On 6 December, Nasdaq opens a Night Session running from 9 in the evening until 4 in the morning. NYSE Arca will open an Overnight Session on the same hours, on the same date. Cboe has filed for near 24x5 trading on EDGX and is targeting the same month. So three of the largest US equities venues are converging on a single month. At that point we’ll be in a world where the market is open 23 hours a day, 5 days a week.
The mechanics are interesting. A trade executed on Arca at 10pm on Monday 7th December will carry a trade date of Tuesday the 8th and go to DTCC with a clearing date of the 8th. A trade executed at 1am on the 8th carries the same date. The calendar day and the trade date have come apart, and they’ll stay apart every night from December onward. NYSE’s documentation works through one consequence where a trade in the Overnight Session on the last calendar day of a month counts toward the following month’s trading tiers, because by the time it exists the trade date has already rolled forward.
It would be easy to read this as just a bit of market structure arcana, a bid for Asian retail flow. But I don’t think it is. What’s being changed is the coordinating device the rest of finance was built on top of, and I doubt the long tail of financial institutions that depend on it has noticed.
The filing that went first
On 2nd April, NSCC filed to move its Universal Trade Capture system to a 24x5 operating model, open for trades from Sunday at 8 in the evening to Friday at 8 in the evening. The SEC approved it on 27th May. It went live on 28th June.
NSCC states that it
“is not currently proposing any changes to its risk management rules or margin/Clearing Fund methodology in connection with the move to 24x5.”
Overnight activity gets handled through the existing framework, treated much like the overnight and premarket activity NSCC already clears. The Margin Requirement Differential charge, which runs on a 100 day historical look back, is described as already capturing the risk of accumulated trades through the day “including any overnight trading session.”
What the whole arrangement rests on is a forecast. NSCC
“believes that overnight trading volumes will increase gradually and steadily over the next few years as ATSs and Exchanges expand and normalize overnight trading hours, as opposed to seeing an immediate significant increase in volumes.”
That isn’t exactly what the curve has looked like so far. Blue Ocean ATS, where most overnight equity trading happens today, went from $5.95 billion of notional in 2022 to $382 billion in 2025, and in the first two months of 2026 was running 244% ahead of the prior year, with a single session on 2nd February clearing $10.5 billion. Those are still small numbers against a regular session, but they’re not a gradual and steady curve.
There’s an operational detail that matters I think. Trades accepted before the final Good Night Message, at roughly midnight, fold into the start of day margin collection. Trades arriving after that go into intraday monitoring instead. NSCC’s intraday systems generate and monitor volatility and mark to market exposures on a 15 minute basis between 6am and 11pm. The overnight session runs until 4am. So for part of the night, the high frequency monitoring window and the trading window don’t overlap, which suggests to me there is some risk sitting inside a framework calibrated on a market that closed at 8pm.
I’m not suggesting negligence. NSCC has committed to evaluate and propose enhancements, and said it would do so before the exchanges expand their own hours, which is now three months away. But it does feel like the risk architecture for a 23 hour equity market currently rests, at least partly, on a stated expectation that not much will happen at 3am.
What the business day holds up
The reason I think this matters beyond clearing is that the concept of a business day was never just about a trading schedule. It is the shared assumption underneath a great deal of machinery that nobody thinks of as time dependent, until it stops working.
Value dates, interest accrual, funds transfer pricing and the margin cycle all assume it, along with every operational control built around a cutoff. Think about how month end assumes it too. The 11:59 cutoff exists because a bank’s books have to stop moving at some point, so that everything downstream has something fixed to reference.
Sam Sidhu of Customers Bank put the operational version of this pretty plainly at Jackson Hole this week. Banks like sticky deposits, he said. Sticky deposits mean friction, and friction creates franchise value. At 2am, no treasury team is sitting there managing liquidity. Yuval Rooz of Digital Asset told the same conference about the other end of it. On one of the first weekend repo trades, a participating bank discovered its systems couldn’t book a trade with a Saturday trade date.
That’s a system that had correctly encoded an assumption everybody shared, meeting a world where the assumption stopped being true.
The last time the two layers came apart
I like looking for historical precedent, and we have seen something similar before, although in the opposite direction. Its actually unusually well documented because the NYSE gave Congress a chronology in 1971.
The trading floor of the late 1960s coped fine. What could not cope was everything behind it, where securities were still physical certificates moved by hand. Volume in 1967 ran a third above 1966; on 25 days in 1968 it exceeded the record set in October 1929. Fails to deliver reached $4.13 billion at the end of 1968. The Exchange’s own account is candid about one aggravating factor, and it is worth reading for anyone currently confident that new technology will absorb a step change in throughput. The crisis stemmed in part from
“hasty efforts to apply sophisticated computer technology to operations problems which had not been adequately analyzed in advance.”
The response was to take hours out of the market. From 22nd January 1968 trading in all US securities markets was cut to 4 hours a day, 10am until 2pm. Settlement went from T+4 to T+5 in February. From 12th June the exchanges closed one day a week, on Wednesdays, and stayed shut on Wednesdays until the last day of the year. The 4 hour day continued through 1969 and was not fully restored until around May 1970. 1968 had 12% fewer trading hours than 1967, through the elimination of 26 trading days and the shortening of 28 others, and still set a volume record.
The cost was significant. More than 160 NYSE member organizations went out of business across 1968 to 1970. About 80 self liquidated or left the securities business; most of the rest were merged or absorbed, frequently in transactions the Exchange itself arranged. The Exchange intervened directly in the affairs of nearly 200 member firms, more than half of those dealing with the public.
The schedule changes may have bought time, but the plumbing was the actual answer. The Central Certificate Service began operations on 21 June 1968, nine days after the first Wednesday closure, transferring ownership of four listed issues by computerized book entry. By February 1969 it covered around 1,200 NYSE issues. It became the Depository Trust Company in 1973, NSCC followed in 1976, and the market got the settlement infrastructure it still runs on.
So, as I read it, the precedent is this. The last time the trading layer outran the settlement layer, the market’s immediate answer was to shorten the day, and its lasting answer was to rebuild the infrastructure underneath. This time the sequence is running backwards. The infrastructure went first, in June, and the day gets longer in December.
Which is progress, if the assumption about volumes holds. Although I’m not sure it does. Overnight equity trading is already growing at a rate that does not look particularly gradual. Blue Ocean ATS, where most of it happens today, went from $5.95 billion of notional in 2022 to $382 billion in 2025, and in the first two months of 2026 was running 244% ahead of the prior year, with a single session on 2 February clearing $10.5 billion. Those are still small numbers against a regular session, but they don’t feel like a gradual and steady curve.
Somebody has to be awake
I think this actually lands on bank balance sheets rather than on exchange P&Ls.
An always on market is cheap for the venue and expensive for everyone financing positions inside it. The exchange runs a matching engine that doesn’t care what time it is. The bank on the other side has to fund, margin, monitor and staff around a clock that no longer stops, and the cost of that will be felt on the balance sheet rather than just in headcount.
The underlying obligation is already pretty large. BIS Working Paper 1089 found that participants in Fedwire collectively used an average of $630 billion of intraday liquidity every business day between 2008 and 2018, with a maximum near a trillion. JD Risk Solutions and UBS put the all in annual cost of holding those buffers at roughly $600 million for a top tier Fedwire participant, applying a credit spread of about 100 bps to an average buffer near $60 billion. Intraday liquidity is expensive, and the reason banks can size it at all is that the day has a shape.
Now take the shape away on one side only. Equities trade at 2am from December. The repo that funds equity positions does not. Broadridge’s distributed ledger repo platform, the fastest moving collateral infrastructure in the market and the one carrying roughly $365 billion a day, runs on a US business day calendar because triparty repo does. Treasury markets keep their own hours.
The clock was built on purpose
Did you know that before 1883 the United States ran on more than 140 local times?
There were roughly 80 separate railroad time standards layered on top of those, because a railway could not publish a timetable across towns that each kept their own noon. On 11 October 1883, the General Time Convention met at the Grand Pacific Hotel in Chicago and voted to adopt four zones. On 18 November, every railroad clock in the country was reset, including those that had already passed midday, which is why it is remembered as the Day of Two Noons.
There was no act of Congress involved. There were plenty of objections. The Indianapolis Sentinel complained that the sun was “no longer to boss the job,” and that people would have to marry by railroad time and die by railroad time. The courts sided with the railroads. Most cities adopted within a year, but Detroit held out until 1905, 22 years later. Congress finally put standard time into federal law with the Standard Time Act of 19 March 1918, 34 years and 4 months after the industry had settled the matter itself.
I think a couple of things in that story are worth thinking about. The first is that a private industry imposing a coordination standard ahead of the law is not that unusual, and it can work. The second is why it worked. Everybody moved at once, on one day, and the standard was adopted by the whole system rather than by the fastest-moving part of it.
But that is clearly not what is happening now. The venues go in December. Clearing already went in June, on an explicit assumption about volumes and without changing its margin models. Repo, Treasuries, corporate credit, bank treasury operations and the accounting calendar aren’t going at all. The railroads built the shared clock in a single coordinated move. The industry dismantling it is doing so one filing at a time, and I suspect the seams between the parts that moved and the parts that didn’t are where this may go awry.
Directionally, I think this is right. Capital, information and risk genuinely are 24 hours now, and a market that closes at 8 exports its worst moments to whatever venue is still open. Kevin Kennedy of Nasdaq is right that the question has moved from whether this happens to whether it is done properly. My worry is narrower. “Properly” has been redefined to mean “the venues are ready,” but I’m not sure the venues were ever really the binding constraint.
Everything downstream of a trade, the accrual, the margin call, the funding, the reconciliation, the general ledger, was engineered against a clock that stops. From 6 December, for the venues at least, it doesn’t. NSCC told the Commission it expects overnight volumes to increase gradually and steadily, and the risk architecture of a 23 hour equity market now rests on that.
References
Trading venue schedules
NYSE Extended-Hours Trading FAQ (updated August 2026) — Overnight Session hours, trade-date and clearing-date mechanics, monthly tier treatment
SEC Release No. 34-105532, NYSE Arca Rule 7.34-E(T) (filed 12 May 2026)
SEC Release No. 34-105199, Nasdaq Night Session approval (10 April 2026)
Nasdaq aims to debut 23/5 trading on 6 December 2026 — Markets Media, 23 April 2026
Cboe files proposal for near-24x5 US equities trading — 16 March 2026
Blue Ocean Technologies, overnight markets overview for the Financial Information Forum — 29 March 2026
Clearing and settlement
SEC Release No. 34-105210, SR-NSCC-2026-006 — NSCC 24x5 Universal Trade Capture filing, 2 April 2026
SEC Release No. 34-105565, approval order — 27 May 2026
DTCC’s NSCC goes live with 24x5 clearing — 29 June 2026
The 1968 paperwork crisis
Crisis in the Securities Industry: A Chronology 1967–1970 — New York Stock Exchange, filed with the House Subcommittee on Commerce and Finance, 30 July 1971
Remarks of Commissioner Richard B. Smith — SEC, 8 February 1971
Providing a Public Service — DTCC, March 2021
Standard time
Whose Time is it Anyway? A Brief History of Standardized Time Zones in the United States — Library of Congress Law Library
The Day of Two Noons — Library of Congress
Intraday liquidity and collateral
Intraday liquidity around the world — BIS Working Papers No. 1089, April 2023
Optimising Intraday Liquidity Management — JD Risk Solutions with UBS Investment Bank, 27 June 2024
Broadridge Distributed Ledger Repo processes $8.0 trillion in July — 10 August 2026