Here is the part of the atomic settlement pitch that nobody wants to sit with. The story everyone tells is that instant, trade-by-trade settlement is the finish line: no delay, no counterparty risk, no two-day window where things can go wrong. Faster is better, and blockchain finally makes faster possible. Fine. But in the US cash equities market, that story quietly assumes the thing it is supposed to improve is broken. It is not. And the Deloitte and SIFMA whitepaper on digital cash and the future of US securities settlement is refreshingly direct about why.
My sharper take: in a netted market, atomic settlement is not an upgrade you win, it is a liquidity cost you pay. The right question is never "how fast," it is "when do I want to net and when do I want to go atomic." Those are different products, and anyone building settlement infrastructure who treats them as one thing is selling the wrong future.
The netting you throw away
US equities clear through a central counterparty that takes enormous gross trading volume and nets it down to a small net obligation per member per day. That is the entire point of the machinery. A member who buys and sells the same name all day does not move the full cash and the full securities for every ticket. They move the difference at the end.
Settle every trade gross and instantly, and that compression disappears. Now each participant has to have the full cash and the full securities on hand at the moment of each trade, because there is no end-of-day net to lean on. The whitepaper and the accompanying SIFMA analysis make the point plainly: atomic settlement removes netting, and removing netting raises the liquidity and funding burden on every participant in exactly the markets where netting was doing the most work.
I have watched this exact movie in cross-border payments. We spent a decade trying to kill prefunding, the dead capital stranded in nostro accounts so a payment can clear on demand. Gross atomic settlement in a netted equities market is prefunding wearing a new hat. You free capital from the settlement window and immediately re-trap it as intraday liquidity you must hold against every ticket. That is not a win. That is the same problem relocated inside the thing tokenization was supposed to fix.
So Deloitte and SIFMA land where the math forces them to land: T+1, intraday batch settlement, and atomic DvP will coexist, chosen by product and risk profile. Atomic settlement earns its keep where there is little or no netting benefit to lose and real settlement friction to remove: repo, collateral transfers, bilateral financing, cross-border. It does not obviously earn its keep in an already-efficient, heavily-netted cash equities market, where the thing you would be removing is the efficiency itself.
A correction to what I wrote in September
In September I argued that the SEC's innovation exemption made the settlement window optional. I want to keep the word and sharpen the claim. Optional was right. Gone was never on the table, and I should have said so with more force.
The whitepaper makes the refinement for me. The window is not abolished, it becomes a choice you make per product. And in netted markets you will frequently choose to keep it, because keeping it is what preserves the netting benefit that makes those markets cheap to run. Optionality is the real deliverable of this whole shift. Not speed. The ability to pick settlement timing against the risk and funding profile of the specific instrument.
That reframes the entire conversation. The question stops being "will atomic settlement replace T+1" and becomes "which products move to atomic, which stay netted, and who gets to decide per trade." That is a far more useful question, and it is the one the paper is actually asking.
The cash side is a policy decision
Now the part most people skip to first and understand least. If you are going to settle on-chain, the securities side is the easy half. The hard question is what cash you deliver against it, and the whitepaper puts three candidates on the table: regulated stablecoins, bank-issued tokenized deposits, and a wholesale central bank digital currency (CBDC).
The differences that matter are not technical, they are structural. Each option does something different to the liquidity math above, places intraday credit risk on a different party, and resolves differently if an issuer fails. A tokenized deposit is bank money and can be bankruptcy-remote in the way bank money is. A stablecoin is a claim on a reserve, which is a different legal animal when something breaks. A wholesale CBDC changes who sits at the center of intraday credit entirely. Pick one and you have made a policy decision about who bears risk, whether you meant to or not.
The paper's real ask follows from that: firms should set their cash-side policy deliberately, before the infrastructure makes the choice for them. And the binding constraint underneath all of it is reconciliation. On-chain and off-chain ledgers have to agree in real time for any of this to settle cleanly, and legacy plumbing was not built to reconcile in real time. That bottleneck, not blockchain throughput, is what actually gates the rollout.
What this means for anyone building
So here is where I land. If you are building settlement products for this market, the product is not speed. Speed is a feature anyone can claim. The product is giving the participant the choice of when to net and when to go atomic, with the same cash usable on both paths. A settlement layer that only does gross atomic is a worse version of what we have for most of the volume it would touch. A settlement layer that only nets throws away the genuine friction wins in repo, collateral, and cross-border. The value is in the switch, and in making sure the cash does not fork depending on which way you flip it.
Deloitte and SIFMA have essentially handed the industry a checklist: decide your cash side, solve real-time reconciliation, and stop treating atomic settlement as a prize to be won rather than a liquidity bill to be chosen or declined per product. That is the honest framing, and it is the one I have been arguing toward from the payments side for years. The destination was never "everything instant." It was "everything optional."
Living document. Last updated November 2026.