At Night, the ETF Industry Looks Nothing Like the Day

Sixty-one percent of overnight ETF notional on US-listed ATS venues moves through a leveraged or inverse product. Direxion is the number one issuer, not iShares. One ticker, SOXL, prints more notional in this window than every sector ETF combined. The conventional ranking of the ETF industry inverts at night.

The ETF industry is usually narrated through assets under management. By that lens, iShares, Vanguard, and State Street SPDR sit at the top of every league table. But AUM is a stock measure. The overnight tape is a flow measure, and on this dataset the rankings come out very differently.

The numbers below come from the Sapinover overnight ATS pipeline covering BlueOcean, Bruce Markets, and Moon ATS, filtered to US-listed ETFs printing at least $50,000 of overnight notional. The window is the last thirty trading days through May 26, 2026.

// PART 01The Issuer Inversion

The top issuer on the overnight ETF tape is Direxion at 35% of all overnight ETF notional. The next four issuers, combined, do not quite catch it. ProShares prints roughly 12%. iShares is also near 12%. Invesco is about 10% (almost entirely QQQ). SPDR is about 9%. Vanguard, the second largest passive issuer by AUM, does not crack the top eight at 1.2%.

Click to expand

Both Direxion and ProShares are leverage specialists by product mix. They sell 2x, 3x, and inverse exposure on US equity indices, sectors, and single names. Together they account for roughly 47% of overnight ETF dollars. iShares ranks third, and despite running 74 distinct overnight-active products against Direxion's 56, it prints about a third of Direxion's notional. The product breadth advantage that defines iShares in daytime AUM rankings does not translate to overnight flow.

GraniteShares is the second-tier leverage specialist worth flagging. With 30 active overnight ETFs at $1.03 billion of notional, it sits sixth, ahead of VanEck and Defiance. GraniteShares runs the leveraged single-name books on NVDA, COIN, AMD, and a number of other large-cap names that retail wants to express directional views on after hours. Defiance is a similar pattern at $0.35 billion across 35 products.

// PART 02Leverage Eats the Tape

Classifying every overnight ETF by product type confirms what the issuer breakdown implies. Leveraged products account for 51.4% of overnight ETF notional. Inverse products add another 9.6%. The combined leveraged-and-inverse share is 61 cents of every overnight ETF dollar.

Click to expand

Broad market funds — the SPY/QQQ/IWM/DIA family — print 24.7%. That is the second-biggest slice and the only category that resembles a daytime ETF order book in structure. Commodity ETFs, almost entirely the precious-metals and crude complex, are 7.4%. Sector, fixed income, and crypto sit below 5% each.

The overnight ETF tape is not a passive-fund tape. It is a leveraged-bet tape with a passive-index core.

The 90-day window softens these numbers slightly. Across the past three months, leveraged-plus-inverse is 53% of overnight ETF flow rather than 61%, but the structural picture holds. This is not a one-month artifact driven by a specific event. It is the stable shape of the tape.

// PART 03One Ticker vs the Sector Slate

One number condenses the story. SOXL, the Direxion Daily Semiconductor Bull 3X Shares, printed $7.90 billion of overnight notional over the last thirty days. The entire sector-ETF universe — every State Street Select Sector SPDR, every iShares sector fund, every Vanguard sector index — printed $1.32 billion over the same window. One leveraged single-sector product is six times the combined overnight notional of the entire sector-ETF slate.

Click to expand

On its own, SOXL is 23.7% of all overnight ETF notional. Combined with its inverse pair, SOXS at $1.38 billion, the semiconductor leverage complex is roughly a third of the entire overnight ETF book. The next-largest single-name leverage products are SQQQ at $0.93 billion, the leveraged NVDA pair, MUU and NVDL, together near $1.63 billion, and the Tesla leveraged pair TSLL and TSLS. All of these are concentrated in technology and AI-adjacent names.

The flip side: the actual passive semiconductor sector ETFs — SMH at $0.49 billion and SOXX at $0.62 billion — combined are about 14% of SOXL alone. Investors who want overnight semiconductor exposure are using 3x leverage, not 1x sector funds.

// PART 04What Is Almost Absent

Three categories that anchor daytime ETF activity are nearly invisible overnight.

The pattern is consistent across the three absences. When an asset class has a natural overnight venue elsewhere (futures for rates, spot or perps for crypto, individual stocks for sector views), the ETF wrapper does not get used. ETFs are used when there is no liquid alternative for the exposure being expressed, and that exposure overnight is overwhelmingly leveraged single-name or index.

// PART 05Three Implications

The shape of the overnight ETF tape has consequences for three distinct audiences.

A leveraged single-name product can move three times the dollar volume of the underlying sector while being entirely invisible on a share-counted league table.

// PART 06Three Regimes in Eight Months

The snapshot above is one regime. The full eight-month dataset contains at least three, and the contrast between them is itself the story.

Click to expand

Fall 2025 (Sep – Nov) was a leveraged-and-inverse base case, similar in shape to the current window. Leveraged share averaged the high thirties to mid forties. Broad market filled another quarter to a third. Commodity and crypto wrappers were minor contributors. Fifteen of the dataset's top regime-pure days (one category > 50% of that day's ETF flow) sit in this period.

Winter 2026 (Jan – early Feb) is the most dramatic event in the entire dataset. Silver and gold ETFs printed enormous overnight notionals through January and into the first week of February. SLV appears six times in the dataset's top-twenty single-day single-symbol notionals, all inside a six-week window. GLD appears three times in the same window. The single biggest day was February 2, 2026, when SLV printed $980 million and GLD printed $810 million in the same overnight session — together more than $1.79 billion in two metals on a single tape.

On Feb 2, 2026, SLV ($980M) and GLD ($810M) cleared $1.79B of notional in a single overnight session — the largest two-metal print in the dataset.

The macro context lines up exactly with the tape. Silver had run 68.5% in the first month of 2026; gold added 29.5% in the same window. On January 30, the Trump administration nominated Kevin Warsh as the next Fed chair. Markets read it as hawkish. CME hiked margin requirements that same day. The combined effect was a forced liquidation: SLV fell roughly 28% intraday — its worst single-day drop since 1980 — and GLD collapsed from $470 to $445. The February 2 overnight blowout we see in the tape is the rebound trade. The cash session that day recovered spot silver by 6% and spot gold by 5.6%, and the overnight ETF wrappers cleared near-record notional as participants re-established or reversed positioning.

After mid-February, commodity share declined steadily as metals consolidated. By the most recent thirty days, commodity ETFs are 7.4% of overnight ETF notional, down from peaks above 30% on the heaviest days of the Winter 2026 regime.

Spring 2026 (late Mar – May) is the current leveraged-single-name regime. Direxion's market share of overnight ETF notional rose from 21% (all-time) to 25% (90-day) to 35% (30-day) — a near-doubling concentrated in the last two months. ProShares and iShares each lost roughly half of their respective shares over the same window. The catalyst stack: an AI capex narrative driving semiconductor exposure (SOXL), the NVDA earnings cycle dragging single-name leverage flow forward, and continued retail demand for 2x and 3x products on the largest US technology names.

One day in this regime stands out as a cross-asset macro pivot. On April 2, 2026, QQQ ($520M), SPY ($440M), and SOXL ($460M) all printed in the top-twenty single-day notional list simultaneously. Broad-and-leveraged co-movement of that scale in a single overnight session is the signature of a market repricing both directional and beta exposure together. That kind of day is rare in the dataset — most sessions are either broad-driven or leverage-driven, not both at once.

Click to expand

The window-comparison chart makes the regime shift visible at a glance. Leveraged share moves from 40% (all-time) to 39% (90-day) to 51% (30-day). Commodity share collapses from 19% to 15% to 7% across the same windows. Inverse moves with leveraged but at a lower amplitude. The structural picture — that the overnight ETF tape is a leverage tape with a passive-index core — holds across all three windows. But the specific mix depends heavily on what macro is happening in the broader market.

// PART 07Methodology

The dataset is the Sapinover overnight ATS pipeline, which aggregates session-level prints from BlueOcean ATS (MPID OCEA), Bruce Markets, and Moon ATS. All three venues operate during the same overnight session window, approximately 8 PM to 4 AM ET, covering the gap between the US regular-session close and the start of US pre-market trading. BlueOcean had first-mover advantage in this window and remains the dominant venue by adoption and notional, which is reflected in the aggregate mix. Bruce and Moon are smaller but operate the same hours and trade overlapping symbol sets.

The thirty-day window for the headline numbers runs April 26 through May 26, 2026 — thirty consecutive trading sessions ending the day before publication. The 90-day window runs February 25 through May 26. The all-time window covers September 2, 2025 through May 26, 2026, a total of 138 trading sessions. The universe is US-listed ETFs as identified by the Sapinover symbol master (Asset_Type = ETF), with a $50,000 minimum notional filter applied at the row level to suppress noise from very-low-volume prints. The filter is applied uniformly across all three venues.

Issuer classification is regex-based on the company name field of the symbol master, using the same forty-five-pattern dictionary that powers the ETF Breakdown panel on the Sapinover session summary page. Product-type classification (Leveraged, Inverse, Broad Market, Commodity, Sector, Fixed Income, Crypto) is derived from a combination of the company name, the leverage multiplier field, and a small set of keyword tests. Edge cases — products that are both leveraged and inverse, such as SOXS or SQQQ — are classified as Inverse to match common practitioner usage.

Macro context for January 30 and February 2, 2026 (Fed chair nomination, CME margin hike, SLV/GLD intraday crash and rebound) was cross-referenced against CNBC reporting on the metals selloff, the Federal Reserve's public communications archive, and contemporaneous market commentary. The Sapinover overnight tape captured both the crash and the rebound; the citations are external confirmation of why those days produced record overnight notional.