How Investment Banks Actually Make Their Money
Goldman Sachs made $53B in 2024. Most of it wasn't from trading. Here's how investment banks really generate revenue—and who pays for it.
Written by AI. Jin Seo

Photo: AI. Ines Cienfuegos
Goldman Sachs generated over $53 billion in revenue in 2024. If your first instinct is to picture trading floors and flashing screens, you're thinking about the wrong business. Less than half of that number came from trading. The steadier, larger piece came from advising on corporate deals and managing wealthy clients' money — activities that sound considerably less cinematic but pay considerably better.
That gap between the popular image of investment banking and its actual economics is what a recent Martik Finance video sets out to close, and it's a gap worth taking seriously. Because how these firms actually make money tells you something important about market structure, competitive concentration, and who ultimately absorbs the cost.
The Four Engines
The Martik Finance breakdown identifies four revenue streams: M&A advisory, underwriting, trading and market making, and asset and wealth management. Each works differently. Each has a different risk profile. And understanding the differences matters more than most financial explainers let on.
M&A advisory is the one that gets banker stereotypes going. When one company buys another, an investment bank gets hired to manage the whole process — valuing the target, structuring the deal, navigating tax and legal land mines, and keeping both sides from walking away during the months of back-and-forth that major acquisitions require. The fee is a percentage of deal size, and the percentage compresses as the deal grows. For mega-deals above a billion dollars, that might be roughly 0.5%. On a $10 billion acquisition, 0.75% works out to $75 million — for a team that might number a few dozen people, over six to twelve months.
What the video gets right, and what deserves more emphasis than it gets, is that this is structurally self-reinforcing in a way that looks a lot like oligopoly. The video notes it almost in passing: "Once a bank has advised a company on one major deal, it usually gets the call for the next one. That's why the same handful of names show up again and again on the biggest deals in the world." According to Reuters, Goldman Sachs topped the global M&A league tables with $1.48 trillion in deals in 2024 — and that dominance isn't accidental or purely meritocratic. It's compounding. A track record of closed deals generates access to the next deal, which generates the next track record. New entrants don't just need capital and talent; they need decades of closed deals they don't have yet. That's not a market you disrupt with a lower fee schedule.
Underwriting is where the risk becomes tangible. When a company goes public or issues bonds, the bank doesn't just introduce buyer to seller. It often buys the securities from the company first, at an agreed price, then resells them to investors. If demand is strong, the bank pockets the spread. If demand evaporates — if the market turns between pricing and distribution — the bank sits on securities nobody wants. That's real balance-sheet exposure, not advisory risk.
The video uses a fabricated IPO example to illustrate scale, but the real market provides plenty. Arm Holdings' 2023 IPO raised approximately $4.9 billion on the Nasdaq, with a syndicate of banks led by Barclays, Goldman Sachs, JPMorgan, and Mizuho sharing underwriting duties. Typical IPO underwriting fees run around 3% of proceeds. On a deal that size, the syndicate collectively earned roughly $147 million — for absorbing pricing risk and placing shares with institutional investors who trust them to deliver quality deal flow.
That last phrase is the key one.
Trading and market making is the business that gets misread most often. The video's airport currency exchange analogy is apt: "You're not predicting whether the dollar rises tomorrow. You're just always ready to buy or sell, and you profit from the small gap between the buying price and the selling price repeated thousands of times a day." Modern investment banks are not, for the most part, making directional bets with house money. The Volcker Rule — passed as part of the 2010 Dodd-Frank Act in response to the 2008 financial crisis — sharply constrained banks' ability to make speculative proprietary trades in the United States. What they do instead is provide liquidity: when a pension fund needs to move $500 million of stock without cratering the price, a bank's trading desk absorbs that flow, profiting from spread and service fees rather than market prediction. Goldman's global banking and markets division, which covers trading, market making, and investment banking fees, generated nearly $35 billion in 2024, with record equities performance.
Asset and wealth management is the one the video correctly identifies as the "smartest model" structurally — and the reason is simple: it's a subscription. Banks managing money for wealthy individuals, pension funds, and institutions typically charge 0.5% to 1% of assets annually, regardless of market direction. M&A fees require closing a deal. Underwriting fees require a live capital raise. Wealth management fees arrive every year whether anything happens or not. Goldman's asset and wealth management division brought in over $16 billion in 2024, against a managed asset base that had grown to over $3 trillion.
The Fee Question
Here's what the video frames as a justification and I'd frame as an interesting tension: the four factors that explain why investment banks command the fees they do — relationships, distribution, balance sheet, and trust — are also precisely the factors that prevent competition from driving those fees down.
The video puts it directly: "Put those four together, relationships, distribution, balance sheet, and trust, and you get a business that's almost impossible to disrupt with a clever app or a smaller fee. That's the real moat."
That's accurate. It's also worth sitting with. These aren't moats that exist because investment banks provide uniquely efficient service. They're moats that exist because the business runs on information asymmetry, access, and accumulated credibility that takes decades to build. A CEO choosing an advisor for a $10 billion acquisition isn't running a competitive bidding process on fee percentage. They're calling the banker who has been in the room for their last three deals, knows their board dynamics, and has never leaked a price-sensitive conversation to the press.
That's not a criticism of how investment banks operate. It's a description of how trust-based oligopolies sustain themselves. The fees reflect the concentration of capability — which is itself a product of the concentration of past fees funding the talent, capital, and relationships required to do the next deal. The circle is fairly tight.
What Investment Banks Are Not
One thing the video does genuinely useful work on is taxonomy. "Investment bank," "hedge fund," and "commercial bank" get conflated constantly, and the conflation distorts most popular criticism of the financial sector.
A commercial bank takes deposits and makes loans. A hedge fund takes capital from wealthy investors and institutions and makes active market bets to generate returns. An investment bank — at its core — is a facilitator: it advises on deals, raises capital, provides liquidity, and manages money for institutions and high-net-worth clients. Some firms, like JPMorgan, operate all three models under one roof, which adds to the confusion. But the business models carry different risks, different regulatory frameworks, and different relationships to public money.
When public criticism lands on "banks gambling with deposits," it's usually describing either hedge fund behavior or pre-2008 proprietary trading practices. The latter were significantly curtailed by Dodd-Frank. The former is a different industry entirely.
The distinction matters not because investment banks deserve protection from scrutiny — they don't — but because accurate criticism requires accurate framing. The more interesting question isn't whether Goldman Sachs trades too aggressively. It's whether a business model that compounds competitive advantage so efficiently, where every closed deal makes the next deal easier to win, produces market outcomes that serve anyone beyond the firms themselves.
That question the Martik Finance video doesn't pursue. Someone should.
Jin Seo covers business, finance, and economic policy for BuzzRAG.
We Watch Tech YouTube So You Don't Have To
Get the week's best tech insights, summarized and delivered to your inbox. No fluff, no spam.
More Like This
What Fintech Affiliate Models Could Teach iGaming
Line Peteri argues fintech's education-first affiliate model and retention focus offer a playbook iGaming has ignored for a decade. Here's what that gap actually costs.
Rockefeller's Wealth-Building Secrets Revealed
Explore how John D. Rockefeller built a $400 billion fortune through strategic discipline and overlooked industries.
America's Economic Cycles: Reinvention or Stagnation?
Examining America's economic pivots every 80-90 years and their impact.
Taiwan Spent Billions on DRAM and Lost Every Time
Taiwan dominates logic chips but repeatedly failed in DRAM memory. Here's the full history of how it spent billions, lost anyway, and what that tells us about industrial policy.
Investment Banking Interviews Demand 8-Hour Days of Prep
An investment banking recruiter breaks down the reality: 30 emails a night, 2% response rates, and behavioral questions harder than technical ones.
Apple Card's Chase Transition: What Actually Changes
Goldman Sachs is out, Chase is in for $19B. But the Apple Card's real problem isn't who issues it—it's whether it can evolve beyond its 2019 vision.
The Iceberg Index: Rethinking AI's Economic Impact
MIT's Iceberg Index reveals AI's hidden economic impact, challenging job-centric narratives.
Otter.ai's Voice AI Vision: A 10-Year Bet
Sam Liang's Otter.ai bets on voice tech as the future, boasting 35M users and $100M ARR. Will voice replace typing?
RAG·vector embedding
2026-08-16This article is indexed as a 1536-dimensional vector for semantic retrieval. Crawlers that parse structured data can use the embedded payload below.