Four Structures, Four Power Systems: The Internal Politics of Banks, Hedge Funds, Prop Shops, and HFT
Why Jane Street's $39.6bn beats Goldman on a per-head basis, how Millennium's pod model can fire a team on a drawdown limit, and what the $100m talent war means for where the strongest quants end up.
Non-bank trading firms booked a combined $114bn in 2025, a rise of 45% over the prior year, with the proprietary-trading portion climbing roughly 60% to $84.3bn while market-making added 20% to $30.2bn. Banks still ran the larger pool at $260.7bn, up 13%.
Those figures usually get read as a market-share story, and they are. They are also an organizational story about who holds power inside each type of firm, because the money tells you exactly where authority sits.
Every one of these structures is competing for the same hundred or so people capable of generating alpha at scale, and each one runs a different internal political system. Two variables decide the politics in all four cases. The first is who controls capital allocation.
The second is who absorbs the loss when a position goes against the book. Once you know how a firm answers those two questions, its compensation design, its firing mechanics, its non-compete length, and the loyalty it can command all follow in a predictable way. What looks like four pay tiers from the outside is really four separate machines for deciding who eats and who leaves.
The bank: the franchise owns the profit and the trader rents the chair
Goldman Sachs ran $31.1bn through its trading operation in 2025 and still lost the efficiency contest. Jane Street cleared $39.6bn with roughly 3,500 people, and JPMorgan, whose trading headcount runs into the tens of thousands across equities and fixed income, came in at $35.8bn. The bank concentrates enormous capital and then distributes credit for it across a committee, and that committee is where the real power lives.
The Volcker reforms pulled proprietary risk-taking off bank balance sheets after 2008, and the desks that wanted to keep eating their own profit and loss left to form hedge funds. What stayed behind shifted toward structuring and executing complex institutional flow, while the independent firms took over retail wholesaling, exchange-traded products, and high-frequency market-making.
The result is that a bank trader no longer owns the risk on the screen in front of them. When the institution holds the balance sheet, the institution captures the marginal dollar of alpha, and the trader ends up with a bonus-pool number decided two levels above their seat.
The pay design hardens that arrangement rather than softening it. A first-year analyst earns a base around $105k to $110k with a bonus averaging 60% to 70% of that base, well below the levels of the 2021 cycle. A growing share of associate compensation arrives deferred, vesting across two to three years so the firm can claw it back if the person walks early.
The up-or-out clock is genuine, and a vice president who fails to reach executive director inside roughly four years can forfeit unvested deferred compensation entirely. Authority therefore belongs to whoever manages the franchise relationship and sits on the compensation committee, rather than to whoever posted the highest desk number that year.
The exits show the pressure that creates. Paulo Costa, twenty-nine years old and freshly promoted to managing director at Goldman, left within months for Millennium in London to run a dividend-futures book, with an inducement reported above $20m. Mark Gibbs left Citi for Brevan Howard back in 2019, and the quant-equities team built around him absorbed the disruption that followed. A structure that keeps the upside for itself trains its most capable people to leave for a structure that pays them for the risk they already know how to carry.
For full access to daily quant articles (12.5$ a month):
The multi-manager platform: the clearest political economy on the buy side
Millennium, Citadel, Point72, and Balyasny run the pod model, and it is the most transparent power system on the buy side precisely because it removes the soft variables and replaces them with two hard ones, capital and drawdown. Pass-through fees mean the limited partners cover essentially every cost the platform incurs, compensation included, so the platform’s own economics depend on never letting a single pod open a hole in the book.
That arrangement hands the risk officer more day-to-day control over a portfolio manager’s survival than the portfolio manager’s own returns provide.
A pod that breaches its drawdown limit, frequently set in the range of five to seven and a half percent peak to trough, has its capital pulled, and the team can be cut in the middle of the year regardless of where the wider fund sits. Survival at a pod shop is governed by the worst month, not the best year, and the people who last are the ones who treat the risk limit as the actual product.
Capital allocation is the currency that decides standing, and it is stated openly. Pay reduces to a formula of gross market value multiplied by percent return multiplied by percent cut. Base salary settles around $250k and barely moves across an entire career, so the meaningful money lives in the cut, usually a negotiated 10% to 20% of net profit and loss, and in the size of the book a manager is trusted to run.
