SAPINOVER OS v2.0 :: INITIALIZING...
000%
Nasdaq is buying LeveL. Cboe bought BIDS in January 2021. The two assets are opposite businesses, and the reason to own either one is about to change.
When we wrote about Nasdaq's agreement to acquire LeveL Markets, we treated it as an unusual event. It is, though the word unprecedented needs more care than we gave it. Exactly one US exchange operator has bought an NMS Stock ATS outright in the era when that category has had a legal definition, and every one of those qualifiers is carrying weight.
Cboe Global Markets completed its acquisition of BIDS Trading in early January 2021. Cboe described BIDS as the largest block-trading ATS by volume in the United States, said the purchase price was not material, and disclosed that BIDS had generated roughly $41 million of net revenue in the twelve months to September 2020. The strategic rationale was stated plainly at the time: the deal gave Cboe a foothold in the off-exchange segment, which even then accounted for more than forty percent of US equity volume. It followed a partnership the two firms had run in Europe since 2016.
Relax any single qualifier and other cases appear. Drop the word outright and you reach 2008, when the International Securities Exchange took a 31.54% stake in Direct Edge Holdings alongside Knight, Citadel Derivatives Group and Goldman Sachs at 19.9% each. Direct Edge ran as an ECN until its venues became exchanges in 2010, and the ISE Stock Exchange became a Direct Edge subsidiary rather than the other way round. Drop the word buying and you reach 2009, when NYSE Euronext and BIDS Trading launched the New York Block Exchange as a fifty-fifty joint venture, SEC-approved that January and closed in February 2013 for want of volume. The exchange-plus-BIDS combination was tried, as a partnership, twelve years before Cboe bought the firm.
Drop the word US and the claim fails outright today. The London Stock Exchange Group holds just over half the economic interest and about ninety percent of the voting power in Tradeweb, which is separately listed and which owns Dealerweb, an NMS Stock ATS that files its own Form ATS-N and carries 3.40% of ATS notional over our window. LSEG reached that position through its acquisition of Refinitiv rather than by buying a dark pool, which is the point: an exchange group can end up controlling one without ever setting out to. Drop the word buying in a different direction and TMX Group built rather than acquired AlphaX US, launched in January 2025 and running at 0.22%. And the pattern is not confined to operators with a US rulebook: the Tokyo Stock Exchange took a five percent stake in Blue Ocean Technologies, whose ATS runs the US overnight session, in 2023, two years before Nasdaq reached for LeveL.
Counting all four, exchange-affiliated venues already touch 15.30% of ATS notional rather than the 11.69% that BIDS and LeveL will hold between them. The honest version of our original claim is narrower and more interesting than the one we made: outright acquisition by a US exchange operator is rare, exchange ownership of an off-exchange venue is not, and the two facts point in different directions.
Nor did Nasdaq arrive at LeveL cold. Its own announcement records a strategic minority investment made in 2021, the same year Cboe closed on BIDS. Both of the largest US exchange operators took a position in an off-exchange venue in the same year. Cboe moved to control within months. Nasdaq waited five years.
Both are called dark pools. In the data they are opposite businesses. Measured across the 32 complete FINRA reporting weeks from 2025-12-15 to 2026-07-20, meaning weeks where both tiers had published and the market-wide totals are whole:
Figure 1 · The same label, opposite businesses
Four measures, LeveL on the left and BIDS on the right, each row scaled to its own larger value.
LeveL carries 7.96% of all ATS notional against 3.73% for BIDS, and reaches 13,238 symbols against 10,394. But the number that matters is the print. BIDS averages 551 shares and $34,194 per execution. LeveL averages 43 shares and $2,964. That is a factor of 12.9 in size, and it runs the other way in count: LeveL does 25 times as many trades on roughly twice the notional.
Figure 2 · Where each venue sits
Average print size against share of ATS notional, for the largest venues. The two acquisition targets sit at opposite ends of the same market.
So Cboe bought size, in 2021, and Nasdaq is buying flow. A block venue is a service business: relatively few executions, each negotiated, each large, revenue tied to the presence of natural counterparties on a given day. A continuous crossing book is an infrastructure business: constant small executions across nearly the whole tape, revenue tied to connectivity and habit rather than to any individual trade.
That difference shows up in stability, which was the core of our original piece. LeveL's weekly share varies with a coefficient of variation of 0.030. BIDS runs at 0.105, roughly three and a half times as volatile, which is exactly what a block business looks like when one large week lands or fails to.
Underneath the statistics are two different social structures, and this is the part the numbers only hint at.
BIDS is a room. Its heritage is institutional: two large asset managers sending indications of interest at each other, a negotiation that has historically run human to human or human to an automated firm-up, with the object being to put up a block that would move the market if it were worked in the open. Broker-dealers, central risk books and agency desks have since arrived to put up blocks of their own. The value proposition is legibility of the counterparty. You are interacting with somebody structurally similar to you, on the opposite side, and the venue exists to find that person.
The risk in that room is asymmetry of scale. If the counterparty is working a multi-day order and you are working a single block, you can be adversely selected for no reason other than being the smaller of the two. That is a size problem rather than a speed problem, and it is why block venues live or die on minimum quantities, firm-up protocols and the discipline of who is allowed to see what.
LeveL is exhaust. Its flow is broker-piped, the residual that sits between a broker's own internal ATS and the lit exchanges, plus retail exhaust passed along the chain. It is extraordinarily broad, reaching 13,238 symbols, and it is not of size: 43 shares a print is not a block by any definition. What it offers is the opposite of a room. It is a utility, and its ownership history reflects that, assembled through the 2022 merger of two operators and run as a broker-neutral venue rather than any single firm's pool.
So the same word covers a negotiation room where the point is knowing your counterparty, and a pipe where the point is that nobody needs to know anyone. Cboe bought the first, and what five years of exchange ownership did to it is measured in section 04. Nasdaq is buying the second.
Both venues describe themselves in public, in the same form, on a schedule the SEC sets. Form ATS-N has two amendment classes that matter here. An updating amendment is housekeeping. A material amendment is what a venue files when it changes what it actually does, and it carries a plain-English cover page saying what changed. Reading those cover pages end to end is the closest thing available to a changelog for a dark pool.
The counts are lopsided. BIDS has filed 35 Form ATS-N documents since 2019, of which 7 are material. LeveL has filed 80, of which 21 are material. In 2024 LeveL filed seven material amendments and BIDS filed none.
Figure 3 · Material amendments per year
ATS-N/MA filings only. BIDS Trading L.P., CIK 1368727, file 013-00031. LeveL Markets, LLC, CIK 1609177, file 013-00192. Counted from EDGAR submission histories.
The gap is a difference in kind rather than a scorecard. BIDS filed a small number of discrete additions to a stable architecture: actionable indications of interest between sponsors and sponsored firms in 2019, buyback instructions written against the Rule 10b-18 safe harbor in 2020, a cancellation feature for those indications in 2021, mixed-lot execution and an Overtime function in 2022, and port fees plus invitation notifications in 2023. The design underneath has not moved. Conditionals are non-tradeable messages the matching engine uses to locate a counterparty, and the venue, in its own words, “does not broadcast information in connection with Conditionals, it only sends directed signals.” That is a negotiation room, and it has been the same negotiation room throughout.
LeveL's filings read as continuous iteration on one question: who is allowed to meet whom. Transition management brokers admitted as a distinct subscriber class in 2021. Sponsored access and outsourced trading brokers in 2022. An extended firm-up conditional order in 2023. In October 2024, the ability for subscribers to rank potential counterparties and change the matching logic that would otherwise apply. In July 2025, a new segmentation regime.
That regime is worth quoting, because it is the mechanism this series keeps circling. LeveL partitions subscriber flow into channels “based upon trading behaviors, business units within the subscriber's organization, or a combination thereof and not solely an identifier such as an MPID.” The operator may also “in its sole discretion establish new channels or consolidate a subscriber's existing channels without notice,” giving as its own example a split between “passive resting” and “aggressive IOC” orders where it judges them meaningfully different. All channels, the form states, are set up by LeveL Markets and not by subscribers.
Segmentation by identity asks who sent the order. Segmentation by channel asks how the order behaves, and lets the venue answer.
Two more amendments land close to the transaction. In February 2026 LeveL admitted institutional accounts directly, stating that it “does not require all subscribers to be broker-dealers,” naming asset managers and hedge funds as subscriber types, and describing a routing service offered by Fidelity as a LeveL Markets affiliate alongside a clearing arrangement with National Financial Services. The same amendment records the cessation of the LeveLUp service, removes the counterparty-selection disclosures attached to it, and adds a waterfall to the Contra Priority List, under which an order steps from one counterparty set to the next by executed shares, elapsed time, or both. In May 2026 a further amendment added a minimum fill size and a maximum contra side order quantity, the latter available only in unilateral, bilateral and multilateral arrangements.
A cap on how large your counterparty may be is a direct answer to the asymmetry of scale described above. It is the sort of control a block room would offer. It now sits in the order specification of the venue we have been calling a pipe.
One filing on the other side deserves the same attention. In March 2026 BIDS filed a material amendment whose purpose was to disclose changes in access to confidential trading information between itself and its parent. It records that certain Cboe personnel provide shared services to BIDS and have access to historical confidential trading information, and that the historical confidential trading information of the BIDS ATS is stored on database systems that Cboe maintains. The same amendment waives fees on executions the subscriber has categorized as originating from a request for quote platform.
That is the ordinary texture of integration, and it reached the public precisely because BIDS filed a material amendment to put it there. It is also the only direct evidence available of what five years of exchange ownership looks like inside a separately operated pool, and the picture is specific rather than dramatic: the venue keeps running as a venue, while shared services and stored data accrete at the parent. Nasdaq has said LeveL will keep its own management team and maintain structural separateness. The BIDS file shows what five years of that arrangement produced at another operator, and it is a filing rather than a press release.
The last filing to note is a closure. On March 13, 2026 LeveL Markets filed a Form ATS-N-C for Luminex, the ATS it had merged with in 2022, with cessation effective March 27, 2026. The weekly tape agrees, and Luminex last reports for the week beginning 2026-03-23. Whatever Nasdaq is acquiring, it is not that venue. The buy-side block pool was wound down four and a half months before the agreement was announced, and the continuous book is what remains.
If the two venues are opposite businesses, the simplest place to see it would be in what they trade. It is not there. Of the fifteen largest names by notional at BIDS, 12 also appear in the fifteen largest at LeveL. Across the top fifty the overlap is 40. The sector mixes are close enough to be the same distribution, the largest single divergence being 3.6 percentage points, and the concentration profiles nearly match: the ten largest names are 15% of BIDS and 14.2% of LeveL.
The difference is granularity, and it survives every control we could put on it. Sorting both venues into market-capitalization bands and recomputing average print size inside each one, BIDS prints larger everywhere, by roughly eight times in the largest companies and thirteen to fourteen times through the middle of the market.
Figure 4 · Average print by market-cap band
Covered names only. Market capitalization is available for 37.8% of BIDS notional and 35.6% of LeveL notional over the window, so this is a comparison between the two venues rather than a profile of either.
So they are not fishing in different ponds. They are the same pond at different resolutions, and that is a stronger reading of the block and exhaust distinction than the one we started with. Exhaust is by definition the residue of the same orders. A parent order in a large technology name gets worked, part of it looks for a counterparty in a negotiation room, and the remainder is sliced small and piped. Both venues see that name on the same day. One sees it in hundreds of shares at a time and the other in tens.
The weekly tape averages every trade in a symbol week into one number, so a reader could object that the 12.9x print gap says nothing about actual blocks. FINRA publishes a second dataset that answers this directly: monthly block counts per venue, defined as executions of at least 10,000 shares and at least $200,000. On that strict definition, over the seven months from December 2025 through June 2026, BIDS printed 106,178 blocks averaging 27,276 shares, ranking #1 of 34 reporting ATSs with 20.06% of all ATS block share volume. The venue Cboe called the largest block ATS still is one, by roughly a fifth of the segment.
The surprise is the venue in second place. LeveL, at 12.24%, printed 66,493 blocks averaging 26,587 shares, nearly the same size as a BIDS block. Blocks are 17.31% of BIDS share volume against 5.56% of LeveL's, a gap of 3.1 times. That is the honest measure of block intensity, and it is far narrower than the 12.9x average print gap suggests, because averages are dragged by the small prints that dominate each venue's trade count. Definitions matter here: BIDS is 24.94% blocks by the share test alone, 23.79% by the dollar test alone, and 17.31% by both. Any block figure quoted from this dataset should say which test it used.
The history explains the small weekly prints. In 2019, blocks were 53.1% of BIDS share volume. By the first half of 2026 that share had fallen to 17.4%. But the block book itself barely moved: 16,991 blocks a month then, 15,874 now, down just 7%. Total volume around it nearly tripled, from 893 million shares a month to 2,493 million, up 179%. The denominator exploded and the room stayed the same size. That decline was filed before it was printed: the mixed-lot execution of 2022 and the request-for-quote fee waiver of March 2026, both in the amendment record above, are the doors the non-block dollars came through. LeveL converged from the other direction: its block share more than doubled in a single step, from 2.34% in 2023 to 5.8% in 2024, and has held near that level since. The same convergence section 03 found in the filings, a pipe adding block controls while a room accretes flow, is visible here in the executions. BIDS is still the largest room in the market. What changed under exchange ownership is that the building grew a pipe around it, and the pipe now carries nearly five shares for every one negotiated in the room.
Figure 5 · The denominator, and the block book
BIDS total share volume per month (bars, left axis) against blocks as a share of its own volume (line, right axis), strict definition, 10,000 shares and $200,000 both required. FINRA blocksSummary, monthly, 2019 through June 2026.
Two limits on these measurements. The market-cap comparison in Figure 4 covers only the names our symbol master can match, 37.8% of BIDS notional and 35.6% of LeveL notional. And the block dataset is monthly and venue-level, with no symbol dimension, so it cannot say which names the blocks were in; its seven-month window overlaps the 32-week weekly window but does not match it exactly. Within those limits the two datasets agree: BIDS prints 12.9 times larger on average, and carries about three times the block intensity, in the same names.
Venue design is downstream of one question, and every innovation in off-exchange structure is an answer to it: what do I know about the party on the other side of my fill, and what can I do with that knowledge?
There are broadly two answers. The first is to make the counterparty legible, by sorting participants into segments and letting each side choose what it faces. A venue that groups principal electronic market makers in one tier and retail or institutional interest in another is selling exactly that legibility. The second answer is to make speed irrelevant, by slowing the match down. A periodic auction converts a race into a negotiation. So does a conditional invitation, or a firm-up window. Both approaches attack the same problem from opposite ends, and almost every recent ATS design is some combination of the two.
The reason this matters is that the US equity market is not one population of traders. It is five constituencies with genuinely different problems, and flow moves in the direction that solves the problem of whoever is holding it.
| Constituency | What it wants from a venue | Where it prefers to be |
|---|---|---|
| Principal electronic market makers Citadel Securities, Jane Street, Hudson River Trading, IMC, Virtu and peers | To know who it is facing and price accordingly, tuning interaction to a measured signature of how that flow behaves after the fill | Retail payment arrangements, single-dealer platforms, hosted rooms, and any ATS whose segmentation it can profile |
| Broker-dealers agency flow, riskless principal, central risk books | To internalize first, then reach outward at declining preference | Own ATS, then other ATSs, then the exchange as the residual |
| Institutions asset managers working parent orders | Size without signaling, and a counterparty of comparable scale | Block venues, conditional and auction mechanisms, broker GUIs and algorithms |
| Retail held orders through a retail broker | Price improvement and certainty, decided by the router rather than the investor | Wherever the broker routes it, predominantly wholesale internalization |
| Everyone, together | Nothing in particular. The open field. | The lit exchange |
That last row is the point. The exchange is the one place where all five interact at once, which is what makes it a fair venue and also what makes it the hardest place to optimize. If your business is pricing counterparty risk, an undifferentiated pool is the worst environment you can trade in, because you are quoting the average of a population you cannot see. Every off-exchange structure of the past decade, from payment arrangements to single-dealer platforms to hosted rooms to tiered ATSs, is a way of replacing that average with something specific.
The five rows are not a census of investor types, which is why the list is short and stays short. They classify orders by how they reach the market, and every segment a reader might think is missing folds into a row once that is the question being asked. The registered investment adviser is the cleanest case. Its held order, routed by a custodial platform, is the retail row in every respect a venue can observe: the router decides where it goes and the adviser has surrendered discretion over the execution. Its not-held order, worked through a broker's desk or algorithms, is the institutional row. The same adviser in the same account can occupy two rows in one afternoon, and the held or not-held instruction is the entire difference. Hedge funds split the same way. Most are institutions working parent orders; the few that run their own execution infrastructure trade like principal market makers, and venues profile them accordingly.
The amendment record in section 03 shows the venues read the map the same way. Corporate buybacks did not get a room of their own; they got Rule 10b-18 buyback instructions at BIDS in 2020, a compliance constraint bolted onto the institutional row. Transition managers were admitted by amendment in 2021 as a subscriber class with an opt-out, and outsourced trading desks followed in 2022. When asset managers and hedge funds themselves arrived as direct subscribers at LeveL this February, the filing slotted them into the existing structure of channels and counterparty lists. New segments keep arriving, and each one has been filed as an admission into an existing row rather than a sixth constituency.
Read the two acquisitions against that map and they stop looking like the same deal. BIDS serves the institutional row: legibility through scale and negotiation. LeveL serves the broker row: the residual of the internalization waterfall, priced as a utility. An exchange operator that owns both kinds of venue is no longer only running the open field. It is running the open field plus two of the structures that exist because the open field is hard to optimize.
This section is a reading of how the market is organized rather than a measurement of it. The constituencies are well understood by practitioners, the routing preferences follow from published order-handling disclosures and venue design, but the motives attributed to each group are our characterization, and no firm named here has described its behavior in these terms to us.
The strategic logic is defensive in both cases. Lit exchange market share in US equities is mature and expensive to move; two decades of competition on pricing, colocation and order types have left the marginal point of share costing more than it returns. Off-exchange volume is where an exchange operator can still grow its share of US equity execution, by owning venues on the other side of the line.
The concentration data makes the point sharper. Across 36 reporting NMS Stock ATSs, the top five hold 51.0% of notional and the top ten hold 71.9%. This is not a fragmented market where share can be assembled cheaply. There are perhaps six or seven venues worth owning, most of them captive bank pools that are not for sale. LeveL and BIDS were among the few large, genuinely independent, broker-neutral books available to a buyer.
If the Nasdaq transaction closes, the two exchange-owned pools would together account for 11.69% of off-exchange notional. That is a meaningful share of the segment to sit inside exchange holding companies, and it is a change in the structure of the market rather than merely a change in ownership.
Here is the part that separates 2026 from 2021, and the reason we think the timing of the Nasdaq deal is not incidental.
In June the SEC proposed rescinding Rule 611, the trade-through rule, along with Rule 610(e) on locked and crossed markets. The comment file closed on August 17. We read all 172 records the Commission hosts: Rule 611 appears in 95 percent of substantive letters and 610(e) in about three quarters, and the professional center of gravity is conditional rather than opposed.
Rule 611 is what gives a displayed quotation its compelling force. It requires order flow to respect the best displayed price wherever that price sits, which is the mechanism that converts an exchange's quote into an entitlement to interact. Remove it and the quote becomes information rather than an obligation.
In a market without that obligation, routing decisions rest more heavily on relationships, measured execution quality and negotiated terms. Many practitioners expect the consequence to be growth in bilateral arrangements: more internalization and more negotiated liquidity, plus access structures that were pointless while the trade-through rule made the lit quote unavoidable.
That is precisely the environment in which owning an off-exchange venue is worth more than owning a faster matching engine. An exchange operator holding a large, broker-neutral ATS owns a relationship layer and a routing habit that does not depend on the protected quote surviving. Read that way, both acquisitions are hedges against the same regulatory outcome, one placed five years early and one announced six days before the comment file closed.
There is a second reason, and it is the one we would watch most closely, because it is the only part of US equity market share that is not yet spoken for.
The overnight session is small. Measured on the same complete weeks, the three overnight-dedicated ATSs, BOATS, Bruce and Moon, together account for 2.18% of all ATS notional, or $636B. That is a rounding error against a market where the top five venues hold half of everything.
It is also growing faster than anything else in the off-exchange market.
Figure 6 · The one pool still up for grabs
Overnight-dedicated ATS venues as a share of all ATS notional, first eight weeks of the window against the last eight.
The overnight venues went from 1.51% of ATS notional in the first eight weeks of the window to 2.96% in the last eight, a rise of about 97 percent in roughly seven months. No mature venue in this market moves like that, because mature venues are fighting over allocated share. Overnight share is not allocated yet.
Note what the overnight session is made of. Every venue in that chart is an ATS. That is not an accident of history: the exchange rulebook is heavier in those hours, with quotation obligations, auction mechanics and protected-quote duties that an ATS simply does not carry, which is why 24X required exemptive relief to operate as an exchange overnight while three ATSs were already trading. An exchange operator that wants to be present in a 24-hour market has an easier path through an ATS than through its own rulebook.
Which is what makes Nasdaq's framing of the LeveL deal worth rereading. It described an always-on market, and announced the acquisition alongside a new business unit, Digital Liquidity Networks, built to house it. On our reading Nasdaq is acquiring a wrapper that can trade at hours its own rulebook makes awkward, in a segment where market share is still being created rather than merely traded between incumbents.
How much share is actually in play? The honest answer is that nobody knows, but the shape of the arrival is predictable from the map above. Overnight volume today is overwhelmingly retail and retail-adjacent, because held retail orders are the easiest flow to route into a thin session. The next two constituencies to arrive are the ones with the least structural friction: broker-dealer central risk books, which can warehouse a position overnight and unwind it in the morning, and institutions dabbling at the edges with small child orders while they build a record of what execution quality looks like at three in the morning.
On our reading, each of those two groups arriving in earnest is worth roughly another percentage point of ATS notional over the next year, which would put the overnight session somewhere near four percent rather than the 2.18% it holds today. That is a judgment rather than a projection from a model, and we state the reasoning so it can be argued with: the growth to date has come from a single constituency, the two constituencies most likely to follow are identifiable, and the infrastructure gates that held them back, clearing and the extended SIP, have both now resolved.
There is a gap in that session the block data makes visible. All three overnight venues are pipes. Combined, they printed 6,830 strict blocks over the seven months, 0.97% of their own share volume, while BIDS alone printed more than fifteen times as many during the day. If institutions do arrive overnight, they bring the problem the daytime market built block venues to solve, size without signaling, into a session where no negotiation mechanism exists at any hour. Nobody offers a conditional order at two in the morning.
Cboe already has one leg of that infrastructure on the SEC's docket. On May 27, 2026 the Commission approved EDGX's proposal to extend its equity trading day to 23 hours, five days a week, running Sunday 9 p.m. ET to Friday 8 p.m. ET with a one-hour operational pause each weekday evening. Cboe has stated a launch target of December 2026, subject to industry-infrastructure readiness. The exchange side of a 24/5 offering is therefore no longer a proposal, and Cboe is the only US exchange operator that has an approved 23-hour equity venue and a large block ATS under one holding company.
Which invites a speculation we will label as one. The regulatory step is smaller than the exchange one. EDGX changed its trading day through an SRO rulebook amendment filed under Section 19(b)(1) and Rule 19b-4, published for notice-and-comment and approved by the Commission on accelerated basis; the record above shows how that unfolded across two amendments and nearly two months. An ATS operator changes what it does by filing a material amendment to Form ATS-N under Rule 304, which posts on EDGAR and takes effect within a defined period unless the Commission acts. BIDS itself has filed seven such amendments since 2019, each shipping a discrete piece of functionality. Extending conditional invitations and firm-up windows into hours the venue already operates in would be a change of that same kind.
The tactically obvious path would tip a toe rather than plant a flag. A single subscriber-class opt-in, a narrow list of names, a defined evening window, an early-stage block instruction with a minimum quantity, all disclosed in a cover page paragraph the way LeveL's February 2026 admission of institutional accounts was. Test whether the demand is there before pricing the venue around it. Section 03 already documents the template: the 2022 mixed-lot amendment, the 2023 extended firm-up conditional, the 2026 waterfall on the Contra Priority List; each opened one door without rebuilding the building.
Nothing in the EDGX 23-hour filing addresses BIDS, and no operator has said anything about it. But the ingredients now sit together at one operator: an approved 23-hour exchange, a block ATS that already carries the mechanics, a filing regime that lets it add hours by cover page rather than by SRO rulemaking, and a public rationale for reaching hours a rulebook alone cannot cover. If the overnight session keeps growing, the first mover that brings a room to the night gets to define what institutional overnight trading looks like, and the 2021 acquisition would hold one more option nobody priced at the time.
Put the two forces together and the logic compounds. If Rule 611 goes, execution moves toward relationships and bilateral terms, which favors whoever owns the venue and the routing habit. If the session extends toward 24/5, an entirely new pool of share opens in the one wrapper that can reach it without a rulebook rewrite. An exchange operator that owns a large, neutral ATS holds a position in both scenarios; its own order book alone reaches neither.
Which returns the piece to its title. The interesting question was never why exchanges buy dark pools. It is what they turn them into. Cboe bought the largest block room in America, and five years later five of every six shares it prints are something other than blocks, while the room itself, measured in block volume, is intact and still first in the market. Whatever Nasdaq is buying today, the BIDS tape is the only preview anyone has of what it becomes. Ownership did not choose between the room and the pipe. It built the pipe through the room, because pipes scale and rooms do not.
One precedent is not a pattern. Cboe and BIDS is a single prior transaction, in a different rate environment, at a price Cboe called not material, involving a venue with a fundamentally different business model. Reading it as a template for what Nasdaq is doing is an interpretation, and two data points can always be connected by a line. The nearest thing to a second data point cuts against us: NYSE Euronext and BIDS ran the New York Block Exchange as a joint venture from 2009 and closed it in 2013 for want of volume. An exchange pairing with an off-exchange venue is not a guarantee of anything.
Rescission may not happen, or not in this form. The comment file is genuinely split. Several exchanges support rescission only with conditions, at least one major wholesaler opposes it outright, and the asset managers largely asked the Commission to slow down. A threshold-based or phased outcome is realistic and would blunt the bilateral thesis considerably.
The overnight growth rate cannot continue and may not be structural. Doubling from a base near one and a half percent is easier than doubling from ten. Part of that rise reflects venues maturing and adding symbols rather than genuine demand migration, and a single volatile stretch of overnight sessions can move a seven-month trend line materially at this scale.
Our participant map is a simplification. Real firms occupy several rows at once. A large bank is simultaneously an agency broker, a central risk book, a market maker and the operator of its own ATS, and its behavior in any given interaction depends on which desk is holding the order. The taxonomy is useful for reasoning about flow direction and misleading if read as a description of firms rather than of roles.
Exchanges may buy ATSs for duller reasons. Revenue diversification, an accretive multiple, a technology team, or a customer list are all sufficient motives on their own, and none requires a regulatory thesis. The market-structure reading is ours, and neither company has framed it this way.
The window. All venue figures come from FINRA Rule 4552 ATS Transparency Data at the security level, across 32 complete reporting weeks from 2025-12-15 to 2026-07-20. FINRA publishes Tier 1 on roughly a two-week lag and Tier 2 on roughly four weeks, so any week missing Tier 2 market-wide is excluded entirely rather than partially counted. This is the same method as our original LeveL piece, so the two are directly comparable.
A trap worth naming. The MPID LATS belongs to The Barclays ATS, not to LeveL. LeveL ATS reports as EBXL. A first pass of this analysis matched on the wrong ticker and produced an entirely incorrect comparison before the venue name was checked. Any venue-level work on this dataset should resolve MPID to ATS_Name before drawing a conclusion.
On rank. Our original piece described LeveL as the third-largest ATS, which is by share volume. Over this slightly longer window LeveL is third by shares and fourth by notional, because Sigma X2 edges it on dollars. Both statements are correct and the measure is stated wherever the rank appears.
The precedent, and its qualifiers. “NMS Stock ATS” is a category created by the 2018 Regulation ATS amendments that introduced Form ATS-N, so claims of the form “the only one” are bounded by that date. The ISE stake in Direct Edge Holdings is from the completion announcement of December 2008, which states the 31.54% and 19.9% figures and records the ISE Stock Exchange becoming a wholly owned subsidiary of Direct Edge Holdings; the SEC published a related notice of filing in the Federal Register that November. The New York Block Exchange was approved by the Commission in January 2009 and NYSE Euronext announced its closure effective February 2013. LSEG's position in Tradeweb is from Tradeweb's own Form 10-Q, which reports LSEG holding a majority of the economic interest and roughly ninety percent of the combined voting power, with Tradeweb in turn the majority member of Dealerweb LLC. Dealerweb and TSX Alpha U.S. both file Form ATS-N and both report under NMS Tier 1 and Tier 2, so both are counted on the same footing as BIDS and LeveL. Nasdaq's 2021 minority investment in LeveL is stated in Nasdaq's own announcement of the acquisition. The Tokyo Stock Exchange's five percent investment in Blue Ocean Technologies is from the August 7, 2023 joint announcement of the partnership; Blue Ocean's ATS supplies part of the overnight data used in this piece, and Blue Ocean is a Sapinover data partner.
Filings. Amendment counts and every quotation in section 03 come from EDGAR: BIDS Trading L.P., CIK 1368727, file 013-00031, and LeveL Markets, LLC, CIK 1609177, file 013-00192, the latter formerly filing as Kezar Trading. Only ATS-N/MA filings are counted as material. Each change is attributed to the material amendment that made it, never to an updating amendment that later restated it. The Luminex cessation is Form ATS-N-C filed March 13, 2026, effective March 27, 2026, and it is corroborated independently in the weekly tape, where Luminex last reports for the week beginning 2026-03-23.
Block data. Block figures come from FINRA's monthly blocksSummary dataset, MPID by month by size bucket, June 2016 to June 2026. “Block” unqualified means the strict bucket, at least 10,000 shares and at least $200,000 in the same execution; the share-only and dollar-only definitions give higher figures and are quoted as such where used. The dataset has no symbol dimension and its monthly window overlaps but does not exactly match the weekly tape's window.
What weekly averages cannot show. FINRA publishes one average print per security per week, not individual executions. Any single large trade inside a week of small ones is averaged away before publication. Print-size figures here are therefore sound as a comparison between venues and unsound as a statement about whether either venue prints blocks. We use them only for the former.
Filing regimes are not the same. An exchange changes its rulebook by filing under Section 19(b)(1) and Rule 19b-4, which the Commission publishes for notice-and-comment and either approves or disapproves. An ATS changes what it does by filing a material amendment to Form ATS-N under Rule 304, which posts on EDGAR and, absent Commission action, takes effect after the operative delay in the rule. The practical consequence for the section 08 discussion: extending an existing ATS's mechanics into hours the venue already operates in is procedurally a smaller step than a rulebook change on an exchange.
EDGX 23-hour approval. Cboe EDGX Exchange, Inc. filed under Section 19(b)(1) and Rule 19b-4 on March 31, 2026; Amendment No. 1 filed May 26, 2026 superseded the original proposal in its entirety and was granted accelerated approval by the Commission on May 27, 2026. Cboe has stated a launch target of December 2026 subject to industry-infrastructure readiness. The overnight-block observation in section 08 is speculation, labeled as such in the body, and Cboe has not tied EDGX or BIDS to any block-venue plan in any public statement.
Sector and market-cap coverage. Company metadata is matched from our own symbol master, which covers 37.8% of BIDS notional and 35.6% of LeveL notional over the window. Both mixes are computed across covered names only and are reported as a venue-against-venue comparison, never as an absolute profile. The uncovered remainder is not assumed to resemble the covered part.
Overnight scope. Only BOATS, Bruce and Moon are counted as overnight-dedicated. Interactive Brokers' Eos ATS operates in overnight hours, but FINRA reports its notional without a session breakdown, so none of it is attributed to the overnight session here. Over this window IEOS reported $26.7B against $636B for the three dedicated venues, so the omission changes the overnight total by about four percent at most, and our overnight share is a floor. The larger IBKR ATS, at $1.01T across all hours, is a different venue and is likewise not session-split by FINRA.
