What is the best way to settle a large crypto OTC trade?

A large crypto OTC trade is settled in one of seven ways: one side goes first, the trade is split into tranches, a neutral third party holds one or both legs, both legs move inside one custodian or exchange, a tri-party arrangement links two custodians through a settlement network, both legs swap in a single onchain transaction, or a hashed time-locked exchange links two chains. The criterion that separates them is whether the two legs are tied together so that one cannot complete without the other, and who or what you have to trust for that tie to hold. No method is best on every count: the right one depends on where each asset sits, whether a bank transfer is involved, how well you know the counterparty and how much time and paperwork the trade can bear.

An eighth tool belongs in the picture for a different job. An onchain payment router divides one payment among several parties, which is how a broker's or an introducer's fee can be paid out of one leg at the moment that leg is paid. A payment router does not exchange two assets against each other, so it is compared below for the payout only.

The comparison uses six criteria: who is exposed to whom and for how long, what must be trusted, speed, assets and chains covered, minimum practical size, and the record left. Rules on custody, escrow and money transmission vary by country, so the legal status of each arrangement is a question for counsel in the jurisdictions of the two parties.

What are the ways to settle a large crypto OTC trade?

There are seven families of methods for exchanging the two legs of an agreed OTC trade, and they differ in whether a party, an institution or a piece of code stands between the first transfer and the second. Once price and size are agreed, one party owes an asset and the other owes payment, and whoever delivers first carries the full value of the trade until the other leg arrives. Standard-setters call that principal risk, the risk that a counterparty loses the full value involved in a transaction, and they call the cure delivery versus payment: a mechanism that links the two transfers so that one happens only if the other does. The Committee on Payment and Settlement Systems at the Bank for International Settlements framed the concept for securities in a 1992 report. Every method below accepts principal risk, shrinks it, or removes it by linking the legs.

Which assets, chains and trade sizes does each method fit?

Each settlement method works only where both assets can reach it: an atomic swap needs both tokens on the same chain, a book transfer needs both parties at the same custodian, and only the methods with a person or an institution in the middle can handle a bank transfer on one side. That constraint usually removes several options before any preference comes into play.

Two points in this table decide many trades. Native bitcoin lives on the Bitcoin network and most dollar stablecoins live on smart-contract chains, so a stablecoin-for-bitcoin trade is a cross-chain trade unless the bitcoin leg is delivered as a wrapped bitcoin token, which is a claim on bitcoin held by a custodian or locked by a protocol. And none of the minimum sizes above is a published threshold: where a method has a floor, the floor comes from fixed fees, legal drafting and onboarding checks, which weigh less as the trade grows.

What does each method expose you to on a $3,000,000 trade?

On a $3,000,000 stablecoin-for-bitcoin trade, the side that goes first carries $3,000,000 of counterparty exposure, six tranches cut that to $500,000 at a time, and the linked methods bring counterparty exposure on the principal to $0 while moving the exposure to an agent, a custodian or contract code. The assumptions of this worked example are the following. A fund buys 30 bitcoin from a seller at an illustrative price of $100,000 each and pays $3,000,000 in a dollar stablecoin. An introducer brought the two parties together for a stated fee of 0.50% of the trade, $15,000, which the seller has agreed to bear, so the seller nets $2,985,000. The durations are illustrative: a transfer cycle is taken as 1 hour 10 minutes, made of six bitcoin confirmations at an average of ten minutes per block plus ten minutes of handling.

Loss of principal is not the only cost. Under tranches, the trade stays open for 7 hours instead of 1 hour 10 minutes, so the price agreed at the start has to be honoured across six rounds while the market moves. Under a hashed time-locked exchange, nobody can lose the principal, yet one party can decline to complete and leave the other with funds locked until a refund time: with the illustrative locks of 12 and 24 hours, a 2% move in bitcoin over that window is worth $60,000 on this trade. Under the institutional methods, the question becomes how a balance is held and protected at the institution, which is a due-diligence question with documents to read.

When does going first or splitting into tranches make sense?

Going first makes sense when the party that receives first is a desk or counterparty whose standing you have already checked and are willing to rely on for the full amount, and tranches make sense when you are not, and no linked method is available. With an established desk, pre-funding is routine: the client sends its leg, the desk confirms receipt and delivers, and the client's protection is the desk's regulatory status, balance sheet, track record and terms of business. Its limit is plain: for the length of one transfer cycle, the first mover is an unsecured creditor of the receiver for the whole trade.

Tranches keep the same mechanics and reduce the amount at stake. The parties agree a tranche size, one side sends the first tranche, the other answers, and the sequence repeats. Exposure falls in proportion to the number of tranches, and the time the trade stays open rises in the same proportion.

In practice the tranche count is chosen so that the amount at risk is one the first mover could absorb, and the trade still closes inside a single session. Two points deserve a line in the trade terms: whether the price is fixed for all tranches or reset for each one, and what happens to the unsettled balance if one side stops midway.

What does an escrow agent, a shared custodian or a tri-party arrangement change?

An escrow agent, a shared custodian or a tri-party arrangement replaces exposure to the counterparty with exposure to a regulated or professionally accountable institution, which is usually the stronger and better-documented credit. The three differ in where the assets sit before the trade and in how much has to be set up.

A neutral third party, an escrow agent or a law firm acting under a written escrow agreement, receives one or both legs and releases them according to instructions both sides have signed. When the agent holds both legs, neither party is exposed to the other at any moment: the stablecoin and the bitcoin each arrive, the agent checks the amounts, and each party receives what it is owed. When the agent holds one leg only, typically the payment, the seller delivers once the deposit is confirmed and the agent releases on proof of delivery. This method reaches furthest: it works across chains and with a bank transfer on one leg, and it produces the fullest file. Onboarding and drafting take time, and each agent sets its own fee and minimums, so it suits a trade whose size justifies them and a first trade between parties with no history.

Settlement inside one custodian or exchange applies when both parties already hold accounts at the same institution. Nothing moves onchain: the custodian debits the stablecoin from one account and the bitcoin from the other and credits each to the opposite side on its own books. It is fast and avoids network fees and confirmation waits. The point to confirm is whether the institution books the two legs as one linked settlement or as two independent internal transfers that each client instructs separately. Only the first is delivery versus payment; the second is one side going first with a shorter wait.

A tri-party arrangement serves parties, or a party and a trading venue, that do not share a custodian. Assets stay in custody with each party's own custodian, balances are made visible to the other side for trading, and a settlement network connected to the custodians moves both legs after the trade. Custodians that operate such networks describe the benefit as lower counterparty risk, because assets stay segregated in custody until settlement, and less pre-funding. Every participant must be onboarded with a member custodian, so the route pays off across many trades more than on a single one.

How do atomic swaps and hashed time-locked exchanges work?

An atomic swap and a hashed time-locked exchange both tie the two legs together with code so that either both transfers happen or neither does, the first inside one transaction on one chain and the second across two chains through a shared secret. Trust moves to the contract logic and to each party's own operations.

In an atomic swap, both tokens are on the same smart-contract chain. Each party authorises a swap contract to move its token, and a single transaction transfers the stablecoin one way and the other token the opposite way. If either transfer cannot execute, the whole transaction reverts and both parties keep what they had. Exposure on the principal is zero for both sides. On Ethereum, the developer documentation published by the Ethereum Foundation gives slots of 12 seconds and epochs of 32 slots, which is 6.4 minutes per epoch, with finality declared at epoch checkpoints, so a cautious desk waits a quarter of an hour or so before treating a large settlement as final. The limits are the contract itself, which should be an audited and widely used one, and the scope: for the trade in the example, the bitcoin leg would have to be 30 units of a wrapped bitcoin token, and the buyer would then rely on that token's backing or redeem it for native bitcoin as a further step.

A hashed time-locked exchange links two separate chains. One party generates a secret and publishes only its hash. That party locks its asset on the first chain under two conditions: the counterparty can claim it by presenting the secret, or the owner can take it back after a time limit. The counterparty checks the lock, then locks its own asset on the second chain under the same hash with a shorter time limit. The first party claims the second asset, and doing so reveals the secret onchain; the counterparty uses the revealed secret to claim the first asset. If the exchange stalls, each side takes back its own asset once its time limit passes. The Bitcoin improvement proposal that standardises these scripts describes exactly this pair of conditions: a designated party spends by disclosing the preimage of a hash, and the funder recovers the funds after a timeout.

In the example, the seller would lock 30 bitcoin on Bitcoin with a 24-hour refund time, the fund would lock $3,000,000 of stablecoin in a contract on the stablecoin's chain with a 12-hour refund time, and the two claims would follow. Two limits matter at this size. First, the party holding the secret decides whether to complete, and can watch the market until the shorter time limit approaches. A 2019 paper by Han, Lin and Yu showed that such a swap is equivalent to an American call option granted without a premium, and proposed adding a premium to make it fair. Second, each party must be online and must claim well before its deadline, with enough margin for network congestion. A hashed time-locked exchange therefore suits technically equipped parties who accept the lock-up and the optionality, or who price them into the trade.

How is the introducer's fee paid under each method?

The introducer's fee is either forwarded by one of the parties after settlement or paid at source as a division of one leg, and each settlement method allows one or the other. In the worked example the fee is $15,000, 0.50% of $3,000,000, borne by the seller out of the stablecoin leg, which leaves the seller $2,985,000. Where the fee is forwarded, the introducer carries a small exposure of its own: $15,000 owed by a party that has already been paid in full.

• One side goes first. The seller receives $3,000,000 and sends $15,000 to the introducer as a separate transfer, or the buyer is instructed to make two transfers of $2,985,000 and $15,000.
• Six tranches. The fee follows each round: $2,500 per $500,000 tranche, leaving the seller $497,500 each time, and $15,000 in total after the sixth.
• Escrow agent or law firm. The release instruction names two payees for the stablecoin leg, $2,985,000 to the seller and $15,000 to the introducer, and the agent pays both on release.
• Book transfer at one custodian. The fee is a third internal transfer if the introducer holds an account there, or a withdrawal to the introducer's address after settlement.
• Tri-party with a settlement network. The network settles the two legs between the trading parties; the fee is normally invoiced and paid separately.
• Atomic swap. A swap contract may be written to send part of one leg to a third address in the same transaction; a standard two-party swap is followed by a separate transfer.
• Hashed time-locked exchange. The locks pay one claimant each, so the fee is a separate transfer once the seller has claimed the stablecoin.
• Onchain payment router. The stablecoin leg is paid once and divided at that moment: 99.5% to the seller and 0.5% to the introducer.

The amounts are the same under every method and sum to $3,000,000. What changes is whether the introducer depends on a second action by a party already paid in full.

Where does an onchain payment router fit, and where does it not?

An onchain payment router fits the payout of one leg among several parties, and it does not fit the exchange of the two legs. A router takes one payment in one token and pays each recipient a share that was fixed in advance, in a single transaction. The seller, the broker and the introducer each receive their part of the dollar leg without one of them forwarding money to the others. The bitcoin still has to be delivered against that payment by one of the seven methods above.

Shaka is one such router. Each party's share is set as a percentage of the deal, each recipient can sign their share before the payer pays, and signing costs the parties nothing. The payer pays once, through a payment link, and every party is paid their share at the same moment in one transaction; Shaka routes the payment and never holds the funds. Destinations are fixed and signed before the payment, so the money cannot be redirected at the last minute. The payment is onchain and final, every party can see that the others were paid, and a payment record can be kept for the file. A deal has at most 20 recipients. In the example, the stablecoin leg would be set at 99.5% for the seller and 0.5% for the introducer, which pays $2,985,000 and $15,000 together.

Where it sits in a trade follows from the callout. It is useful on the leg that is paid second, for instance when the seller delivers the bitcoin first to a buyer it knows, or tranche by tranche, and the dollar leg has several beneficiaries: a desk, a broker, an introducer, or several entities on the selling side. It is not a substitute for an escrow agent, a custodian or a swap contract, it does not hold funds against a condition, and it has no role when the dollar leg is a bank transfer.

How do you choose a settlement method for your situation?

The choice of settlement method depends on four facts: whether the two assets can meet on one chain or one set of books, whether a bank transfer is involved, how well the parties know each other, and how much setup time the trade can bear. Start with what is technically possible, then pick the residual exposure you can document and accept.

For a bank transfer on one side, the timing is set by banking hours and by the finality rules of the payment system used, which differ by country and by system, so the crypto leg is usually released only once the receiving bank confirms good funds.