Prime Brokerage: The Same Position, Financed by Five Banks at Once

Think of a family that hires one contractor to handle the plumbing, the electrical, the roof and the annual inspection, all under one ongoing account — one relationship instead of four, one point of contact who already knows the house inside out. Two very different things can go wrong with that arrangement. The contractor itself could go bankrupt while holding a deposit on your materials. Or — since nothing stops you from doing this — you could quietly sign the identical contract with four other contractors on four other houses, each one certain it's the only one financing your renovation.

Prime brokerage is a bank playing that contractor role for an institutional client, almost always a hedge fund. Both of those failures have actually happened, thirteen years apart, to two of the best-known names in finance — and they broke the relationship in opposite directions.

What prime brokerage actually bundles

A prime broker is a bank, or the broker-dealer arm of one, that bundles financing, securities lending, custody and capital introduction into a single relationship for one client — instead of that client sourcing each service separately from whichever desk happens to offer it.

The bundle usually holds five pieces:

Piece	What it does	Indian example	Global example
Financing	Lends the client cash against securities as collateral, so a position can be bought with borrowed money	Edelweiss/Nuvama providing margin funding to a domestic Category III AIF running a long-short equity book	Morgan Stanley Prime Brokerage financing a US hedge fund's leveraged long position
Securities lending	Lends out shares the client needs to borrow to sell short, the same function covered in Securities Lending: The Share Sold Short More Than Once	A domestic broker sourcing shares through NSE's SLB window on a fund's behalf	Credit Suisse's prime services desk borrowing shares from its own custody pool to lend to a client
Custody & clearing	Clears the client's trades and holds the resulting securities	A SEBI-registered custodian and clearing member, held under the same banking group but each separately registered, executing and safekeeping trades for a Foreign Portfolio Investor	Goldman Sachs acting as both clearing broker and de facto custodian for a hedge fund's US positions
Capital introduction	Connects the client to prospective investors — family offices, funds of funds, allocators the bank already relationship-manages	A SEBI-registered placement agent introducing a domestic AIF to institutional allocators, kept as a separate function from custody or broking	A prime broker's capital-introduction team hosting an annual investor conference for its hedge fund clients
Reporting	Consolidates positions, P&L and risk across the client's activity at that bank into one feed	A custodian's daily holding statement to an AIF's SEBI-mandated compliance officer	A prime broker's daily risk and margin report to a hedge fund's CFO

One relationship, five services, one client. That's the organizing idea — and it's a genuinely different axis from the two you've already seen split a trading floor. Equities and FICC sort by what's being priced. GECD sorts by what skill it takes to price it. Prime brokerage doesn't sort by product or skill at all — it sorts by client, wrapping several different products around one relationship instead of scattering that client's business across several desks.

Why the bundle exists in the first place

The efficiency is real, not just convenient paperwork. Think of financing five separate positions through five separate accounts, each one demanding its own pile of collateral even while some of those positions are winning and others are losing — versus financing all five through a single account that nets the winners against the losers first, and only asks for fresh collateral on the net exposure left over. That netting, called cross-margining, is the core economic reason a hedge fund wants one prime broker rather than five unrelated lenders: it needs less collateral tied up doing the same amount of trading. Capital introduction adds a second, separate reason — a young fund gets easier access to the bank's own network of institutional allocators, in exchange for routing its financing and trading through that bank. The bundle isn't a marketing wrapper. It's a real subsidy: cheaper leverage and warmer investor introductions, in return for concentrating a fund's business with one counterparty.

The same economics show up on both sides of the India/global split, even though the legal wrapper around them doesn't. A domestic Category III AIF running a long-short book across, say, twenty stocks still wants its winning shorts netted against its losing longs before any lender asks for fresh collateral — that netting doesn't require one bundled prime-brokerage license to work, only a bank willing to look at the fund's book as a whole. Edelweiss or Kotak can offer that same cross-margining benefit to an Indian fund by coordinating across their own separately registered financing and clearing arms internally, even though the fund is legally contracting with more than one entity to get it. The efficiency the bundle is built to capture survives the India split; only the single point of contact for capturing it doesn't.

That last clause — concentrating a fund's business with one counterparty — is also exactly where both of the failures below start.

Two different ways to assemble the same bundle

The US, and most global prime brokers, deliver that bundle from inside one legal umbrella — the client signs largely one prime brokerage agreement, and financing, lending, custody and reporting all flow from the same relationship team, even where distinct legal entities sit behind the scenes for regulatory reasons.

India keeps the pieces apart on purpose. SEBI registers custodians, clearing members and stockbrokers as separate categories, each under its own regulations, even when a single banking group like ICICI or Kotak owns all three. A Foreign Portfolio Investor or a domestic AIF assembling a prime-brokerage-style stack in India is really assembling it from several separately registered entities that happen to share a parent — not from one entity holding one bundled license the way a US client experiences it. It's the same functional-separation instinct covered in Legal Entity: The Only Thing a Contract Can Actually Bind: the relationship can feel like one bundle to the client while the obligations underneath it are still split across distinct legal entities, each bound only for its own piece.

The split carries one concrete, load-bearing consequence: a SEBI-registered custodian holds a client's securities for that client's benefit, not as an asset the custodian is free to re-lend or pledge for its own funding needs. That single design choice is the reason the first of the two failures below couldn't have happened, in quite the same shape, inside the Indian structure.

The first failure: the contractor who could re-lend your materials

Think of handing your house keys to a valet who's contractually allowed to lend your car out to other customers overnight, on the understanding you'll always get an identical car back by morning — right up until the valet company itself goes bankrupt while your car happens to be out with someone else. That's rehypothecation: a prime broker's right, written into most prime brokerage agreements, to re-lend or re-pledge the securities a client posted as collateral, rather than sitting on them idle.

In the US, that right is capped: SEC Rule 15c3-3 limits a broker-dealer to rehypothecating no more than 140% of a client's outstanding debit balance. Lehman Brothers' UK arm, Lehman Brothers International Europe, operated under English law instead, where no equivalent cap applied — and many hedge funds, chasing the extra financing flexibility that unlimited rehypothecation made possible, ran their prime brokerage relationship through the London entity rather than the more restricted New York one. When Lehman collapsed into administration in September 2008, those funds' assets were legally tangled up inside Lehman's own bankruptcy estate rather than sitting safely to one side as segregated client property. Some clients waited years, and recovered only a fraction of what they'd posted, before the UK administration process worked its way through their claims.

The lesson the hedge fund industry took from Lehman wasn't "avoid prime brokerage." It was "don't concentrate your whole relationship — financing, custody and rehypothecation risk together — inside one prime broker." Funds that had used a single prime broker before 2008 spent the years after actively adding second and third prime brokers, splitting financing and custody across multiple banks specifically so no one counterparty's failure could freeze the entire book again.

That fix worked exactly as intended, against exactly the risk it was built for. It also planted the seed of the second failure.

The second failure: five contractors, none comparing notes

Archegos wasn't a hedge fund in the regulatory sense. It was structured as a family office, a status that, under the US Family Office Rule, exempted it from registering as an investment adviser and from most of the position-disclosure obligations that would otherwise apply to a fund its size — a structural choice about which legal entity it counted as, in the same spirit as the segregation-of-duties question the Legal Entity essay closes on. Its founder, Bill Hwang, built enormous concentrated positions in a handful of stocks including ViacomCBS and Discovery — not by buying the shares directly, but through total return swaps, a derivative where the bank holds the actual shares and the client receives only the economic return on them. Because Archegos never legally owned a single share, its stake never crossed the ownership thresholds that trigger a public 13D or 13F disclosure filing — the position was, by design, invisible to the market at large.

Here's where the post-Lehman fix turns into the mechanism of failure. Archegos ran this strategy through prime brokerage relationships at multiple banks at once — Credit Suisse, Nomura, Morgan Stanley, UBS and Goldman Sachs among them, the exact multi-prime-broker diversification the industry had adopted to protect itself after 2008. Each bank, as prime broker, saw only the swaps it was financing for Archegos and margined that exposure on its own book, in isolation. None of them had visibility into what the other four were financing for the same client on the same underlying stocks. Five separate, individually reasonable risk assessments added up to a client leverage picture no single bank could see — reportedly north of $20 billion in swap exposure, built on a much smaller sliver of Archegos's own capital.

When ViacomCBS's share price fell sharply in late March 2021 after a large stock offering, Archegos faced margin calls at every one of its prime brokers simultaneously — the same underlying stock, the same client, five separate demands for collateral it didn't have enough of to meet all at once. Banks that moved fastest to sell the swap collateral, Goldman Sachs and Morgan Stanley among them, largely avoided losses. Credit Suisse, slower to unwind, lost roughly $5.5 billion; Nomura took a loss estimated near $2.9 billion. Nothing about any individual prime brokerage relationship was improperly documented or under-margined. The exposure was hidden by the plainest fact of the structure itself: the bundle exists per-bank, not across banks, and nothing in a standard prime brokerage agreement requires one prime broker to ask whether its client has four others.

It's also, structurally, wrong-way risk at the level of the relationship itself: the very event that made Archegos unable to meet a margin call — the stock price collapsing — was the same event that made the collateral backing that margin call worth less, at every bank, at the same time.

Where India's split lands on each failure — and where it doesn't

Run both failures back through the Indian structure and they land differently.

Against a Lehman-style failure, India's split genuinely helps: because a SEBI-registered custodian holds a client's securities for that client's benefit rather than as an asset the custodian's own balance sheet is free to re-lend, an Indian institutional client's holdings aren't exposed to the specific mechanism that trapped funds inside Lehman's UK estate — there's no equivalent of an unlimited-rehypothecation clause sitting inside a custody relationship that's, by regulation, a distinct entity from the bank providing financing.

Against an Archegos-style failure, the same split does almost nothing. Custody, clearing and financing being separately registered entities doesn't give any Indian regulator, or any Indian bank, a consolidated view of a client's total leverage across every prime-brokerage-style relationship it holds elsewhere. Concentration risk in an Indian institutional client's book is, by construction, visible to more than one regulated party rather than sitting entirely inside one bank's private view — but no single party currently aggregates that client's total exposure across every bank it deals with. The Indian structure was built to answer 2008's question. It was never built to answer 2021's.

Why this matters for a Business Analyst

Go back to the two ways the contractor arrangement can break: the contractor who goes bankrupt holding your deposit, and the client running the identical arrangement with four other contractors at once.

A risk model built only around the first failure — checking that a prime broker's custody arm properly segregates client assets, that rehypothecation stays inside its legal limit — is a genuinely necessary model, and it's the one most compliance checklists are built to catch, because it's the one regulators wrote explicit rules about after 2008. It is also, on its own, silent about the second failure entirely. A counterparty risk model, a margin engine or an exposure report scoped around "does this desk correctly track what it finances for this client" will pass every test, every day, for a client who is quietly running the same trade at four other banks — because nothing in that scope asks the question. A BA designing either system has to decide, explicitly, which of the two failures the system is meant to catch, because a model built to catch one gives no protection against the other, and a stakeholder who only remembers "we fixed prime brokerage risk after 2008" may not realize which half of the problem that actually covers.

The Indian custodian split reinforces the same discipline from the opposite direction: it's evidence that a genuinely well-designed structural safeguard against one specific, well-understood failure mode can still leave a completely different failure mode untouched. Solving for the last crisis is not the same as solving for the next one, and a BA reviewing "what controls do we have here" should always ask which specific failure a given control was built to catch — not assume that a control addressing a risk in this business line addresses the risk in it.

Lighthouse Insight

Back to the five contractors, each one certain it's the only one on the job — and back one step further, to the single contractor who went bankrupt holding a deposit nobody could get back.

The industry solved the first problem by teaching every client to spread the exact same relationship across several banks instead of one. That fix was correct, on its own terms, against the risk it targeted. It also quietly built the second problem into the market's structure: a bundle, by design, visible to only one bank at a time, now deliberately multiplied across several. Archegos didn't break prime brokerage. It stood on the industry's own post-2008 lesson and used it exactly as taught, five times over, until the collateral backing all five bundles fell on the same afternoon.