A friend who runs a small systematic fund asked me why his crypto book needed something like four times the cash his equities book did to carry what was, on paper, the same amount of risk. The answer is boring and structural. In equities his prime broker lets him trade a dozen venues out of one margin account. In crypto, every exchange he touches wants its own pile of collateral sitting there before he can place a single order, so the same strategy at similar risk ties up several times the capital. That gap explains more about why institutional crypto stays smaller than the headlines suggest than any licensing question does, and it is worth understanding properly.
It helps to be precise about what a prime broker does in traditional markets, because the term gets thrown around loosely in crypto. An equity prime does a few things that matter here. It custodies your assets. It finances your positions against the whole portfolio, netted, so a long in one name offsets a short in a correlated one and margin is computed on the residual risk rather than the gross. It intermediates execution, meaning you can trade through any broker or venue you like and everything settles back into that single account. And it lends you securities so you can short. Stack those together and a fund posts one pool of collateral and gets access to essentially the entire market. The capital efficiency is enormous, and it is invisible until you work somewhere that does not have it.
Crypto grew up the opposite way. Each exchange is its own vertically integrated stack, exchange plus custodian plus clearinghouse plus margin lender, all inside one company. There is no shared clearing layer behind the venues, so the only way a venue can be sure you will pay for what you buy is to hold your money before you trade, which is where prefunding comes from. If you run a strategy across five exchanges, you keep five separate piles of collateral, each sized for the worst day that venue alone might see, and none of them offset each other. A hedged book that would carry modest margin at an equity prime can eat multiples of that in crypto, because every leg gets margined in isolation and a chunk of your cash sits stranded at venues you are not even using that day.
Prefunding also concentrates counterparty risk in exactly the wrong place. Your collateral is committed, and it is committed to the balance sheet of a trading venue. When FTX failed, the funds that lost the most were often the ones running the most sophisticated cross-venue strategies, because those strategies require leaving capital everywhere.
What crypto primes actually offer today
Credit intermediation is the core product. Instead of facing each exchange yourself, you face the prime, and the prime faces the venues. You post collateral once, with the prime, and it extends you trading limits across the exchanges it has relationships with. The prime bears the venue default risk and charges you for that through fees and financing spreads. This is a genuine improvement over raw prefunding. It also means the quality of your protection is exactly the quality of the prime's own credit and its own agreements with each venue, which is worth reading the actual documents about rather than assuming.
Off-exchange settlement is the second piece, and I think the more important one. In these arrangements your collateral sits with a custodian, often in a tri-party structure, and the exchange extends you a trading limit against assets it can verify but does not hold. Trades accumulate and then settle on a net basis on some cycle, hourly or daily depending on the setup. Your exposure window to the venue shrinks from all the time to the gap between settlements. Several large exchanges have come around to supporting versions of this because institutional clients refused to onboard without it.
Cross-venue margining is the third piece and the least mature. A few primes will look at your positions across venues and margin the net, the way an equity prime would. But the netting happens on the prime's own books, not at a clearinghouse, so the benefit is capped by the prime's balance sheet and risk appetite. In practice you get partial offsets with conservative haircuts rather than true portfolio margin.
Where the model still breaks
The missing layer is central clearing. In equities the prime is backstopped by a clearinghouse, and the clearinghouse is backstopped by a mutualized default fund that every member pays into, so a single failure gets absorbed by a very large pool. In crypto the prime's own equity is the backstop, and crypto primes are tiny next to bank prime brokerages. That makes credit expensive, shallow, and quick to disappear. I am not sure anyone fixes this without building something that looks a lot like a clearinghouse, which would require the big venues to give up vertical integration, and prefunded customer float is a wonderful business if you can keep it, so they are in no hurry.
The 24/7 market makes it harder still. Traditional settlement runs on batch cycles with banking hours underneath. Crypto trades through the weekend while the fiat rails sleep, so any settlement network has to handle the Saturday problem, meaning a venue can get into trouble at a moment when nobody can move dollars. Stablecoins patch part of that, but then your treasury operation becomes the bottleneck, and shuffling collateral between chains and custodians in the middle of a Sunday night is its own category of operational risk.
Underneath everything sits the underwriting problem. Extending credit to a client who trades on an offshore venue means underwriting the venue too, and exchange balance sheets remain mostly opaque. Every credit officer in the space prices that opacity in, and it shows up as lower limits and wider spreads than the same client would get in any traditional asset class.
How I would evaluate a crypto prime
If you are thinking about trading through one, the questions that matter are mostly about failure states rather than features, because the pitch decks all look the same and the differences live in the legal agreements.
- Where does my collateral physically sit, who controls the keys, and is it segregated or sitting on someone's balance sheet.
- If a venue fails between settlement cycles, who eats the loss, me or the prime, and is the answer in the agreement or only in the marketing.
- What rehypothecation rights does the prime have over my assets, and under what conditions do they expand.
- How is cross-venue margin computed, and can haircuts or limits change unilaterally, and with how much notice.
- What actually happened to client credit lines during the last serious drawdown, as opposed to what the policy document says should happen.
That last question is the one I would weight most heavily, because it points at the nastiest failure mode in the whole structure. Credit in crypto is procyclical in a way that equity prime credit mostly is not. When volatility spikes, the prime's own exposure to venues and clients balloons, so it cuts limits, and it cuts them at precisely the moment your strategy needs the line most, because dislocations are widest when everything is on fire. If a strategy only pencils out at your fair-weather credit limit, it does not pencil out. Size to the line you would realistically have on the worst day and treat everything above that as a bonus.
The broader point is about what to watch. Institutional adoption gets framed as a regulatory story, and licensing does matter at the margin. But plenty of institutions already hold every approval they need and still run crypto books at a fraction of the size their mandates would allow, because the return on capital after prefunding drag does not compete with what the same dollar earns elsewhere. What changes that is plumbing, settlement networks getting adopted, netting getting broader, credit getting cheaper as the backstops get stronger. None of it produces exciting announcements, so it mostly goes unnoticed. If you want to gauge how far along it really is, put the five questions above to a prime and count how many get answered from the documents instead of the deck.