Netting and Compression for Bilateral Crypto Derivatives

The CFTC issued an advance notice of proposed rulemaking on October 5, 2026, sketching two frameworks that would route leveraged retail crypto transactions onto a new subcategory of designated contract market, intermediated by a futures commission merchant and cleared by a derivatives clearing organization, with comments due December 14. It rests on Section 2(c)(2)(D) of the Commodity Exchange Act. It does not address netting, and it does not address bilateral trading. Those two omissions describe the state of this plumbing accurately: the venue path is being specified, and the bilateral book is being left to document itself.

Three Operations, One Word

Netting names three different things, and conflating them is how an institution pays for one while believing it bought another.

Payment netting is operational. Same-day, same-currency obligations offset so fewer transfers leave the building. It reduces settlement traffic and nothing else.

Close-out netting is legal. On a default, every transaction under one master agreement terminates and collapses into a single net sum owed one way. It reduces exposure and the capital held against it, and it is not a feature of any system. It is an assertion in a contract that an insolvency court has to honor.

Compression is portfolio surgery. A set of economically redundant trades is replaced with fewer trades carrying identical net risk, reducing gross notional and line items without changing market exposure.

So close-out netting reduces what is owed when a counterparty fails, and compression reduces how many contracts exist. Netting alone leaves gross notional and its operational load untouched; compression alone tidies the book while default exposure stays gross.

The Netting Set Is the Unit of Account

All three depend on one object: the netting set, meaning the collection of trades that collapse together. It is the unit of account the whole system is denominated in, and its boundary is drawn by a different mechanism depending on where the trade lives.

On a venue, the boundary is drawn by where collateral sits. The margin engine holds one net position per account per instrument and nets by construction. Enforcement is immediate, automatic and free, and conditional on the venue's solvency and its terms of service.

Bilaterally, the boundary is drawn by a document. The master agreement names the entities and the transactions in scope, and the net is produced by a court applying that document under insolvency law. ISDA's analysis concluded close-out netting is likely enforceable for digital asset derivatives in major jurisdictions including England and Wales and New York, with US protection running through the Bankruptcy Code safe harbors at sections 555, 556, 560 and 362(b). Enforceability still turns on each counterparty's local insolvency law.

Same word, two enforcement locations, opposite failure profiles. The engine-enforced version is strong until the venue fails. The contract-enforced version is weak until documented and opined on, and then it survives exactly the event that destroys the other.

Why Compression Barely Exists Here

Compression is an algorithm over a netting set, so it cannot run until that set's edge is stable and both sides agree on it. Three preconditions follow and crypto satisfies none of them well.

Standardization first. Rates compression works because a five-year dollar swap has identical terms wherever it trades, making redundancy detectable. Bilateral crypto derivatives vary in reference index, settlement asset and collateral terms, so genuinely offsetting positions are rarer than gross numbers suggest.

Multilateral consent second. Canceling and replacing trades requires every participant to agree to termination and novation, which needs a service with critical mass sitting on a netting opinion base.

Then the awkward part: venues already get this free. A perpetual position is one net line, so no gross notional accumulates to compress. Compression is a problem created by bilateral documentation, which is why it appears exactly where the documentation is thinnest.

The payoff is not small. ISDA modeled an illustrative $160 million portfolio of bitcoin and ether non-deliverable forwards and options and found that faster settlement plus compression cut valuation adjustments by roughly 40 to 45 percent with volatility held constant, moving a capital proxy from about $6.3 million to about $3.7 million.

The Break That Only Appears at Close-Out

The operational failure here is specific and quiet: the two sides disagree about which trades are in the set.

One desk books five transactions under a single master with one legal entity. The counterparty booked the fourth against a different group entity, or under a separate agreement. Nothing malfunctions. Confirmations match, valuations agree, margin calls reconcile to the penny, because the trade economics are correct on both sides. The disagreement is about membership, and membership gets tested once, at close-out, when gross appears where net was expected.

The control is therefore not an economic reconciliation but a reconciliation of the agreement identifier and contracting legal entity, trade by trade, against the counterparty's own record of those fields.

Two checks belong beside it. Confirm a netting opinion actually covers the counterparty's jurisdiction and the instrument type, since ISDA's June 2026 review notes that close-out netting for ledger-based collateral is not yet addressed in the opinion frameworks at all. And confirm which entity in a group faces you, because netting sets do not span entities however the relationship is managed.

Controls, Costs, and What Is Improving

Costs sit in two places and get measured in one. Net exposure drives credit and margin, which everyone watches. Gross notional drives line-item count, reconciliation work, dispute surface and, for bank counterparties, exposure measured on gross. The second is what compression attacks and what rarely appears in a budget.

The collateral questions alongside this stay open and are worth naming rather than resolving here: perfection and priority of security interests, enforceability of on-chain title transfer in insolvency, and conflict-of-law rules for ledger assets.

What is improving is the legal substrate, which is the slow part and has to come first. The UK's Property (Digital Assets etc.) Act 2025 confirmed digital assets can be the subject of personal property rights, several US states have adopted UCC Article 12, the CFTC issued tokenized collateral guidance in December 2025, and ISDA has published model provisions for tokenized variation margin.

For a technical owner the instruction is narrow. Make the agreement identifier and contracting entity first-class fields on the trade record rather than reference data, because the netting set is the unit everything is denominated in, and no amount of correct economics will tell you where its edge is.

For informational purposes only. Not an offer to buy or sell any security. Available only to accredited investors who meet regulatory requirements.

‍