Crypto’s off-exchange custody test: who owes what between settlements?

Collateral can remain outside an exchange while a claim against that exchange grows. The reverse is also possible: an institution can keep assets with a custodian while accumulating trading losses it must pay. Custody answers where assets are held. Settlement answers when obligations are discharged.

That distinction matters more than the network label in Bitget’s November 5 announcement of its Open Custody Network. The company describes asset mirroring that lets eligible users access spot, margin and futures markets while underlying assets remain with participating external custodians or banks. Those are company-described capabilities, not independently tested outcomes.

The revealing detail is the settlement clock: four-hour, eight-hour, daily and loss-triggered arrangements. Each changes how long obligations may remain outstanding and when somebody needs liquidity. None, by itself, establishes which arrangement is safer.

Follow the assets, then the obligations

The basic flow has three stages. First, eligible collateral sits with an external provider. Second, the exchange recognises trading capacity against that collateral through the mirroring arrangement. Third, trading generates obligations that must be reconciled and settled under the applicable agreements.

Mirrored trading capacity is not a second pool of underlying assets. Nor should an exchange balance be confused with money already received into external custody. For a profitable trader, an unsettled gain can represent a receivable whose payment still depends on the settlement process. For a losing trader, externally held collateral does not extinguish the obligation to pay.

The precise payer, recipient, transfer authority and treatment of balances depend on the contract and product. Spot purchases, margin borrowing and futures profits and losses do not generate identical obligations. A general description of asset mirroring cannot establish every legal or operational step.

There is also a chronology worth preserving. Bitget’s explanatory documentation displays October 15, preceding the November announcement. Sygnum announced its Bitget integration with Protect on September 24. November’s network announcement should not be read as the starting date of every underlying relationship.

Four hours, eight hours, one day

The October-dated Bitget documentation describes configurable four-hour settlement for Copper, eight-hour settlement for Cactus, and daily settlement for Fireblocks and Bitfire. These are documented arrangements, not evidence of observed completion times.

Four-hour settlement can reduce the scheduled interval over which obligations accumulate, assuming each cycle completes and comparable trading continues. The corresponding operational demand is frequent access to transferable liquidity, reconciliation and payment authorisation. A shorter cycle offers little comfort if a disputed balance or failed transfer carries exposure into the next one.

Eight-hour settlement allows a longer interval between scheduled discharges. Under otherwise identical conditions, more market movement can occur before the next cycle. It also gives a payer more time between funding events. Neither point establishes that obligations will be twice as large as under a four-hour arrangement: positions, prices and permitted netting determine the amount.

Daily settlement concentrates scheduled payment activity into fewer events. Where agreements permit netting, offsetting gains and losses within the period may reduce the final transfer. But a favourable end-of-day balance can conceal a much larger intraday obligation. Daily settlement can leave a recipient waiting longer for funds while requiring a payer to meet a larger single payment if losses persist.

Consider a hypothetical trader whose loss rises steadily after a completed settlement. More frequent successful settlement would discharge portions earlier; daily settlement would defer payment to the daily cycle, absent other controls. If losses subsequently reverse, earlier settlement might instead require cash that a later net calculation would not. This is a liquidity trade-off, not a forecast about any provider.

Importantly, the interval between scheduled cycles is not necessarily the time to settlement finality. Calculation, agreement, authorisation and completion may be separate steps. The relevant exposure window ends when the obligation is discharged—not when the clock reaches the scheduled hour.

A loss trigger is a different kind of clock

Bitget documents an OSL arrangement in which loss settlement is triggered when losses reach at least 15% of custodial assets, rather than following a scheduled cycle.

That is an event-based rule. It does not provide the same elapsed-time bound as periodic settlement. Below the threshold, the loss-trigger provision alone does not tell a reader when an obligation will be discharged. During a sharp move, the threshold could be reached quickly.

Nor is 15% a guaranteed ceiling on losses or unsettled exposure. Detection and payment take time; prices can move while a transfer is being arranged. The disclosed trigger also should not be extrapolated into an identical rule for paying profits.

The incentive differs from a fixed schedule. A loss trigger links payment activity to deterioration in the account rather than the passage of time. That may avoid some routine transfers, but it can demand liquidity precisely when markets are difficult. Whether that trade-off works depends on monitoring, collateral valuation, transfer authority and the response after a trigger—details a percentage alone cannot settle.

Collateral value is not payment liquidity

Sygnum’s September announcement lists bitcoin, ether, stablecoins and US Treasuries among eligible collateral. This illustrates another distinction: an asset accepted as backing is not necessarily the asset needed to discharge a trading obligation.

If the required payment differs from the collateral held, the process may require conversion, financing or another agreed mechanism. A valuable collateral portfolio can therefore coexist with a shortage of immediately deliverable settlement assets. Stablecoins can also introduce issuer, redemption and network dependencies; they are not interchangeable with every payment obligation simply because their reference value is a dollar.

External custody can be useful where an institution wants trading access without placing all supporting assets directly with an exchange. But segregation, payment authority and access after a default require examination of the actual agreements and applicable law. A custody label does not establish recovery rights or timing.

Bitget’s documentation explicitly acknowledges continuing trading, market, custody and settlement risks. Eligibility, jurisdiction and final agreements also govern access; the announcements do not establish availability to US readers.

The test is completed settlement

Independent trade coverage of the November launch describes the institutional audience and mirroring model. It does not establish adoption volumes or independently validate risk reduction.

The evidence supports a mechanism-level conclusion: off-exchange settlement risk changes with the size and duration of outstanding obligations, and with the liquidity available to discharge them. It does not support approval of any arrangement as demonstrably safer; that judgement must be withheld.

The decisive questions are concrete: who owes the balance, in which asset, by when, and what happens if payment fails? Custody separation changes where collateral sits. Completed settlement—not the announcement, mirror balance or scheduled clock—determines when a particular obligation is gone.

Sources and References

• Bitget: November 5 Open Custody Network announcement, via GlobeNewswire.
• Bitget: OCN documentation, dated October 15.
• Sygnum: September 24 Bitget–Protect integration announcement.
• The Digital Assets Edge: November 5 OCN coverage.