Skip to main content
000%
BETA// LIVEPlatform in beta. Feedback welcome.Share Feedback →
Sapinover - Click to go home
ETF Market Structure · Part II

The Eight Hours When the ETF Industry Inverts

Between 8 PM and 4 AM the exchanges are shut and the dark pools are quiet. What is left is a market roughly a five-hundredth the size of the day tape, and almost nothing about it resembles the industry that trades while the sun is up.

Key findings
  1. Leverage goes from a tenth of the tape to more than half. Leveraged and inverse products are 10.8% of all-hours ETF volume on ATS venues and 56.7% overnight. Products at 3x or higher alone are 32.0% of the night tape.
  2. The leaderboard flips. SPY is the largest ETF on ATS venues by a factor of two and falls to fifth overnight, sending 0.8% of its own volume into the session. SOXL is fourth by day and first by night, sending 32.5%.
  3. So does the issuer ranking. Direxion is fifth by day at 4.6% and first overnight at 27.8%. State Street runs the opposite path, from 36.6% to 9.5%. Vanguard, fourth by day, does not appear in the overnight top eight at all.
  4. Some funds do most of their business at night. KORU clears 44.0% of its ATS volume in the overnight session and MUU 42.9%. For SPY and IVV the figure is under 1%.
  5. This is a small market. Overnight is 4.5% of ETF ATS notional, and ATS is itself a minority channel. The night tape is structurally distinct, not economically large, and we are explicit about that below.

Part I established that ATS venues carry about an eighth of daytime ETF volume, with most of the rest split between lit exchanges and dealer off-exchange prints. This note removes the first two channels and asks what is left. Read Part I.

Take the ETF tape and delete the exchanges. Delete the dealer prints too. What remains is the eight hours between the US close and the pre-market open, when the only venues running are three alternative trading systems built for the overnight session. Over 28 tier-complete weeks those venues cleared $239.9 billion in ETF notional. It is a rounding error against the $5.30 trillion the same funds traded across all ATS venues in the same period. It is also, by composition, a different industry.

One channel left

The overnight session is the one part of the trading day where the venue question answers itself. Lit exchanges are closed. The dealer internalization that carries a quarter of daytime ETF volume has no retail order flow to internalize at two in the morning. Blue Ocean, Bruce Markets and Moon are what is open, and for these hours an ATS is not a minority channel competing with exchanges. It is the market.

That makes the session a useful natural experiment. The same funds, the same reporting regime and the same measurement, with two of the three execution channels switched off. Whatever is different about the night tape is a statement about who is awake and what they are trying to do, rather than about routing economics.

Leverage takes over

Exhibit 1
Structure of ETF volume, all hours against overnight
Share of ETF ATS notional by product structure. 28 tier-complete weeks, December 2025 to June 2026.
All hoursevery reporting ATS89.1%7.7%Overnight8 PM to 4 AM ET43.3%45.6%11.1%Plain (1x)LeveragedInverse
Leveraged and inverse products run 10.8% of ETF volume across all reporting ATS venues and 56.7% in the overnight session. Source: Sapinover analysis of FINRA Rule 4552 data.

Across every reporting ATS venue and all hours, leveraged products are 7.7% of ETF notional and inverse products 3.1%. Nine dollars in ten go through ordinary one-times funds, which is roughly what anyone would predict from the shape of the industry. Restrict the same measurement to the overnight venues and leveraged jumps to 45.6% and inverse to 11.1%. Products geared at three times or more, on their own, are 32.0% of everything that trades overnight.

The reading we would offer, and we are labelling it inference rather than measurement, is that leveraged and inverse products are daily-reset instruments. They are designed to be held for a session and their tracking degrades over longer horizons, which makes them natural vehicles for someone positioning around an event that resolves before they can trade again. An investor in Asia reacting to a US earnings print at what is for them the middle of the working day is exactly that someone. We can see the footprint clearly. We cannot see the account behind it.

The leaderboard flips

Exhibit 2
Where the top overnight funds rank during the day
Rank by ATS notional, all hours on the left and overnight on the right. Log scale.
ALL HOURSOVERNIGHTSOXL #4#1 SOXLQQQ #2#2 QQQSLV #11#3 SLVTQQQ #15#4 TQQQSPY #1#5 SPYGLD #7#6 GLDSQQQ #22#7 SQQQMUU #42#8 MUUSOXS #36#9 SOXSAGQ #45#10 AGQEWY #12#11 EWYKORU #76#12 KORURank by ATS notional. Vertical scale is logarithmic. Amber = leveraged or inverse.
The funds that lead the night tape are mostly not the funds that lead the day tape. KORU is 76th across all hours and 12th overnight. SPY runs the other way, from first to fifth.

SPY is the largest ETF on ATS venues by a wide margin, at $1.27 trillion of notional over the window, roughly double the next fund. Overnight it is fifth, at $10.4 billion. Put differently, eight tenths of one percent of SPY's ATS activity happens in the overnight session. IVV, which Part I showed behaves almost identically to VOO during the day, sends under 1% as well and ranks 43rd at night.

SOXL runs the opposite way. It is fourth across all hours and first overnight, and the reason is not that it grows at night but that everything else shrinks: 32.5% of its ATS volume prints in the overnight session. KORU, a 3x Korea fund that ranks 76th during the day, does 44.0% of its business overnight and lands 12th. MUU does 42.9%.

The pattern is not simply leverage. SLV is a plain silver fund and it is third overnight, with a quarter of its volume in the session. GLD and EWY behave similarly. What SLV, GLD, EWY and KORU share is an underlying that keeps trading while US equities are shut, whether that is a metal with a live global benchmark or a Korean index in its own cash session. Leverage and offshore reference prices are two different routes to the same place.

The issuer inversion

Exhibit 3
Issuer share of ETF volume, all hours against overnight
Top eight issuers by overnight notional, with their all-hours share for comparison.
ALL HOURSOVERNIGHTDirexion4.6%27.8%ProShares3.6%16.8%iShares25.8%15.4%State Street36.6%9.5%Invesco12.6%9.2%Tradr0.6%3.2%GraniteShares0.6%3.1%Roundhill0.5%2%
Direxion moves from fifth to first and State Street from first to fourth. Vanguard, fourth across all hours at 5.3%, does not place in the overnight top eight.

Ranked across all hours, the ETF industry looks the way the AUM tables say it should. State Street leads at 36.6% of ATS notional, iShares follows at 25.8%, then Invesco and Vanguard. Direxion is fifth at 4.6%.

Overnight, Direxion is first at 27.8% and ProShares second at 16.8%. The two largest issuers of the day, State Street and iShares, hold 25% of the night tape between them against 62% by day. Vanguard, which is the second-largest asset manager on earth, does not appear in the overnight top eight. Three names most readers will not recognise from any AUM ranking, Tradr, GraniteShares and Roundhill, take 8.3% of the session between them, more than triple their combined daytime share.

How small this is

Everything above is a statement about composition, not about size, and the distinction matters. The overnight session cleared $239.9 billion of ETF notional over 28 weeks against $5.30 trillion across all ATS venues in the same period. That is 4.5%. Part I established that ATS venues are themselves a minority of the total tape, so the overnight session is a fraction of a fraction. Composing the two figures suggests something around half a percent of all ETF trading, though the two come from different samples and different units, so treat that as an order of magnitude rather than a measurement.

The reason to care is not the dollars. It is that this is the only window in which a particular kind of participant trades without competition from the institutional machinery that dominates the rest of the day, which makes the composition unusually legible. For an issuer, it is also the one part of the tape where a product can be a market leader without being a large fund.

What this shows, and what it does not

The composition figures are measured. The explanation for them is not.

What we measured
  • Where ETF volume printed, by venue and by week. FINRA Rule 4552 ATS transparency data, with the overnight session defined as the three venues that run it.
  • Product structure and issuer for every fund, from our own ETF reference data.
What we did not measure
  • Who is trading. The Asia-hours explanation for the leverage tilt is an inference from what the products are and when they trade. We do not see accounts, domiciles or routing.
  • Anything outside ATS venues. This is an ATS-only measurement throughout, which is appropriate for the overnight session because the exchanges are closed, but it means the all-hours comparison is also ATS-only and not the full consolidated tape.
  • Recent weeks. FINRA publishes Tier 1 on about a two-week lag and Tier 2 on about four, so the window stops at June 22, 2026 where the tier data is complete. Including partial weeks would understate volume and distort the mix.
  • Causation in either direction. We do not know whether these products draw overnight flow or whether overnight demand called them into existence.
  • Execution quality. Nothing here says whether the overnight session is a good place to trade. Spreads, depth and price improvement in these venues are a separate question we have not addressed.
  • Whether this holds through a different regime. One window, one volatility environment. The leverage share in particular moves with the market.

Methodology

Figures come from FINRA weekly ATS transparency data published under Rule 4552, aggregated across 28 tier-complete weeks from December 15, 2025 to June 22, 2026. The overnight session is Blue Ocean (MPID BLUE), Bruce Markets (BOSS) and Moon ATS (MOON), the three venues operating roughly 8 PM to 4 AM ET. All-hours figures cover every reporting ATS over the same weeks, so both sides of every comparison use the same source, the same universe and the same units. The universe is the 1,136 US-listed ETFs in our reference data with any ATS print in the window. Product structure and issuer come from our own ETF reference dataset. Ranks are by notional. These figures are ours and should be attributed to our analysis, not to FINRA.

This is Part II of a two-part series on where ETF volume actually trades. Part I covers the daytime tape and the channels that carry it. Deeper dives on the overnight session: the leverage tape, the flow lab, and the structural map. Sapinover builds intelligence on U.S. equity market structure and the overnight session. This analysis is for informational purposes only and is not investment advice or a solicitation.