ETF Overnight Record

BlueOcean (BOATS) alone printed $13.52B / 487M shares on June 23, ranking #1 of 157 sessions since September 2, 2025 on both metrics. The prior BlueOcean peak was $11.95B on June 8. This deep-dive isolates the ETF subset of the night and runs four independent analyses against it.

Single-name equities carried the remaining $7.97B, led by Micron (MU) at $2.69B and SanDisk, Marvell, Intel, NVIDIA behind it. The ETF complex was the second leg of the same trade: leveraged and thematic vehicles expressing the semis/memory thesis. The rest of this report is the ETF subset only.

Direxion's 3x semis bull/bear suite (SOXL and SOXS) was the night's expression vehicle. Roundhill's single-stock leveraged products (MUU on Micron, MULL on Marvell) carried the memory-thematic flow. Vanguard — the largest ETF issuer in the world — was a rounding error here at 0.8%. The overnight ETF tape is a different product universe than the one that dominates daytime AUM rankings.

One row is worth pausing on: Inverse 3x carried 72.9M shares against only $418M notional. That's SOXS (semis bear 3x) at $25-share prices being hammered with massive lot sizes as a hedge wrap on the long-semis trade. The share-count leaderboard and the dollar leaderboard tell different stories — which is why we rank by notional throughout the rest of this report.

The structure is sharper than the silhouette implies. Cluster C3 — only 17 ETFs — captures the actual trade: mid-leverage (~1.9x average) names like SOXL, MUU, KORU, DRAM that printed $3.35B. Cluster C4 is the large broad-index hedge layer (QQQ, SPY, EWY) for another $1.21B. Everything else — 361 ETFs across the four other clusters — added up to $0.99B combined. Sixty percent of the ETF notional fits in a basket of seventeen ETFs.

Catalyst events don't recruit new ETFs to the overnight tape — they amplify the names already there. The model is unambiguous: 88.5% of the explainable variance in overnight notional comes from each ETF's typical level of overnight activity. Leverage and inverse exposure barely move the needle once you control for habit. The reason leveraged ETFs dominate the tape isn't because their leverage attracts overnight flow on catalyst nights. It's because they are the ETFs that live on the overnight tape every night.

Methodological note: R² is in-sample. The point of this analysis is feature attribution, not forecasting. An out-of-sample fit would be lower; ranking of feature importance would be stable.

The correlation tree is one of the strongest results in this report. The 14 ETFs in Group 4 don't share an obvious surface attribute (different issuers, different leverage factors, different geographic exposures) but their overnight activity moves together. When semis catch a catalyst, that whole basket lights up as one trade — which is precisely what June 23 looked like.

/ Weekly run-rate

/ Tier 1 vs Tier 2 breakdown

/ Ticker-level head-to-head

The aggregate 37.9% figure averages across the full 1,137-ETF universe and is pulled down by hundreds of thinly-traded ETFs that show up on FINRA dark pools but not on overnight venues. Looking only at the ETFs that are actively traded overnight, the structural picture sharpens: for the top 25 ETFs by 4-week overnight notional, the overnight venues hold the majority of ATS flow in 16 of 25 names. Five names exceed 80% overnight share. The mega-cap broad-index complex (QQQ, VOO, SPY, IWM) is no exception.

Three structural findings emerge from running the comparison at the venue level rather than the per-symbol level:

1. 01

   The overnight ETF tape is a tactical venue, not an allocation venue.

   Two-thirds of notional was in 2x and 3x products. Issuer HHI of 2,619 is at the DOJ “highly concentrated” threshold, and the concentration is not in the Vanguards and BlackRocks. It is in Direxion, ProShares, and Roundhill — the leveraged and single-stock-thematic issuers.

2. 02

   Catalyst nights amplify existing names. They do not recruit new ones.

   The Random Forest is unambiguous: 88.5% of the predictive signal for overnight notional is the ETF's own habitual overnight activity. Building a franchise on the overnight tape on normal nights is what captures the catalyst nights. Issuers absent from the overnight tape during quiet periods will not appear when the trade ignites.

3. 03

   Levered and single-stock ETFs are a separate liquidity universe.

   SOXL, MUU, TQQQ, DRAM, KORU — they trade lit and on overnight ATSes. They do not print in UBS, Morgan Stanley, JPMorgan, or any of the 14 other institutional dark pools FINRA covers. This is a structural distinction by product type, not just by time of day. Workflow desks routing levered ETF flow cannot rely on conventional dark-pool sourcing — the inventory is not there.

4. 04

   The overnight ATSes are no longer a niche. In Tier 2 ETFs they are at venue parity with the daytime dark-pool complex.

   Same 4-week window, head to head: 3 overnight venues processed $32.6B in ETF notional vs the 16 institutional dark pools FINRA publishes at $53.4B. 38% of all tracked institutional ATS ETF flow now goes overnight, and that is a floor. In the Tier 2 category specifically (leveraged, inverse, single-stock, niche), our 3 overnight venues moved $18.6B against the FINRA 16's $20.3B - 47.8% overnight share. The overnight session has crossed the threshold where it can be modeled separately from daytime; for ETF flow desks, routing and benchmarking are no longer separable problems.

Download requires an active Sapinover session (any tier including free trial). Anonymous requests are redirected to the subscribe page.