From 23/5 to 24/7: What an Always-On Equity Market Demands of Issuers and the Community

Executive summary

Our core asks

Where we are: 23/5 is built, 24/7 is not

The weekday build is almost done. Once 23/5 launches, the market will trade 115 of the week's 168 hours, and 49 of the 53 dark hours will fall between Friday 8 p.m. and Sunday 9 p.m. ET.

The build-out so far

Rules that are already set for 23/5

What the tape actually shows

Until this month the size of the overnight session was a matter of estimate. DTCC put it at about 1% of daily notional and projected 1% to 10% by 2028. On September 10 the staff of the SEC's Office of Analytics and Research filed a memo to the roundtable file that measures the session from CAT rather than from the public tape. It is the first official account of what trades at night.

In August 2026, overnight trading was 0.9% of total NMS share volume, averaging 144.6 million shares a night. That is up 359% from a year earlier and down 27% from July. Both directions matter. The session has grown roughly fivefold in a year, and it still moves enough month to month that no single print should be read as a trend.

Breadth is wider than the headline share implies. An average of 4,208 NMS stocks traded overnight each day in August, 8,335 traded at least once during the month, and 1,814 traded on every trade date. Concentration is severe all the same. The top ten names took 43.4% of overnight share volume and 50.7% of dollar volume, against 10.3% and 21.1% in regular hours, and the overnight and regular top-ten lists overlap by only about four names a day by dollar volume. Whatever trades at night is not a scale model of what trades during the day. The staff also report that foreign accounts were 37% of overnight share volume and 24% of dollar volume, the highest of any session.

The same gap has an hourly version, and it lands on the busiest part of the night. In the staff data presentation accompanying the memo, overnight volume is heavily front-weighted: the first ten minutes is the largest single bucket of the session and the first hour carries 21% of overnight volume. Volume then declines through the night and reaches its smallest bucket between 3:50 and 4 a.m., as flow is redirected to the 4 a.m. open.

Two proposals to change this are live and neither is settled. One would move the maintenance pause from 8 to 9 p.m. back to 7 to 8 p.m., on the reasoning that 8 to 9 is prime trading time in Asia. The other would shorten the pause itself, on the argument that a restart does not need a full hour. Either would put the tape back over the busiest hour of the night, and both are worth watching more closely than the December 6 date itself.

Two ways of counting, two different markets

Share counts and dollar values describe different populations of the same session, and the gap between them is wide enough to change what a reader concludes. Across the three overnight ATSs Sapinover collects from, 2,269 million shares and $104.8 billion traded on the 21 August trade dates. Names under $1 were 35.0% of the shares and 0.15% of the money. Names above $200 were 4.0% of the shares and 46.1% of the money.

The same split shows up in single names. One leveraged semiconductor ETF accounted for roughly 11% of all overnight notional on this tape in August, more than QQQ, NVDA and TSLA combined. A share-volume statistic and a notional statistic will therefore support opposite characterizations of the same session. Any rule calibrated to overnight activity, whether a price band, a tick size or a fair access threshold, should state which measure it is calibrated to.

What the night selects

Size is the easy question. The harder one is what the session is made of, and there the difference is not marginal. Measured across the 1,881 names that traded on this tape in August, leveraged and inverse ETFs took 25.9% of overnight dollars against 4.3% of the same names' daytime dollars, six times their day weight, from 311 instruments. In common stock, technology was 74% of overnight dollars against 49% by day. Healthcare, energy, consumer staples and real estate barely register at all.

The sector picture is the same story told in the underlying shares. Reading only common stock, so that the ETF wrappers above do not double-count, the night concentrates into technology and drains almost everything else.

The chart below plots every name in both sessions. If the night were a thinner copy of the day, the cloud would follow a single line. It does not.

The same picture read one name at a time gives the sharpest version of it. Overnight dollar volume as a percentage of a name's own daytime dollar volume ranges from 20% down to 0.003%, a spread of nearly six thousand to one, across names that are all heavily traded in the regular session.

What sits at the top of that list is consistent rather than random. Korea exposure, memory and semiconductors, China-listed names, and metals and crypto proxies are the instruments whose own reference markets are trading while the United States is closed. What sits at the bottom is the domestic core of the day session: large banks, telecoms, broad index funds. That is a statement about what trades, not about who trades it, and the venue data cannot answer the second question.

Narrowness follows from that selection. Ten names carry 51.9% of overnight dollars on this tape and 26.2% of their own daytime dollars, and the overnight curve sits above the regular-session curve at every rank. The staff figure for the regular session across all NMS names is 21.1%, so measured against the whole market rather than against itself, the night is narrower still.

The policy consequence is direct. A parameter calibrated on the day session's composition will not describe the night session. Price bands, tick sizes, fair access thresholds and market maker obligations all assume a distribution of instruments that the overnight tape does not have. As the session grows, the rules written for the day will need to be tested against what the night is actually made of.

Today most overnight flow still runs through ATSs that operate without a consolidated NBBO. That changes on December 6.

What 24/7 means

True 24/7 requires four layers to run continuously: trading venues, consolidated market data, clearing and settlement, and cash movement. From December 6, the first three will run five days a week. The fourth, cash, will not reach even six days until 2028 or 2029, and the Federal Reserve calls that expansion an interim step toward 24/7/365.

What the roundtable weighted

The roundtable ran six hours across a staff data presentation and three panels. Counting the market structure vocabulary across all of it, 42,395 words, gives a rough measure of where the industry's attention actually sat on the day the session was signed off. The word "overnight" was said 200 times, more than twice any other term, and is set aside so the rest can be read.

The roundtable vocabulary, grouped by purpose

Expand

Participants and liquidity took a third of the counted vocabulary, the largest share, which fits a session whose case rests on demand rather than on rulemaking. Market structure took a quarter, and almost all of it is about permission: exchanges, ATSs, the SIP and the tape. Technology and operations took 15%, the work of actually running the thing through the night.

Two shares are worth pausing on. Risk and investor protection drew 19% of the vocabulary for a session that is under 1% of share volume, and inside that group the abstract terms outrank the specific ones: risk and resiliency were said far more often than price bands or halts. And issuers, at 8%, is the smallest group while holding almost every genuinely unresolved question of the day, including tokenization, the weekend, the cash leg and capital formation.

The same terms, weighted

Expand

Read term by term, the four largest are exchanges, investors, ATSs and the SIP. The day was spent on who may operate and on whose plumbing. Asia appears 46 times, ahead of liquidity and ahead of every term for a domestic participant type except retail, which is the demand story surfacing in the language rather than in the figures.

The imbalance worth sitting with is the last one. One issuer representative sat in a room of roughly thirty participants, and he opened by reframing the entire exercise: everyone else in the room has a choice about trading 23 hours a day, and issuers do not, because their shares get traded whether they want them to or not. That asymmetry is the subject of the next section.

The issuer lens: a market with no close

Much of how a public company operates assumes the market will close. 23/5 makes that closed period short. 24/7 removes it. The table below shows what issuers lose at each step.

Issuers used to decide when the market heard their news. They are losing that control. Nasdaq's 2015 guidance asks issuers to release news at 4:05 p.m. rather than 4:01 so the closing price can print first. Separately, Nasdaq listing rules require 10 minutes' notice to MarketWatch between 7 a.m. and 8 p.m. For news released outside those hours, notice is due by 6:50 a.m. That schedule assumes nothing trades between 8 p.m. and 4 a.m. From December 6, that assumption is false.

Our expectation is that releases will move into the remaining quiet time: the 8 to 9 p.m. pause and the weekend. 24/7 removes both, so issuers need a new approach built on three things:

• Pre-notification. Notice obligations that apply at every hour.
• Coordinated halts. A standing news-pending halt that works across venues at any hour.
• A realistic Reg FD test. A clear view of what "broad, non-exclusionary" dissemination means when the first reaction may come from an algorithm in Singapore at 2 a.m. Eastern.

The 4 p.m. auction is the one price nearly every contract refers to: index levels, fund NAVs under Rule 22c-1, equity award valuations, TSR performance periods, M&A exchange ratios, convertible settlement and ATM pricing. Under 23/5 it survives, and the market keeps trading around it. Under 24/7 there are two choices. One is to keep a single daily reference auction at a fixed time. The other is to accept that the reference price may be days old by the time a weekend event is priced in. Issuers should push for the first option.

With limit orders only and 20% static bands, a small order at 3 a.m. can print a large move. Data vendors and news headlines report that print without context. For a mid cap with little overnight liquidity, a thin-session print can move retail margin calls, trigger stop orders and give activists a talking point. Issuers will need overnight price monitoring and a prepared response for moves that do not reflect fundamentals.

The case for 24/7 is access: non-US and retail holders can trade in their own hours. The cost is visibility. Some of that flow will reach the stock through tokenized wrappers and offshore platforms. There, the link between the token holder and the registered shareholder (voting, dividends, 13F and 13D visibility) depends on the structure of the wrapper, not on any choice the issuer made. Issuers have a direct interest in the terms on which their shares are tokenized, including whether their consent is needed.

Major challenges: the weekend is a settlement problem first

The hardest part of 24/7 is not keeping trading venues open. It is that money, collateral and back-office processes stop over the weekend. A trade can be matched on a Saturday, but nothing that settles it can move until the following week.

Three of these problems decide whether 24/7 can happen at all: weekend cash, the settlement calendar and weekend hedging. Everything else on the list can be solved with more staff, money and rulemaking. Those three need either new infrastructure or a different settlement model.

The connectivity layer was built around a reset

Almost every institutional order, fill and drop copy moves over a FIX session, and most sessions reset daily or weekly. FIX does not require that reset. Firms do it because it creates a maintenance window and limits how much history a reconnecting counterparty can ask to replay. 24/7 removes the window and leaves three open problems.

The FIX Trading Community is working on this. The protocol itself already supports continuous sessions, so the unresolved questions are operational: how far back recovery is supported, and how a counterparty learns that a session is going down. A 24/7 market needs machine-readable maintenance notices and session failover that every venue and broker supports.

The upside: moving overnight volume into lit markets

About 1% of daily notional already trades overnight, on venues with no consolidated quote. An always-on market moves that volume into a market that is consolidated, surveilled and open to everyone.

Roadmap: measure 23/5, pilot weekends, then scale

The sensible path has four phases, and each should start only when the previous one has produced evidence. The earliest realistic date for broad 24/7 trading in NMS stocks is after Fedwire and NSS move to six days in 2028 or 2029. Even then, Saturday still needs a solution.

Recommendations

The most urgent step is updating issuer rules for the night session before December 6. The most important step is agreeing that 23/5 data, not competition between venues, decides the timing of 24/7.

For the Commission (Day 2)

For exchanges, SIPs and DTCC

For issuers: before December 6