The Inevitable Pull of Real-Time Settlement

When Oracle reported Q1 results last September, the stock moved 36.7% before the New York open. Most of the gain happened while US equity markets were closed — price discovery playing out on alternative venues with a fraction of normal liquidity while most participants watched from the sidelines. Four weeks later, 24X National Exchange launched as the first SEC-approved 23/5 stock exchange, and the infrastructure to trade US equities around the clock went from theoretical to operational.

The four months since have compressed a timeline the industry expected to play out over years. Nasdaq filed with the SEC in December to extend trading to 23 hours a day, targeting the second half of 2026. NYSE Arca received accelerated approval for 22-hour operations. DTCC’s clearing corporation goes live with 24×5 processing on June 28 — Sunday 8pm ET through Friday 8pm ET. Cboe launched nearly 24-hour Russell 2000 index options in February. Every major exchange and clearinghouse in the US has converged on the same structural endpoint: continuous markets with a single one-hour maintenance window, 8 to 9pm Eastern, Monday through Thursday.

The exchange schedules are the visible change. The operational question is what happens to everything that currently runs during the eight hours between market close and market open.

What eight hours subsidized

Trading operations follow a rhythm that batch processing built into the architecture. Markets close and batch processes kick off — trades match, reconcile, and settle while risk recalculates, compliance reports generate, system patches deploy, and data warehouses refresh. On a good night this takes four to six hours. On a bad night — end of quarter, corporate actions, a surge in settlement exceptions — it takes all eight.

This cycle isn’t a quirk of legacy systems — it’s the architectural assumption the entire operational stack was built on. Trade capture systems write to databases designed for batch extraction. Reconciliation engines compare end-of-day snapshots, and reconciliation takes as long as it does because every counterparty delivers data in its own format, with its own identifiers, on its own schedule. Before you can match, you have to translate, and that translation work is what fills the window. Margin systems compute against a fixed close-of-business position. Compliance surveillance runs pattern detection against completed-day activity. The trading applications themselves can process at any hour, but the plumbing between them can’t.

T+2 settlement gave two full days of processing time. T+1 compressed that — the industry spent $30 billion adapting — but the nightly batch window survived, narrower but intact. Under 24/5, that window compresses to sixty minutes between 8 and 9pm Eastern, and even that exists primarily for system health checks. The DTCC has been explicit: the one-hour pause is for maintenance, testing, and clearing activities that can’t run during live trading, not a replacement for the nightly batch cycle.

And some of the workloads filling that window have been growing while the window around them narrowed. Fractional share reconciliation is among the most significant. The clearing layer cannot process sub-share quantities — NSCC stated in the Federal Register last September that it “is only able to accept trades for clearing in units of full shares” — so every fractional trade runs through a two-tier accounting system: broker sub-ledgers tracking fractional ownership against whole-share positions at the depository, with manual reconciliation at every boundary. Retail fractional trading now accounts for more than 20% of US equity volume, and manual intervention rates reach 20% at firms running legacy infrastructure. That reconciliation runs in the overnight window, and the volume flowing through it has been compounding while the window itself has been shrinking.

The one-hour constraint

At a mid-to-large broker-dealer, trade reconciliation — matching internal records against exchange confirmations, custodian statements, and clearing reports — regularly consumes two to three hours. Most of that time goes to normalization: every counterparty speaks a different format, uses different identifiers, delivers on a different schedule, and when you’re reconciling across multiple asset classes, hundreds of counterparties, and dozens of venues, the translation work is what takes hours. Once data speaks the same language, matching is fast.

Risk recalculation follows a dependency chain that starts with reconciliation. End-of-day VaR, stress testing, and regulatory capital calculations rely on final positions, which depend on completed matching. Risk numbers often aren’t available until three or four hours after market close, which is why margin calls go out in the morning.

Compliance surveillance operates on a similar lag. Trade surveillance systems — detecting spoofing, layering, wash trading — analyze completed sessions against algorithms calibrated for normal-hours liquidity and volume patterns. Extended hours, with thinner order books and wider spreads, produce different statistical signatures that can generate false positives or mask actual manipulation.

System maintenance — patches, upgrades, database optimization, failover testing — requires downtime that doesn’t exist in a 23-hour trading day. One hour shared with clearing operations isn’t enough for a database migration or a major software update.

Fractional share reconciliation adds to the pile: every exception requires unwinding the aggregation that mapped individual retail orders to the whole-share blocks submitted for clearing, and those exceptions run through the same overnight window as everything above.

None of these compress into sixty minutes, which is why the firms ready for June 28 aren’t trying to compress their batch windows — they’ve eliminated them.

What the first movers found

The firms furthest ahead share a pattern: they stopped asking how to make batch processes faster and started asking which ones needed to exist at all. The answer was fewer than expected. Their trading applications stayed in place, and what changed was how data moved between them.

Reconciliation was the first domino. Instead of waiting for the batch window to translate and compare, the leading firms moved to continuous matching — every trade, every position update resolved the moment it arrives. One global asset manager went from twenty-minute fund reconciliation cycles to under ten seconds. The operations team didn’t grow. Breaks that used to require manual intervention largely disappeared once mismatches were caught before they had time to compound — and for fractional trades, where a single break in the whole-share block cascades across every underlying retail position, catching mismatches early was especially valuable.

Risk followed the same trajectory. The NSCC recalculates VaR every fifteen minutes with intraday mark-to-market. J.P. Morgan’s global head of prime services put it plainly: “The ability to calculate margin in real-time is critical.” The Options Clearing Corporation now produces hourly mark-to-market reports during extended hours, with kill-switches that activate if risk thresholds breach — daily risk calculation giving way to continuous monitoring.

Compliance required a different kind of change. When trading spans 23 hours across varying liquidity regimes, surveillance thresholds can’t be static. What flags as spoofing during the opening auction looks different at 2am when the order book is thin, and the firms handling this well can adjust parameters without a code deployment — which matters when an exchange changes a position limit overnight and regulators have issued over $189 million in fines in the first half of 2023 alone for compliance failures (Capco / FINRA enforcement data).

The deployment pattern mattered as much as the architecture. The firms that moved fastest ran new and old in parallel — same data, both paths — until they trusted the new numbers as much as the old ones, then migrated gradually with the ability to fall back if discrepancies emerged. The T+1 transition in May 2024 validated this approach: firms that compressed batch windows incrementally achieved 95% confirmation rates by the 9pm deadline, up from 73% in January 2024, while settlement fail rates improved and the NSCC Clearing Fund dropped 23%, freeing $3 billion in capital. Operations teams didn’t grow even as per-hour processing demands increased, and the exercise revealed that most of what ran during the batch window had inherited batch as default because nothing underneath had been designed for continuous operation.

T+1 also exposed dependencies the industry hadn’t anticipated. When India moved to T+1 in early 2023, 90% of Swift messages from APAC customers settling North American equities were initiated after trade date — FX management proved more operationally complex than anyone had modeled. For 24/5, similar gaps will emerge wherever connectivity hasn’t been tested at speed: corporate actions processing, securities lending recall timing, or settling trades executed at 3am against payment rails that weren’t designed for it. The firms that can absorb new counterparties in weeks have an advantage over firms where each new connection requires months of integration, and the gap widens as APAC participants like Shinhan Securities and Rakuten Securities join venues like 24X expecting connectivity on their timeline.

Access without liquidity

Extended hours create access without creating liquidity. Off-hours trading represented 11.5% of US equity volume in Q2 2025, spread across sixteen additional hours — meaning average hourly volume outside the core session runs at roughly a tenth of regular hours. Thin liquidity means wider spreads, partial fills, and sharp price moves on modest volume. Reg NMS protections — protected quotes, NBBO requirements, route-away rules — don’t apply outside standard hours, so the price discovery mechanism that provides stability during the day is absent at night.

Cryptocurrency markets offer a cautionary parallel. Bitcoin has been three to four times as volatile as equity indices between 2020 and 2024, and crypto has always operated on continuous settlement — every trade settles immediately, meaning participants face instant liquidity demands with no buffer to absorb sudden moves. When prices swing sharply, margin calls and forced liquidations cascade at machine speed. Traditional markets moving toward continuous operations may find similar dynamics during thin hours, moderated by more mature risk management and decades of institutional learning — though the experiment hasn’t run at this scale, and the outcome isn’t predetermined.

The firms preparing for this are building risk frameworks that distinguish between sessions — overnight positions carry different margin buffers, execution algorithms adjust to available liquidity, alert thresholds scale with time-of-day volume. Managing a position book across varying liquidity regimes, from the deep pool of the opening auction to the shallow hours after midnight, is a different operational problem than managing a single-session trading day.

Where the pull leads

Dave Olsen of Jump Trading put it simply: risk doesn’t go away when the market is closed. The market is about to stop closing.

DTCC goes 24/5 on June 28. NSCC testing is live. Nasdaq, NYSE, and Cboe have filed or been approved. 24X is already trading. Almost 60% of securities firms told DTCC they need IT and risk management upgrades. The one-hour window compresses everything that batch processing used to handle — translation, matching, risk, compliance, maintenance — into a period shorter than most firms’ current reconciliation cycle alone. The firms that recognized this early built infrastructure where those processes run continuously, and their trading applications didn’t need to change. The firms discovering it now face the same architectural question with less runway: redesign how data flows between your systems, or try to run eight hours of processing in sixty minutes.

Of the workloads running through that window, fractional share reconciliation deserves particular attention. FINRA’s six-decimal reporting mandate goes live on February 23, adding precision requirements to a reconciliation layer that was already stretching the batch window — and the clearing layer’s structural inability to process sub-share quantities means those requirements flow through workarounds that have been scaling for five years. What happens when that precision meets the one-hour constraint is the next question worth examining.