDA of -79.05× is meaningless for a bank (negative and very large reflects gross deposits/financing exceeding gross financial assets in the calc). Real question: is duration mismatch managed? GS's duration profile and counterparty exposure look acceptable under current disclosures.

(4) Buybacks. Share count is down ~2.9% over 10 years (per scorer). This is modest — JPMorgan and Morgan Stanley have shrunk floats more aggressively. Importantly, GS has bought back stock across a wide range of prices, including at peaks (2021 at $400+, 2024 at $500+). The current price $923.71 is roughly 2.4× tangible book — a price at which a Buffett-disciplined allocator should slow, not accelerate, repurchases. The pattern is closer to "return excess capital after stress test" than "opportunistic concentration at discount." Grade for buyback discipline: average.

(5) Dividends. GS has raised the dividend steadily; current yield is modest (~2%). The dividend has not been cut since the GFC. Reasonable.

Communication quality. Solomon's investor-day presentations have improved since the messy 2020–2022 period. Segment reporting was simplified (consolidated to three segments) which made the franchise easier to underwrite. The firm acknowledged the consumer-banking errors publicly and quantified the losses. That said, Solomon himself has been the subject of meaningful internal-culture friction (DJ side-gig optics, partner departures), and Goldman's leadership cadence under Solomon has been more turbulent than under Lloyd Blankfein. The board's tolerance is high so far.

Insider alignment. Partners and managing directors hold significant restricted stock, with multi-year vesting and clawbacks. Alignment is solid. But comp ratio of 33–35% of net revenues means the employees, not shareholders, are the residual claimants on a good year. This is the structural challenge of the partnership-DNA business as a public company — it is not a Solomon-specific failing.

Capital allocator: C. Above-average versus peers on capital return discipline post-crisis, below-average on the consumer-banking misadventure and on opportunistic buyback timing. The franchise is excellent; the allocator is competent but not exceptional.

Industry

Porter's Five Forces applied to bulge-bracket investment banking:

Rivalry among existing competitors: HIGH. The bulge bracket is GS, MS, JPM, BofA, Citi, plus European universals (Barclays, Deutsche, UBS post-CS), plus elite boutiques (Evercore, Centerview, Lazard, Moelis, PJT) in advisory. In trading, the rivalry now extends to non-bank market-makers (Citadel Securities, Jane Street, Virtu, Susquehanna) which run lower-cost, more electronified models. Fee compression in cash equities and rates is a permanent feature; pricing power survives only in capital-commitment products and complex underwriting. League-table competition is brutal and re-set every quarter.

Threat of new entrants: LOW in advisory, MEDIUM-HIGH in flow trading. Regulatory capital requirements, the Fed's bank-holding-company supervision, and the brand barrier make a new bulge-bracket entrant essentially impossible. But in flow trading, technologists with capital have entered and won — Citadel Securities went from nothing to a top market-maker in a decade. The bar to entry depends entirely on which product line.

Bargaining power of suppliers: HIGH. The "supplier" is talent. Senior bankers, traders, and PMs walk for 20–40% pay rises. The compensation ratio (~33–35% of net revenues) measures exactly this — labor captures a third of the value the franchise produces. Buffett's repeated frustration with investment-bank compensation [6] is the structural critique: the people, not the shareholders, are the residual claimants.

Bargaining power of buyers: MIXED. Corporate clients on M&A advisory have limited power on the first call (they want the best banker for the deal) but high power across the cycle (relationships rotate, fees compete). On trading, institutional buyers (hedge funds, asset managers) have very high power — they can route flow electronically to the cheapest venue. On AWM, client stickiness is higher because of operational lock-in and performance-based persistence.

Threat of substitutes: MEDIUM and growing. Direct issuance platforms, private credit (Apollo, Blackstone, Ares) substituting for traditional leveraged finance, electronic execution replacing voice trading, and secondaries markets disintermediating IPOs — all are real substitution threats that have advanced over the last decade. Private credit alone has grown to ~$1.5T+, displacing meaningful syndicated-loan and high-yield volume that historically went through the bulge bracket.

Value-pool location and trajectory. The capital-markets value pool has fragmented: trading commissions are down structurally, advisory fees are volatile but durable, AWM fees are growing (especially alternatives), and private credit is taking the leveraged-finance pool from public markets. GS's three-segment redesign explicitly chases the AWM pool. The trajectory is correct; the question is whether GS gets to a 60%+ AWM mix before the next trading cycle resets.

Damodaran's structural point [3, 4]. Financial-services firms are regulated, which means part of their economics depends on regulatory choices we cannot underwrite. Comparing P/E ratios across banks without controlling for risk is a category error. And [4] — investment-bank indispensability rebuilt itself through complexity. As long as instruments get more complex, the bulge bracket retains a role. As soon as standardization wins (electronic, exchange-traded), the role shrinks.

Industry verdict: AVERAGE. The franchise tier is excellent for the top two players, but the industry economics are mediocre by Buffett's standard: high cyclicality, labor-captured rents, regulatory exposure, and substitution from private credit and electronic markets. This is not a "toll bridge" or a "branded consumer staple."

Inversion

Playing the short-seller. No softening.

(1) The single event that kills this. A counterparty failure or a trading loss large enough to trigger a regulatory capital-rebuild and a bank-stress-test failure. Goldman's balance sheet is, fundamentally, unknowable from the outside — Damodaran [3] makes the central point that financial-firm assets are marked-to-something, regulated by someone, and investors are trusting that "the regulatory authorities will keep banks and insurance companies in check." When that trust ruptures — as it did in March 2023 with Credit Suisse, Silicon Valley Bank, First Republic — the re-rating is instant and brutal. GS's specific kill-shot would be: a major prime-brokerage counterparty (think Archegos-scale, but for GS instead of Credit Suisse) or a derivatives concentration failing in a way that requires an emergency capital raise at depressed prices. The 2021 Archegos episode cost Credit Suisse $5.5B and started the chain that led to its UBS-forced takeover. It can happen here.

(2) Why the moat is narrower than bulls think. Bulls say "the brand is unassailable." Reality: the brand is people. Top advisory bankers and traders walk every cycle. Centerview, PJT, Evercore, Moelis — every one founded by ex-bulge-bracket partners who took client relationships with them. In trading, Citadel Securities and Jane Street have eaten meaningful share in equities and options market-making in a decade. In flow rates, electronification has compressed margins to near-commodity levels. The moat is real in complex M&A and complex underwriting — both of which are small slices of total revenue and highly cyclical. The "scaled franchise" moat narrative dramatically overstates how much of the revenue is truly moated.

(3) Why management is worse than it appears. David Solomon's record includes: the GreenSky fiasco (acquired 2022 for $2.2B, divested 2024 at substantial loss); the Marcus consumer-banking write-downs; the Apple Card relationship which has been a operational mess and is being exited at a loss; reported partner departures and culture friction; the side-gig DJ optics that distracted the board. The board's defense is that legacy investment-banking compensation pressures and the post-2018 talent battles required leadership flexibility. The bear reads the same evidence as: weak strategic discipline, weak execution on M&A integration, and a CEO who spent shareholder capital chasing a consumer pivot that was strategically incoherent. Buffett's 2014 letter [2] on capital reallocation specifically warns that CEOs with capital trapped in declining operations seldom redeploy well — the consumer experiment was redeployment, and it failed.

(4) What bulls are extrapolating that won't hold. Bulls extrapolate: (a) the M&A activity recovery off the 2022–2023 trough as the new normal; (b) AWM growth at 10%+ persistently; (c) FICC peak-cycle revenues as sustainable; (d) Basel IV finalization being benign. Each of these is a possible future, not a probable one. M&A volumes are mean-reverting around GDP-plus, not exponential. AWM growth at GS lags Blackstone, Apollo, KKR, and BlackRock — the share gain story is not playing out. FICC is volatile by design — peak revenues require the Fed and ECB to keep rates volatile and credit spreads to keep moving. Basel IV finalization could add 5–10% to RWA, mechanically compressing ROTCE. Damodaran [3] is explicit: "riskier banks will look cheaper" — but the market often misprices the risk because back-tested models (Buffett 2008) miss the regime change.

(5) Valuation trap (multiple compression / regime change). GS trades at P/E TTM 20.14 versus a 10-year average of 13.88. That is a 45% premium to its own history. The implied path is: keep peak-cycle earnings at $18B+ and re-rate from 14× toward 18–20× as AWM mix grows. The bear case: trailing earnings normalize down 25–35% as M&A and FICC mean-revert, AND the multiple compresses back toward 12–13× as the cycle becomes obvious. The combination — earnings down 30%, multiple down 35% — implies a stock price of roughly $923 × 0.7 × 0.65 = $420. That is not a tail scenario; it is roughly what GS traded at in October 2022 and October 2023. The "back to historical norm" trap is the most likely bear path.

Bonus bear vector: regulatory regime change. Basel IV finalization is the largest known unknown. The Fed's proposed implementation in 2023 added roughly 20% to large-bank RWA at consultation; the rule was softened, but the direction is settled — capital intensity for the bulge bracket is going up, not down, over the next several years. Higher RWA mechanically compresses ROTCE for any given dollar of net income. In addition, the SEC's ongoing focus on private-market disclosure could reduce the AWM information-asymmetry advantage that justifies premium fees in alternatives. None of these are tail events; they are base-case directions of regulatory travel.

Bonus bear vector: AI displacement of mid-tier banker work. A non-trivial fraction of associate-and-VP work in M&A and ECM is producing pitch books, running comp tables, structuring drafts, and document review. Generative AI is going to compress the labor required for this work meaningfully over the next five years. Bulls argue this raises GS's margin (less labor cost per deal). Bears argue it accelerates the boutique threat — if AI does the junior work, the boutique partner can match the bulge-bracket pitch book with one analyst instead of five, and the moat narrows further. Net effect is uncertain, but the direction of compression is clear.

Bonus bear vector: succession. David Solomon's tenure has been more turbulent than Lloyd Blankfein's, and the partnership talent base has lost notable senior departures over the last several cycles. The CEO transition (whenever it comes) is itself a discrete risk event for a brand-driven business — the next CEO inherits the consumer-banking cleanup, the private-credit competitive threat, and a partner culture in flux. Succession risk in investment banks is materially higher than in industrial businesses because the people are the franchise.

Combining the vectors. The five primary bear arguments and the three bonus vectors do not require a recession or a market crash to play out. They require only that mean-reversion happens, that regulation continues its post-2008 trajectory, and that competitive substitution from private credit and from non-bank market-makers continues at the current pace. None of those is a forecast — each is the continuation of an existing trend. Compounded across two to three years, they justify a price re-rating toward the mid-$400s.

If I am right, the stock could be worth $420 within 24 months.

Lollapalooza Bias Check

Biases active in me as an analyst right now.

Authority bias (HIGH). Buffett's 1987 Salomon letter [1] is sitting in my canon excerpts and is the most quotable piece of source material on investment-bank economics in the entire pipeline. I am almost certainly over-weighting his explicit "far less predictable" framing because it makes for a cleaner narrative. The honest counter: Buffett's view of investment banking has evolved — he held Goldman common stock from the 2008 warrant exercise through 2020, generated very large profits, and only exited when the Apple position made it small relative to the book. Berkshire also held bank stocks heavily for decades. The "Buffett doesn't do investment banks" claim is a simplification. I should treat it as an input, not a verdict.

Anchoring (HIGH). I am anchored on the P/E TTM 20.14 vs 10y avg 13.88 ratio — a 45% premium that implies expensiveness. But "10-year average P/E" for a cyclical includes deep-trough multiples in 2015–2018 and 2022 that reflected market skepticism more than fundamentals. The comparable I should anchor on instead is price-to-tangible-book through-cycle — and there I am also seeing a premium, but a smaller one. Anchoring on a single ratio is a known investment-research mistake.

Confirmation bias (MEDIUM). I started with the prior that "investment banks are too hard" — every quote and every datapoint I selected from the filings and from the canon reinforces that prior. I did not work hard at the bull case. A genuine bull case for GS exists: durable advisory franchise, growing AWM, disciplined capital return, reasonable balance sheet, attractive yield-plus-buyback. I did not steel-man it.

Recency bias (MEDIUM). The 2023 regional banking crisis (SVB, Signature, First Republic, Credit Suisse) is fresh in my mind and is making me over-weight tail risk for GS even though GS's funding profile is structurally different from those failed institutions. Wholesale-funded, large-deposit, investment-bank balance sheets failed; deposit-light, prime-brokerage-heavy GS is a different animal. I am bleeding 2023 sentiment into 2026 analysis.

Incentive (LOW for me, HIGH for the company). I have no skin in the game on GS. But the bias I should flag, per Munger's incentive doctrine, is inside the company: 33–35% comp ratio means the people running the trades and the deals are paid to do them, not to refuse them. That is the single most important Munger lens for evaluating any investment bank.

Net effect on my recommendation. Authority bias and anchoring are pushing me harder toward "Too Hard" than the underlying facts may justify. Confirmation bias compounds it. The honest reading is that GS is a real franchise at a fair-to-full price — closer to Hold than to Avoid. I will land at "Too Hard" because the Buffett-Munger discipline is to skip businesses where I cannot confidently model the next ten years of earnings, and investment-bank earnings genuinely fail that test for me — but I am flagging the biases honestly.

10-Year Outlook

Same fundamental business model in 10 years? Mostly yes. GS will still advise on M&A, underwrite securities, trade FICC and equities, and manage assets and wealth. Mix will have shifted further toward AWM (probably 40–50% of revenue vs ~25% today). The consumer experiment will be a historical footnote.

Customer base larger? Probably. Global capital-markets activity grows roughly with global GDP-plus over long horizons. UHNW wealth pools are growing faster than aggregate GDP, favoring AWM. But the number of counterparties in trading is concentrating into fewer, larger players — which is mixed news for an intermediary.

Profit per customer higher? Hard to say. In advisory, fees-per-deal are roughly flat in real terms over decades — large transactions pay 50–100bps, mid-cap 1–3%, and that range hasn't moved much. In trading, profit per ticket has been declining for two decades on electronification and will continue to. In AWM, fees are compressing slowly (from ~85bps blended to ~70bps) as ETFs and passive allocations grow share, partially offset by alternatives at higher fees. Net: probably flat-to-slightly-down per customer, with growth driven by volume and mix.

Moat wider in 10 years? Probably narrower in trading (electronification, new entrants), stable in advisory (brand-and-people-driven), and possibly narrower in AWM (Blackstone, Apollo, BlackRock are out-growing). The franchise as a whole is most likely flat-to-narrower, not widening.

Single biggest threat? Private credit and direct-lending displacement of leveraged finance, combined with continued electronification of flow trading. Together these compress two of GS's three legacy revenue pools. The AWM pivot is the offset, but it depends on execution against better-positioned alternative-asset managers.

Confidence assessment. I cannot tell you what GS will earn in 2034 with even one-standard-deviation confidence. That alone is the Buffett test failing — Buffett's criterion is identifying businesses where you can predict ten years out, not businesses you find interesting. The franchise is real, the brand is real, the people are talented, but the earnings profile is structurally cyclical, regulatorily exposed, and dependent on macro variables (rates, volatility, M&A activity) that no one can predict.

CONFIDENCE: low

Position Guidance

• Recommendation: Too Hard
• Conviction: medium
• Target buy price: $560 (~1.5× tangible book, the historical bank buy-zone anchor; only a margin-of-safety re-entry candidate, not a Buffett-quality compounder)
• Target trim price: $1,050 (~2.8× tangible book, the upper end of the bank-trim range)
• Position sizing: Zero. Buffett-Munger discipline says skip businesses where you cannot model the next decade with confidence. Investment-bank earnings volatility, regulatory exposure, and labor-share dynamics put GS outside the circle of competence for a concentrated long-term portfolio. If forced to size: maximum 1–2% as a cigar-butt-style entry below $560.
• Conditions to revisit: (a) GS trading below 1.5× tangible book during a sector dislocation; (b) AWM mix exceeding 45% of revenue with demonstrated fee-revenue stickiness through a down cycle; (c) sustained ROTCE above 15% through a full M&A cycle (peak + trough), not a single-year print.