The word “clearing” is doing too much work

Following infrastructure news is hard when the same word covers three different functions.

Last Wednesday the SEC approved a Paxos subsidiary as a clearing agency.

By Thursday morning the same event was being reported as “approved to clear US stocks on blockchain” (CoinDesk), “approved to clear and settle US securities” (Bitcoin.com), the first blockchain-native firm “to provide clearing and settlement services” (The Block), and “secures SEC license to settle securities transactions nationwide” (PYMNTS). Paxos’s own press release described the new entity as a “settlement company” that “receives clearing agency registration”.

None of those headlines is technically wrong. They are also describing at least three quite different functions, in the same sentence, using the same word.

This is the problem with following infrastructure news. The post-trade vocabulary collapses several distinct activities into a single term, and the more important the news, the more aggressively the language collapses. By the time a story moves from the Federal Register to the headlines, the function that was actually approved has been merged with two or three other functions that were not. That matters when the difference between those functions is where the risk sits.

Four things called “clearing”

The first is central securities depository and settlement. The mechanical job of holding the official record of who owns what, moving the record when securities change hands, and making sure the cash and the securities arrive at the same time. In the US this is DTC for equities, Fedwire and DTC for Treasuries, and a long tail of niche infrastructures. Same-day settlement, T+1 settlement, T+0 settlement are all variations on the same function. This is what Paxos PSSC actually got registered to do.

The second is central counterparty clearing. Standing between the buyer and the seller as legal counterparty to both, taking the credit risk of each one defaulting, and managing that risk through margin requirements, default funds and a loss waterfall. CME Clearing, LCH, Eurex Clearing, ICE Clear, all do this. It is the function that absorbs counterparty risk and mutualises losses across non-defaulting participants when the absorption fails. It is also the function PSSC explicitly does not perform. The order says so directly, more than once.

The third is custody and asset servicing. Holding the actual securities, processing corporate actions, dividends, voting rights, tax withholding. Banks do this. Broker-dealers do this. DTC does this for the depository layer. The Paxos Ledger does a version of it inside the PSSC system through a “digitisation” process that creates an entitlement on a private permissioned blockchain corresponding to securities held in PSSC’s DTC account.

Payments clearing. In payments, clearing is the work that happens between the moment a bank tells another bank that money should move and the moment the central bank actually moves the reserves. It is the collection of payment instructions across a window, the netting of those instructions down to a single number each bank owes by the cut-off, and the production of the settlement file that the central bank’s settlement system acts on. Bacs in the UK and ACH in the US run this way: instructions accumulate across the day, the clearing system nets them, and a deferred net settlement happens at fixed times on the central bank’s books. SEPA in Europe runs a deferred-net variant and an instant variant in parallel. Real-time gross settlement systems like Fedwire and CHAPS skip the clearing window entirely and settle each payment individually as it arrives, which is why payments specialists treat them as settlement systems in their own right rather than clearing-plus-settlement arrangements.

A headline that says “Paxos approved to clear and settle US securities” is, strictly speaking, describing the first function and silently implying the second. The implication is wrong. PSSC is not a central counterparty. There is no novation. There is no mutualised default fund. Participants face each other on credit, not the clearing agency. That is a fundamental design choice and it is the most interesting thing in the order, and the headlines as a class do not flag it. Not saying that Paxos intended this, just that the language is constantly sloppy in the industry.

Why the conflation happens

There is a historical reason these functions get bundled together verbally. In US equities, DTC (the depository) and NSCC (the CCP) are both subsidiaries of DTCC, and DTCC also markets itself as “the clearing and settlement infrastructure”. So in the equities world, the two functions sit inside the same parent company, and it is reasonable shorthand to refer to “DTCC clearing” or “DTCC settlement” without much consequence. The same shorthand carries over into how journalists, regulators in informal mode, and even practitioners talk about the space.

That shorthand falls apart the moment somebody builds an infrastructure that deliberately separates the functions. PSSC is doing exactly that. It is taking the depository and settlement piece, building it on a private permissioned blockchain, and explicitly declining to take the central counterparty piece. Participants need to manage their credit exposure to each other. PSSC needs to manage the ledger. Those are two different jobs and PSSC is only doing one of them.

The same conflation happens in derivatives coverage, in a different direction. Crypto futures venues frequently describe themselves as “clearing” their own trades, when what they mean is that the exchange is also the central counterparty (vertically integrated) and the settlement layer is the exchange’s internal book. That is a different architectural choice from the equities model, but the language is identical to a casual reader. “Clearing” again does several jobs at once. If is doesn’t do one of the jobs well, this gets glossed over. Clearing does not equal settlement.

Reading infrastructure news with the language stripped out

A practical filter for any infrastructure announcement is to ask questions.

What function was approved or changed? Who holds the official record of ownership, and how does it move? In the Paxos case, the answer is that PSSC’s private permissioned ledger now has SEC clearing-agency status as the record for trades between its CP Pairs, sitting on top of DTC, which continues to hold the underlying securities.

What credit function was approved or changed? Where is the counterparty risk, who absorbs it on default, and is there mutualisation? In the Paxos case, the answer is nothing changed. PSSC is not a CCP. Participants face each other. There is no loss waterfall.

What custody and asset servicing function was approved or changed? Who holds the assets, who handles corporate actions, who collects withholding? In the Paxos case, DTC continues to hold the securities through PSSC’s omnibus account, with PSSC mirroring the holdings on its ledger and managing corporate-actions processing through arrangements that DTCC raised questions about in its comment letter.

Three questions, three different answers, one approval. That is the texture an infrastructure story actually has. The headlines collapse it into one verb.

Why this matters operationally

For anyone evaluating clearing counterparties (and this is the practitioner audience for this newsletter, so the operational consequence is the point), the conflation is not academic. An institutional approval framework that does not separate the three functions cannot evaluate a counterparty like PSSC properly. It will either approve too quickly, assuming that “clearing agency registration” implies CCP-equivalent risk-pooling that is not there, or it will reject it as not real clearing, missing that the depository and settlement function is genuinely there and is genuinely useful.

Both errors are happening already in the conversations the team has been having through the week. Phil has been reporting that approval committees he has sat in are not built to distinguish between the three. Neil has been pulling on the operational resilience standard for post-trade infrastructure (PFMI Principle 17), which applies to CCPs, CSDs and securities settlement systems alike, so the same regulatory bar applies regardless of which of the three functions it is performing. Wolfgang has been on the capital arithmetic, which only makes sense once you can say which of the three functions the counterparty is performing. Matt has been on the legal layer, where the statutes themselves treat the three differently (Section 17A of the Exchange Act for securities clearance and settlement, the Commodity Exchange Act’s Derivatives Clearing Organisation regime for derivatives CCPs, and a patchwork of trust company and money transmitter law for custody).

The thread tying them together is the language problem, not the architecture problem. The architecture is fine. The problem is that the way infrastructure news is reported, and the way institutional approval frameworks are written, does not separate the three functions cleanly enough for a reader to know what they are looking at.

What gets lost in the noise

The more important point is what the conflation crowds out.

The substantive conversation in clearing right now is not really about whether a particular firm has registered. It is about what becomes possible when the three functions are separated and each one is built with the tools we actually have available now.

Take risk management. The major CCPs already run intraday margin cycles, sometimes several a day, and have done for years. The harder question is the direction of travel. Moving from a handful of scheduled intraday recomputations toward something close to continuous risk visibility has been a cost question more than a methodology question. Full-precision VaR across a real portfolio is expensive compute, and the cost goes up faster than the frequency does. The AI-engineering work going on right now is changing that cost curve. Neural approximations of VaR and SA-CCR exposures, trained against the precise models and validated against them in production, can give you a number so close to the models at a fraction of the compute that risk reduces, all well inside the latency window an intraday cycle has to fit into. That opens up margin frequency, cross-margining across products, intraday risk visibility for the participant as well as the CCP, and over time the amount of capital that has to sit idle against modelled tail events. The bottleneck is computational, not regulatory. The economics of lifting it have moved. There are much better methods available just as far as I can see not implemented in Financial Markets yet.

Take capital. QCCP treatment under Basel III gives a materially lower risk weight on a bank’s trade exposures to the CCP than it would face against a non-qualifying counterparty. The headline 2 percent figure is shorthand for the trade-exposure leg under specific conditions, and the real capital answer depends on the exposure type, the SA-CCR calculation, default-fund contributions, and whether the bank is clearing directly or as an indirect client. That is exactly the point. We use QCCP because it is the operative term, the one that carries a defined regulatory consequence rather than a marketing claim, and the rest of the answer follows the rulebook from there. That precision is the first casualty when the surrounding vocabulary is loose. “Cleared” stops meaning anything specific the moment it covers both QCCP-eligible CCPs and infrastructure that has explicitly chosen not to be a CCP at all.

Take market access. Cross-margining only makes sense once you can say which counterparty bears which risk, across which products, on which depository. A vertically integrated venue cross-margining inside its own book is doing something quite different from a CCP cross-margining across two product silos that each have their own independent depository. Both get described the same way in coverage. The capital and operational consequences are not the same.

There are specific positions and specific solutions inside the wider DCN work on each of these. None of them lands when the surrounding language is still doing too much work. If everyone is using the same word for different things, the specific arguments get drowned out, and the readers most likely to act on them have already scrolled past.

That is why the language problem sits upstream of every other problem in clearing right now. Until the terms separate, the substance does not.

A working reference: what “clearing” and “settlement” actually mean

The rest of this is a reference. The same two words mean different things in capital markets and in payments, and different things again when they are used together. Worth having in one place.

Clearing

In cash equities, clearing is the post-trade work of matching trades, netting obligations, and confirming what each side will deliver on settlement date. In the US market this work is done by NSCC, which is a central counterparty and does novate, stepping between the buyer and the seller. The clearing work and the CCP function arrive bundled, but they are separable in principle, and PSSC has separated them by performing the netting and matching without the CCP layer.

In derivatives, clearing is the same work plus novation. Typically a central counterparty steps in as legal buyer to every seller and seller to every buyer, takes the credit risk on both sides, and manages it with initial margin, variation margin, a default fund and a loss waterfall. This is the function CCPs perform. It is the function PSSC explicitly does not.

In payments, clearing is the exchange of payment instructions between banks and the netting of those instructions across a clearing window. Bacs in the UK and ACH in the US run on this deferred-net model. SEPA in Europe runs both deferred-net and instant variants. The clearing step is what produces the net position each bank owes by the cut-off.

The shared meaning: clearing is what happens between the instruction and the moment value moves. The risk and the netting work sits in the middle.

Settlement

In cash equities, settlement is the final transfer of the security and the corresponding cash, simultaneously where possible (delivery versus payment). DTC settles US equities, Fedwire settles US Treasuries. T+1 became the US standard on 28 May 2024. PSSC enables same-day settlement between its participants on its own ledger, on top of DTC’s underlying cycle rather than instead of it.

In derivatives, settlement is the final movement of margin, variation margin, the cash flows of the derivative itself, and at maturity the final settlement amount or physical delivery. The CCP orchestrates it. The actual cash flows through commercial banks, or through central bank money, depending on the regime.

In payments, settlement is the actual movement of central bank money between bank reserves. Real-time gross settlement systems like Fedwire in the US and CHAPS in the UK settle each payment individually as it arrives. CHAPS is an RTGS system in its own right rather than a clearing system feeding into a separate settlement step. Deferred net settlement takes the netted output of an ACH or Bacs clearing batch and moves the net positions across reserves at the cut-off.

The shared meaning: settlement is the moment value actually moves and the obligation goes away. Everything before it is provisional.

Clearing and settlement together

In capital markets, the combined phrase usually describes the whole post-trade stack: matching, netting, credit risk management (if a CCP is in the picture), custody, final transfer. DTCC describes itself this way because it owns both the depository (DTC) and the CCP (NSCC) under one parent. The shorthand works inside DTCC. It is also the shorthand that hides what PSSC has done.

In payments, the combined phrase usually describes the path of a payment from the originating bank to the receiving bank: the clearing system nets it, the settlement system finalises it across central bank reserves. Faster Payments in the UK, FedNow in the US, and instant SEPA in Europe close the gap between the two, but the two functions are still separable in principle.

None of this is hidden. It is all in the rulebooks. It is just not in the headlines, which is why the headlines this week described PSSC as “clearing US stocks” when what PSSC will actually do is run a depository and a settlement system without acting as central counterparty.

Where it goes from here

Paxos PSSC is the most useful worked example to come along in a while, because it forces the separation. A clearing agency registration with no central counterparty function inside it is a clean test of whether your framework, or your reader, or your headline writer, can keep the three functions apart.

We will keep tracking how the coverage and the frameworks adapt. We need to as we explain our goals and position in the markets.

James

This first appeared in The Clearing Gap, my LinkedIn newsletter, on 4 June 2026. If you want to comment, discuss, or subscribe, that’s the place.