Goldman Sachs: Profits amid public bailout
The Inventory of Risk
In the spring of 2009, while the American economy remained concussed from the financial crisis, Goldman Sachs recorded $45 billion in revenue, early triple its historical average. The firm had just converted from an SEC-regulated broker-dealer to a Federal Reserve bank holding company to access emergency funding, a move that theoretically constrained its risk-taking. Yet there it was, posting record profits amid public bailout money and private ruin. The outrage was understandable but misplaced. Goldman had not simply navigated the crisis; it had positioned itself as the buyer of last resort for risk that others could no longer hold. When the market for mortgage securities seized, Goldman held inventory. When corporations needed to hedge currency exposure and their usual counterparties had vanished, Goldman took the other side. The $45 billion was not a windfall from speculation but the premium charged for providing liquidity when liquidity had fled.
This is the essential mechanic of Goldman Sachs: the firm generates revenue by sitting at the center of institutional capital flows and trafficking in risk. It does this through three distinct but interwoven mechanisms. First, it takes risk directly onto its $1.5 trillion balance sheet, underwriting securities, holding inventory, maintaining counterparty positions, earning spreads that widen precisely when markets become chaotic. Second, it matches risk between parties that would otherwise never meet, intermediating between oil producers seeking price stability and airlines seeking the same, capturing the spread between their divergent anxieties. Third, it structures and sources risk for institutional investors, packaging derivatives and structured products that transfer exposure from those who cannot bear it to those who can. Investment banking, the marquee advisory work of initial public offerings and mergers, contributes roughly 20 percent of revenue through fixed fees calibrated to transaction value: three percent on IPOs, fifty to sixty basis points on mergers. Asset management contributes another 25 percent through management fees on $2.5 trillion in assets, supplemented by returns from principal investing. But the engine room remains sales and trading, which generates roughly 40 percent of revenue through bid-ask spreads on securities inventory and financing margins, producing annual earnings that swing wildly between $9 billion and $18 billion depending on the volatility of the underlying markets.
The economics of this model are brutally simple. Goldman employs approximately $100 billion in equity capital to support $1.5 trillion in assets: leverage ratio that would be illegal for an ordinary industrial company but is the essence of banking. The firm borrows short and lends long, holds inventory when others cannot, and provides immediacy to buyers and sellers who lack the balance sheet to wait for the right counterparty. Profitability depends on the velocity of this inventory turnover and the width of the bid-ask spreads captured in the process. When volatility collapses, as it did between 2017 and 2019, spreads compress and financing margins evaporate. When volatility explodes, as in March 2020, the spreads widen dramatically, but only if Goldman has positioned itself to be net long the volatility rather than caught holding depreciating inventory. The business is, by design, a black box. The inventory changes daily, the positions are hedged in ways that resist simple decomposition, and the earnings resist modeling. This opacity is not incidental to the model. It is the model.
Thirty Cents on the Dollar
There is a statistic that receives less attention than trading revenues or return on equity but which explains more about Goldman’s durability than any quarterly earnings report. Twenty years ago, the firm paid out fifty cents of every revenue dollar to employees in compensation. Today, that figure hovers near thirty cents. This compression, from fifty percent to thirty percent, represents one of the most significant margin expansions in modern financial services, and it has occurred not through automation or headcount reduction but through the economic exploitation of prestige.
Goldman Sachs remains the premier destination for young financiers not merely because it offers the highest salaries, though it offers competitive salaries, but because it offers a currency that competitors cannot manufacture: the signal. To have traded at Goldman or structured products at Goldman or advised on mergers at Goldman is to have passed through a filter that subsequent employers, investors, and society at large recognize as valid. This brand equity functions as a direct subsidy to shareholder returns. Competitors must pay premium compensation to attract talent capable of executing complex derivatives trades or navigating the regulatory labyrinth of cross-border M&A. Goldman pays a discount, or rather, it pays a market wage but extracts surplus value through the non-pecuniary benefits of association. The thirty-cent ratio is not merely a cost structure; it is a competitive moat.
This advantage is fragile precisely because it is intangible. The prestige rests on a century and a half of institutional memory, on the mythology of the partnership culture that persisted even after the 1999 initial public offering, on the selective hiring and the brutal culling that creates an aura of elite competence. If the firm were to suffer a reputational catastrophe, particularly one involving the consumer-facing businesses it has recently entered, the talent discount could evaporate. If the best graduates of the best universities begin to view Goldman as merely another bank, or worse, as a compromised institution, the compensation ratio would revert toward historical norms, compressing the thirty-cent advantage and destroying the margin structure that supports the target 14 to 16 percent return on equity. The brand is not marketing. It is a balance sheet item.
The Leverage Trap
The defining structural constraint shaping every strategic decision at Goldman Sachs is the post-2008 regulatory framework that governs bank capital. When the firm became a bank holding company in 2008, it accepted oversight by the Federal Reserve and with it, the strictures of Basel III capital requirements. Common Equity Tier 1 ratios, the measure of high-quality capital to risk-weighted assets, are now mandated at levels significantly higher than those that prevailed during the firm’s most profitable decades. The mathematics are inexorable: return on equity equals return on assets multiplied by leverage. If regulators reduce leverage by demanding more equity capital against the same asset base, the firm must either accept lower returns or increase the riskiness of its assets to maintain the same ROE.
Goldman has chosen the latter path repeatedly, but it is running out of room. The $1.5 trillion balance sheet is funded by roughly $100 billion in equity, giving a leverage ratio that appears modest by historical banking standards but which remains the engine of the business. Every additional dollar of equity capital that regulators require, whether through stress testing or macro-prudential buffers, directly reduces the return on the existing equity base unless the firm can find risk-adjusted assets that yield proportionally higher returns. This is the tension that animates the strategic confusion of the past decade: Goldman must either accept lower returns that fail to cover its cost of capital, or it must take more risk per dollar of assets, or it must find businesses that generate high returns without consuming balance sheet space.
The search for capital-light revenue has pushed the firm into asset management and, more controversially, into consumer banking. Asset management fees, twenty basis points on $2.5 trillion, require little regulatory capital and provide stable, recurring revenue. But they lack the margin and the information asymmetry of the trading business. Principal investing, where Goldman deploys its own capital alongside client capital, generates higher returns but consumes scarce balance sheet and invites regulatory scrutiny. The firm is thus trapped in an optimization problem with no clear solution: the businesses that justify the leverage are the ones regulators most want to constrain; the businesses that please regulators dilute the returns that justify the equity.
Volatility is the Product
Institutional investors despise unpredictability in their portfolio companies, yet Goldman Sachs sells unpredictability. The sales and trading division, which contributes roughly forty percent of revenue, functions as a volatility harvesting machine. In years when markets move sideways and volatility indexes remain suppressed, the division struggles to generate $9 billion. In years of crisis or dislocation, it can produce $18 billion. The mechanism is straightforward: when uncertainty increases, the bid-ask spreads on securities inventory widen, and the premium charged for providing immediacy expands. Goldman, with its $1.5 trillion balance sheet and its ability to fund positions through repo markets and deposits, can hold inventory when smaller dealers are forced to liquidate. It becomes the market maker of last resort, and it charges accordingly.
This creates a strategic paradox. The firm’s valuation multiple remains depressed, typically trading near or below book value, because investors cannot model the earnings power of the trading division. The volatility is the product, but it is also the poison. Equity analysts attempt to forecast earnings by examining historical trading volumes and volatility regimes, but these models break down precisely when Goldman’s opportunities are greatest. The firm cannot provide transparency into its positions without destroying the alpha those positions generate. If Goldman were to reveal its hedging strategies or its inventory composition in sufficient detail to satisfy investors seeking earnings stability, it would simultaneously reveal information that competitors could exploit or that counterparties could front-run.
The opacity is thus structural, not cultural. It is the necessary condition for the intermediation spreads that drive profitability. Yet this opacity conflicts with the regulatory and investor preference for stable, predictable, fee-based revenue. Goldman has attempted to resolve this tension by expanding into wealth management and consumer banking, usinesses that generate recurring fees and appear more “valuable” to equity markets. But these businesses conflict with the core competency. The institutional sales force that can price a complex interest rate derivative is not the same workforce that can patiently manage the regulatory compliance of a retail deposit base. The risk management systems designed to monitor counterparty credit exposure are ill-suited to detecting consumer fraud at scale.
The Consumer Experiment and the DNA Conflict
In 2016, Goldman launched Marcus, a consumer deposit platform, followed by an aggressive push into wealth management for the merely affluent rather than the ultra-high-net-worth. The strategic logic was sound on paper: consumer deposits provide stable, low-cost funding for the balance sheet, and wealth management generates predictable fee income that trades at higher valuation multiples than trading revenue. Yet the move represents a fundamental category error, or at least a dangerous bet against the firm’s own history.
For 150 years, Goldman Sachs operated exclusively in institutional markets. It did not accept retail deposits; it did not maintain branch networks; it did not field customer service representatives handling disputed credit card charges. The firm’s infrastructure, its culture, its hiring patterns, and its risk appetite were calibrated for a specific ecosystem: corporations, governments, hedge funds, and sovereign wealth funds. These clients understand that Goldman is a counterparty, not a fiduciary. They accept that Goldman will take the other side of their trades when it suits the firm’s book. They understand that advice may be conflicted. This is the implicit contract of institutional finance.
Retail consumers operate under different assumptions. They expect their bank to act in their interest, not to hedge against their mortgage positions. They expect their deposits to be safe not merely because of FDIC insurance but because of an institutional commitment to safety. If Goldman suffers trading losses that threaten its capital position, institutional clients may continue to trade with the firm precisely because they trust Goldman’s ability to manage risk, hey understand that volatility cuts both ways. But retail depositors may flee at the first sign of trouble, triggering a run that destabilizes the funding base. Moreover, the regulatory scrutiny attendant to consumer-facing businesses, fair lending laws, consumer protection bureaus, the operational risk of handling millions of small accounts, introduces liability profiles that the firm has historically avoided.
The deeper risk is reputational dilution. The prestige that allows Goldman to hire talent at a thirty-cent discount depends on an aura of exclusivity and institutional sophistication. If Goldman becomes known primarily as a high-yield savings account provider or a mass-market wealth manager, the brand signal weakens. The talent discount evaporates. The firm becomes indistinguishable from JPMorgan Chase or Bank of America, except with higher cost structures and less operational experience in consumer compliance.
The Central Node Assumption
The most fragile long-term assumption underlying Goldman’s business model is that the firm will remain the central node in institutional capital flows even as technology and market structure evolve. The model assumes that corporations will continue to need Goldman to intermediate their risk, that hedge funds will continue to rely on Goldman’s balance sheet for financing, and that the information asymmetry derived from sitting between thousands of counterparties will persist despite electronification and disintermediation.
This assumption is being tested. Electronic trading platforms have reduced the need for voice brokers in standardized products. Direct indexing and algorithmic execution allow institutional investors to replicate strategies that once required Goldman’s structured products desk. Blockchain technology, whatever its ultimate trajectory, threatens to reduce the need for centralized intermediaries in certain settlement and custody functions. The question is whether these technologies eliminate the need for Goldman’s specific bundle of services, balance sheet, trust, and structural alpha, or merely shift the locus of value to more complex, illiquid, and bespoke transactions.
Goldman’s defense lies in the illiquidity premium. As markets become more automated and transparent, the spreads available in standardized products compress, but the complexity of the products that remain bespoke increases. The firm has spent a century building relationships and balance sheet capacity that cannot be replicated by software. When a sovereign wealth fund needs to hedge a twenty-year exposure to Mongolian currency risk, or when a corporation needs to securitize an idiosyncratic stream of cash flows, Goldman remains one of perhaps three global institutions capable of pricing and holding that risk. The durability of this advantage depends on continued market fragmentation, on the persistence of regulatory barriers that prevent non-bank entities from taking similar risks, and on the continued willingness of institutional clients to pay significant intermediation spreads rather than build in-house capabilities.
The Price of Being Understood
Goldman Sachs trades at a persistent discount to book value, or at best, at book value. This is the market’s verdict on the opacity of the trading business and the regulatory overhang that constrains leverage. The discount represents a cost of capital that the firm must overcome to create shareholder value. Yet attempts to close this discount by becoming more transparent, by shifting toward stable, fee-based revenue, threaten to destroy the very earnings that make the book value worth owning.
This is the final tension: Goldman cannot become a traditional asset manager or commercial bank without losing the high-margin, high-information-asymmetry activities that justify its existence. But it cannot remain a pure trading house without accepting a valuation discount and constant regulatory hostility. The firm is attempting to thread this needle by operating both businesses simultaneously, aintaining the volatility-dependent trading franchise while building stable, recurring revenue in wealth and consumer banking. But this diversification strategy introduces its own fragilities. The capital allocated to Marcus and to wealth management is capital not available for trading inventory. The management attention consumed by consumer compliance is attention diverted from institutional client relationships. The cultural accommodation of retail banking risks diluting the prestige that subsidizes the compensation structure.
Fortune favors the old in finance because trust accrues slowly and dissipates quickly. Goldman’s $1.5 trillion balance sheet and its thirty-cent compensation ratio are the accumulated sediment of 150 years of institutional memory. The firm persists not because it has found a way to resolve the tensions between leverage and regulation, between volatility and stability, between opacity and valuation, but because it has managed to remain profitable despite them. The risk is that these tensions are not cyclical but structural, hat the regulatory environment will continue to tighten, that technology will continue to disintermediate, and that the consumer experiment will erode the brand equity without generating sufficient returns to compensate for the lost leverage.
The question for Goldman Sachs is whether it can remain the buyer of last resort in a world that increasingly prefers buyers of first resort, ransparent, regulated, predictable, and boring. The $45 billion revenue of 2009 was not an aberration but a demonstration of the model’s purest form: when the market breaks, Goldman holds the inventory. But markets are not breaking as they once did, and when they do, regulators ensure that Goldman cannot hold quite as much inventory as it once could. The firm is becoming a more stable institution by force, but stability was never the source of its advantage. The advantage was always the willingness to hold the risk that others feared, to traffic in the volatility that others avoided, and to do so with a balance sheet backed by the implicit conviction that Goldman knew something the market did not. Whether that conviction remains justified will determine if the next 150 years resemble the last.
Whether that conviction remains justified will determine if the next 150 years resemble the last.
The conviction rests on a specific form of knowledge that does not appear on any balance sheet but generates the returns that make the balance sheet possible. When Goldman trades mortgage-backed securities or structures a collateralized loan obligation, it is not merely matching buyer and seller. It is synthesizing information from a thousand simultaneous conversations: the corporate treasurer describing receivables timing, the hedge fund manager revealing leverage constraints, the pension fund stating its duration requirements. This information, fragmentary, embargoed, and often illegal to act upon in public markets, coalesces into a proprietary view of where risk is mispriced. The trader who knows that three corporate clients simultaneously need to sell the same credit exposure can structure a trade that extracts value from that desperation without revealing the desperation to the wider market. The balance sheet exists to warehouse this risk during the interval between recognition and price discovery. That interval has historically been Goldman’s alpha.
Yet the half-life of this informational advantage is lengthening in ways that threaten the model. The electronification of credit markets, the standardization of derivatives clearing, and the transparency mandated by post-crisis regulation have systematically eliminated the pockets of opacity where Goldman once operated. When a corporate bond trades on an electronic platform with visible depth of book, the spread compresses to the marginal cost of capital plus a software maintenance fee. Goldman’s physical trading floor, still populated by hundreds of humans shouting into headsets, becomes an expensive anachronism in markets where price discovery occurs in milliseconds. The firm has responded by retreating up the complexity curve, focusing on bespoke structured products that resist automation: the customized total return swap for a family office, the illiquid private credit facility for a mid-market buyout firm. But these transactions consume disproportionate amounts of the scarce resource, regulatory capital, while generating fees that do not compensate for the leverage foregone. The black box is becoming a smaller box.
Consider the mechanics of a typical structured credit trade in 2007 versus 2024. Fifteen years ago, Goldman might have underwritten a collateralized debt obligation, retaining the equity tranche while selling the senior pieces to insurance companies. The structuring fee, perhaps fifty basis points on a billion-dollar deal, captured the value of Goldman’s proprietary default correlation models. The firm then traded the correlation itself, taking long and short positions against the slices it had created, profiting from the divergence between its view and the market’s. Today, that same CDO has been replaced by direct lending funds and private credit vehicles that operate outside the banking system entirely. The insurance companies have built their own credit teams. The correlation models are commoditized software. The spread that remains is competed away by non-bank lenders who do not face the Supplementary Leverage Ratio that forces Goldman to hold capital against every notional dollar of exposure. The black box still functions, but it processes a shrinking universe of trades.
The fragility of the talent discount, the thirty-cent ratio, deserves closer examination than the headline figure suggests. The compression from fifty cents to thirty cents over two decades appears to represent a structural efficiency, but it masks a more troubling underlying dynamic. In the 1990s, the fifty-cent ratio applied to a smaller revenue base generated by a smaller balance sheet; the absolute compensation per employee was high, but the fixed cost structure was flexible. When markets turned, Goldman could reduce the bonus pool and align labor costs with revenue. Today, the thirty-cent ratio applies to a revenue base that includes substantial fixed costs: the technology infrastructure required for regulatory compliance, the consumer banking operations with their sticky deposit bases and branchless-but-still-expensive customer acquisition, and the asset management platform with its long-dated fund commitments. The ratio has compressed not because Goldman is paying less for equivalent talent, but because the denominator has shifted. The firm is paying thirty cents on a dollar that includes stable, low-margin wealth management fees that require different, less scalable labor inputs.
Moreover, the prestige that subsidizes the compensation is eroding at the margins where it matters most. The analyst and associate classes, the raw material from which future partners are forged, are increasingly viewing Goldman as a training ground rather than a destination. The average tenure of a vice president in the investment banking division has shortened as private equity firms, sovereign wealth funds, and technology companies poach mid-level talent with offers of carried interest or equity that Goldman cannot match within its partnership structure. The firm responds by accelerating the path to managing director, diluting the title’s scarcity value, which in turn reduces the signaling power for those who remain. It is a deflationary spiral masked by the aggregate ratio. When the compensation compression reverses, hen Goldman must pay thirty-five or forty cents to retain the specific talent capable of pricing a complex derivative or executing a hostile takeover defense: he margin expansion of the past two decades unwinds rapidly. The thirty-cent ratio is not a permanent feature of the cost structure; it is a temporary equilibrium sustained by the absence of a viable competitor for the specific bundle of prestige and training, an equilibrium that weakens with each high-profile departure to a tech startup or crypto fund.
The consumer experiment provides the clearest evidence that institutional DNA cannot be rewritten by strategic mandate. Marcus, launched with the ambition of gathering $125 billion in deposits and becoming a top-tier consumer bank, has instead accumulated losses exceeding $3 billion while capturing a deposit base characterized by high sensitivity to interest rates: he flightiest kind of liability. The customers Goldman attracted were not the sticky, low-cost retail depositors who form the foundation of a JPMorgan Chase franchise, but rate-chasers who opened accounts for the high-yield savings rates and departed immediately when competitors offered basis points more. The cost of acquisition, fueled by partnerships with Apple and Intuit, exceeded the lifetime value of the relationships. More damagingly, the operational requirements of servicing these accounts exposed Goldman to a regulatory scrutiny it had never faced. When the Consumer Financial Protection Bureau investigates overdraft fee practices or account opening procedures, the legal and compliance costs fall upon a fixed cost base designed for institutional relationships where such issues rarely arise.
The Apple Card partnership illustrates the categorical mismatch. Goldman provides the balance sheet and the regulatory umbrella; Apple provides the customer interface and the brand. Yet the economics are inverted from Goldman’s traditional model. In institutional finance, Goldman extracts the spread because it bears the counterparty risk and holds the information advantage. In the Apple Card arrangement, Goldman bears the credit risk, the cardholders are Apple’s customers, not Goldman’s, while Apple captures the interchange fees and the data. Goldman becomes a commoditized lender, the functional equivalent of a industrial bank, but with the overhead structure of a premier investment bank. The deposits gathered through Marcus were supposed to provide cheap funding for the trading book, but the regulatory treatment of brokered deposits and the flightiness of the customer base have instead created a funding source more expensive and less stable than the wholesale markets Goldman historically used. The consumer pivot has not diversified the revenue; it has contaminated the balance sheet with liabilities that behave badly precisely when the trading book needs funding most.
Meanwhile, the competitive landscape has shifted in ways that disadvantage Goldman’s specific bundle of capabilities. Morgan Stanley, once the most direct comparable, executed a decisive pivot toward wealth management through the acquisitions of Smith Barney, Eaton Vance, and E*TRADE. These moves diluted Morgan Stanley’s trading culture but created a stable, fee-based revenue stream that now commands a premium valuation and provides low-cost funding through stable brokerage deposits. Goldman, by contrast, attempted to build rather than buy, constructing Marcus from scratch while maintaining the trading franchise. The result is a firm caught between models: too capital-intensive to compete with the pure asset managers, too regulated to compete with the private credit funds, too institutionally focused to compete with the retail banks, and too complex to be acquired. It is the financial equivalent of a buggy-whip manufacturer that has invested heavily in automobile manufacturing but cannot master the assembly line.
The regulatory constraints are tightening not through headline rule changes but through the accretive weight of supervisory expectations. The Stress Capital Buffer, imposed annually based on the Federal Reserve’s hypothetical scenarios, has effectively become a binding constraint that varies with the Fed’s assessment of Goldman’s internal risk models. When the Fed rejects a firm’s model for calculating risk-weighted assets, a process opaque to outsiders, the immediate effect is to increase the denominator of the capital ratio, forcing either deleveraging or capital raising. Goldman has suffered multiple model rejections in recent years, each requiring a sudden shrinkage of the balance sheet or a suspension of buybacks. This creates a ratchet effect: the firm cannot commit to long-dated risk positions if it cannot predict the capital charge that will apply in the next regulatory cycle. The leverage ratio, currently hovering near the minimum requirement, leaves no room for the aggressive inventory accumulation that generated the 2009 windfall. If Goldman had attempted to hold the mortgage risk it accumulated in 2008 under current capital rules, it would have breached regulatory minima and faced forced asset sales.
The bear case, then, is not a sudden crisis but a slow bleed of relevance. As markets become more electronic, the bid-ask spreads compress and the $14-16 billion trading revenue becomes $10 billion, then $8 billion. As talent departs for private markets, the compensation ratio rises toward thirty-five cents, compressing the margin. As the consumer business generates losses rather than stable funding, the return on equity drifts toward 10 percent, below the cost of capital. The firm continues to operate, Goldman will not disappear, but it becomes a utility, a regulated entity that provides necessary plumbing for the financial system but generates returns insufficient to justify the risk of the equity. The price-to-book ratio, already depressed, settles permanently below one, creating a value trap where the firm is too cheap to sell but cannot earn its way out of the discount.
This outcome is not predetermined. The firm retains structural advantages that could sustain the model if deployed with precision. The balance sheet, even constrained, remains larger than any non-bank competitor’s. The relationships with sovereign wealth funds and corporate treasuries, built over decades, cannot be replicated by technology platforms. The ability to intermediate complexity, structuring a $5 billion cross-border acquisition financing in forty-eight hours, remains a scarce service. But these advantages are wasting assets if not employed in markets where opacity and complexity still command a premium. The question is whether Goldman can shrink sufficiently in the commoditized businesses, standardized flow trading, vanilla wealth management, mass-market consumer banking, to preserve the capital and the talent for the illiquid, complex, and risky transactions that remain its only defensible territory.
The alternative is a slow convergence with the universal banking model that Goldman has historically disdained: a firm that does everything adequately but nothing exceptionally, that trades at book value because it has become a book-value business, that employs thousands of competent professionals but few irreplaceable ones. The $45 billion year of 2009 represented the last pure expression of the old model, aximum leverage, maximum opacity, maximum risk absorption. Every year since has been a negotiation with regulators, markets, and technology to preserve fragments of that model within a structure increasingly designed to prevent its recurrence. The conviction that Goldman knows something the market does not, the conviction that justifies the inventory, the leverage, and the opacity, requires markets that are sufficiently complex to reward that knowledge and regulators sufficiently permissive to allow the risk-taking that proves it. Both conditions are eroding. The next 150 years may resemble the last only in the name on the door.