One Share,
Four Prices

A $100 stock closes on Friday. By Saturday afternoon a same-rights token trades it at $106 after company news. A custody-backed receipt sits at $104 because redemptions do not open until Monday. A synthetic changes hands at $107 because its market maker wants a premium for the risk. Three transaction prices, one company, and no rule that says which one is the stock. That scenario is no longer hypothetical in its parts. Every component now exists, is approved, or is scheduled.

Start with the number that ended the shrugging. In August 2026, trading between 8:00 p.m. and 4:00 a.m. averaged 144.6 million shares and $7.4 billion a day, with share volume up 359% year over year. An average of 4,208 NMS stocks traded overnight each day. Overnight equities are no longer a novelty. They are also not yet a new market structure. They are the leading edge of one.

The consequential change is not blockchain by itself. It is fragmentation across clocks, legal wrappers, price-forming mechanisms, and settlement systems, without the connective tissue that Regulation NMS built around a single security. The old market fragmented venues. The new market can fragment the representation of the security itself.

The title above cheats slightly, and the cheat is the point. Sometimes these are four prices for one security. Sometimes they are four prices for four legally different things that a screen invites an investor to treat as one. Telling those two cases apart, before comparing any prices, turns out to be the whole problem. Here is what a weekend looks like once all four constructions are live:

Who actually trades the night

The SEC's own staff, working from CAT rather than the public tape, has now published the first real census of the after-dark market, and the anatomy is nothing like the day. The staff splits the 24-hour "modified trade date" into five sessions: overnight (8:00 p.m. to 4:00 a.m.), early extended, regular, late extended, and non-business day. Each has its own crowd.

Read together, the census says the night is a retail and foreign-account market wearing an institutional venue's clothes. The top of the dollar tape is Micron, SanDisk, Nvidia, and the leveraged semiconductor complex; the top of the share tape is three-to-five-cent Asian-domiciled names. Ninety-four percent of overnight dollar value is US-domiciled while 37% of the shares are not. Market quality matches the crowd: at the roundtable, Citadel Securities put overnight spreads at two to three times the regular session, with median displayed depth of $15,000 against a median order of $800, and two thirds of executing orders arriving non-marketable. DriveWealth said 60% of its overnight flow is new trading interest rather than displaced day flow, and noted foreign ownership of US equities is up 76% in three years. At least some of this is not displaced day flow. On the evidence available, extending the clock appears to be creating trading activity of its own.

The dark quadrant: what even the night tape misses

Here is the part the census made explicit. Of the 144.6 million overnight shares a day in August, the four reporting overnight ATSs, Blue Ocean, Bruce, IBKR Eos, and Moon, carried 111.5 million. Blue Ocean alone ran 93.7 million shares and $4.2 billion a day. The remaining roughly 33 million shares a day, close to a quarter of the night, printed off-ATS: over-the-counter, single-dealer, principal. That is the flow the largest wholesalers and banks internalize against their own books.

Two things about that quadrant, and then it can be left alone. It is not addressable liquidity. It is internalized volume, flow a market maker crossed against its own risk, and it was never something another participant could have traded against; counting it beside ATS volume flatters the night the way counting daytime TRF prints flatters the day. And nobody outside the Commission can reproduce the split at all. Principal reporting carries no day-or-night flag, and FINRA's weekly files are firm totals that are never cut by session. The overnight portion is known only because SEC staff went to CAT, which timestamps everything. That 33 million is not late. It is unavailable.

Three forces, arriving together

The collision comes from three independent projects landing at once: continuous trading, an explicit regulatory lane for on-chain trading, and institutional tokenization plumbing.

Time: the market outgrows the clock, and the weekend gets a cap table

Post-trade is already ahead of the tape. NSCC moved to 24x5 processing in June, and the extended consolidated SIP is scheduled for December 6, with the 8-to-9 p.m. hole shown in Figure 2. Exchanges have said they intend to route to the ATSs during the hours the SIP does not cover. The busiest hour of the night stays ATS-only, off the public tape, by design.

Then comes the weekend, and this time with names attached. On September 29, Bruce Markets announced the first continuous weekend session in US equities, extending its overnight ATS toward round-the-clock access, pending regulatory review and reported for early 2027. The announcement doubled as a cap table: a strategic investment round led by PEAK6, now the majority shareholder, and Robinhood, alongside Apex Fintech Solutions, Fidelity Investments, Nasdaq Ventures, NH Investment & Securities, tastytrade, and Webull, with Nasdaq supplying the trading technology and Apex Clearing the clearing and custody. Read the list again: the largest retail-flow originators, the exchange's venture arm, and the clearer, all on one weekend venue's side of the table. "The world doesn't take weekends off, and now neither will the markets," PEAK6's Jenny Just said in the release. Robinhood's chief brokerage officer made the demand thesis explicit: market-moving news does not check the calendar. Nobody assembles that roster for a science project.

Rails: a second book for the same ticker

On September 17 the SEC issued its innovation exemption, Release 34-106402, a five-year conditional order for "Tokenized Securities Venues." A qualifying venue can match buyers and sellers through permissioned automated market-maker pools without registering as an exchange or an ATS. The order is deliberately bounded: same-rights tokens only, issuer notice and a veto right, no primary issuance, symbol and volume caps (0.25% of average daily volume in Tier 1 names, 2.5% in Tier 2), a three-month pause on a second breach, and public trade disclosure within ten minutes. Synthetic price-exposure products are excluded.

Here is the plumbing break. The order lifts Rules 605, 610, 611, 612, and 613 from the venue itself. Its AMM price is not a protected quotation, does not feed the national quotation machinery, and does not have to peg to the SIP. Read the last one precisely, because it is the one that gets overstated: the venue is not a CAT reporter, while a broker-dealer that routed the order may still owe CAT a record of it. Anti-fraud law still applies throughout. But a tokenized-stock price and the conventional NBBO can now coexist while disagreeing, and the Commission says so plainly.

Plumbing: three flavors of the same share

"Tokenized stock" already means at least three economically distinct things. A native digital security makes the ledger part of the share's legal issuance. A custody-backed receipt locks a real share away and issues a token against it; the holder's claim is on the issuer, not the register. A derivative or exposure token gives contractual economics and, often, none of the shareholder rights. Robinhood's private-company tokens made that last distinction concrete in 2025, when OpenAI said the token was not its equity. The exemption adds a fourth idea: a token required to preserve the traditional share's rights while trading on a different rail. Same security, different mechanism.

The fragmentation dilemma

Put the constructions side by side and the problem stops being abstract.

The key insight is that price fragmentation and security fragmentation are different problems. US equities already trade across many venues; Reg NMS was built for exactly that, and it worked because everyone agreed which share was being traded. Tokenization can create two things that both say "IBM" while one is the common stock, one is a claim on a custodied share, and one is a derivative on an SPV. A national best bid cannot solve a taxonomy problem. Before you ask for the best bid, you have to decide whether two bids are bids for the same legal instrument. Call that step instrument resolution, and note where it sits: instrument resolution precedes price comparison. A consolidated tape answers where a security traded. It does not answer whether two economically similar objects are the same security, and no amount of speed in the second question rescues a wrong answer to the first.

And the referee is reconsidering the whistle. On June 11 the Commission proposed rescinding Rule 611's trade-through protections and the locked-and-crossed provisions of Rule 610(e). Protected quotations were the market's answer to venue fragmentation, the machinery that decided which price counted. The proposal is not final, and this article does not assume it becomes final. But the sequence deserves saying out loud: in the same season the Commission exempted a fifth construction from the order-protection regime entirely, it proposed to retire that regime's central rule for the other four. "Which price gets to be called the market" is not a side effect of tokenization. It is now an open policy question in its own right.

The fashionable word is interoperability. The less fashionable word is reconciliation, which is what people discover they needed all along.

What breaks when the four prices disagree

• Price formation becomes conditional. A book prices from competing orders; an AMM prices from pool state and a formula, which may not consider outside quotes at all.
• Best execution gets harder. A broker's duty does not vanish because the venue is exempt. Smart routing may have to choose not just among venues but among representations of the same company, only some of which the customer can actually convert between.
• Reference-price controls go circular. Moon's own filing says it plainly: "there is no NBBO during the overnight session." So Moon computes its own CBBO from its top of book plus whatever external market-center data it receives, and its 20% collar leans on a reference captured before the night begins. By Sunday that anchor is two days old, which is exactly what Figure 1 draws.
• Finality becomes overloaded. "Settled" on a ledger is not the same as legal title, a transfer-agent record, or a DTC position moving. Hong Kong's SFC already requires intermediaries to disclose whether on-chain or off-chain settlement is final, in its November 2023 tokenised-securities circular. US practice will need an answer as crisp.
• Surveillance splits. The exempt venue is not a CAT reporter; broker-dealer participants keep their own CAT obligations for the orders they carry in, and the venue's public trade feed sits apart from all of it. The trail exists, in pieces, in different hands. A manipulator does not need the markets to be legally identical. Correlation will do.

The regulatory-arbitrage question sits in that cap. At 0.25% or 2.5% of daily volume, one exempt venue cannot replace an exchange. But the cap limits volume, not the informational weight of a price. Price discovery is not allocated in proportion to turnover. The marginal trade after unexpected news carries more information than ten thousand routine prints at noon, and a small venue that is the only liquid market open at 3:00 a.m. or on a Saturday can become the reference for a much larger market that is closed. That is how 0.25% gets a loud voice.

The tip of the spear

The standards layer is not waiting for the dust to settle. A disclosure before the claim: Sapinover participates in the FIX Trading Community's digital-asset work, so read this section as a participant's account of a work program rather than an independent audit of one. FIX is the messaging language beneath most institutional orders and, quietly, most crypto matching engines. Its digital-asset sessions are working through how tokenized-security workflows differ across the native, receipt, and derivative constructions, and how an off-chain FIX order triggers an on-chain action through oracle and interoperability layers into token standards such as ERC-3643, with the confirmation finding its way back. Adjacent groups carry 24/7 session semantics, agentic AI in trading, expressive algorithm design, and the unglamorous SOP and ISO alignment that decides whether any of this interoperates or becomes seventeen incompatible pilots. Those are agendas rather than finished standards, and the distinction is worth keeping.

The pattern to notice: the bridge between TradFi and DeFi will be written in message specifications and identifier registries before it is written in headlines. Whoever defines how an off-chain order triggers an on-chain action, and how the confirmation flows back, owns the choke point of the fusion. The mosh pit is real, but somebody still runs the soundboard.

Which is the argument for leaning in now rather than waiting for the dust. The specifications drafted this year become the constraints everyone trades inside for the next decade, and they are being written whether or not the people who spent two years on their heels about 24/7 show up to the room.

Where Sapinover sits

Sapinover already measures the one venue-class that had no clean consolidated record: the overnight ATS tape, across Blue Ocean, Bruce, Moon, and IBKR Eos, reconciled to the FINRA weekly transparency data and read against each venue's Form ATS-N. We built the goggles because the tape did not exist and the question was real. The fragmentation now arriving is the same question, multiplied. As time, rails, and token wrappers split the same share across four constructions, someone has to answer five boring questions in milliseconds: what instrument is this, what rights does it carry, where is the best executable price, which ledger makes ownership final, and who is responsible when the records disagree.

That mapping layer, not any single AMM or exchange extension, is the piece the emerging market is missing. Measurement is the neutral vantage, and it is the one we keep.

The catalyst calendar