00 / The Same Number
Two constructions, printed the same day by different parties, returning the same number.
Strategy's announcement posted at 4:02 PM Eastern on July 23, carrying a set of net-of-claims metrics with definitions effective July 24. Computed from the figures published that afternoon, their Amplification and the tracker's return 1.5289x. Identical to four decimals. Every 1.52 and 1.53 and 1.54 circulating since is a timestamp, not a difference.
Under Strategy's new definitions the "Net BTC" numerator and the CEBE numerator are the same. Given the same instrument figures they return the same number, to the dollar. 843,775 BTC in the reserve, carried at $54,905M. Debt of $6,754M and preferred of $15,464M subtracted and a USD Reserve of $3,225M added back. A Net Reserve of $35,912M left standing behind the common equity.
The agreement is real, and it is conditional. The condition is visible in the price. MSTR closed at $93.63. At $93.63 every convertible note is out of the money and STRK is out of the money. Every senior instrument is a claim under both constructions. Nothing sits in the share count under one and the claim line under the other. The two numbers match because the market has not yet asked them the question that separates them.
That question is a rally. Strategy's new net of claims measurement treats a convertible instrument's classification as a function of where the common trades. In the claim line while out of the money and in the share count once it is not. The measurement moves when the moneyness moves. The CEBE numerator has no such switch. It reads the standing state of the claim and holds it, and the first instrument to cross its threshold in a rally is the point where the two figures separate. Only the price can rejoin them.
The constructions share a principle and differ in their definitions, and at this price those definitions read every instrument identically. Everything that makes the agreement fragile was published the same afternoon, in Strategy's own words.
01 / The Definitions
The definitions live on Strategy's own notes page, with a stated effective date of July 24, 2026, published the afternoon before.
Three of the new objects describe the reserve. Net BTC takes the bitcoin holdings and reduces them by the notional amount of the out-of-the-money convertible notes and other debt-like instruments, plus the notional amount of the outstanding perpetual preferred, excluding any in-the-money STRK, less the USD Reserve. Preferred denominated in a currency other than dollars is struck at exchange rates as of 12:30 PM New York time on the most recent Friday. Net Reserve is the same construction carried in dollars: BTC Reserve, less the same two notional deductions, plus the USD Reserve. Its definition carries a qualification in Strategy's own text, that the notional value of the preferred may not equal its liquidation preference or redemption amount, and that accrued and unpaid dividends are not included. Net BPS, reported in sats and in dollars, is Net BTC over Fully Diluted Shares Outstanding.
Fully Diluted Shares Outstanding is the fourth object and the one the other three depend on. It is basic shares outstanding plus the shares from exercise of all outstanding stock options, settlement of all restricted and performance stock units, conversion of all in-the-money convertible notes, and conversion of any in-the-money STRK. Strategy distinguishes it by name from Assumed Diluted Shares Outstanding, and states the distinction directly: apart from options, only in-the-money instruments are converted into the count, while out-of-the-money convertibles and preferred are deducted at notional in Net BTC instead. The in-the-money test is struck on the most recent price during market hours, which the company's own footnote says may differ from the closing price on the date presented.
Two existing names were redefined rather than retired. mNAV is now the market price of the class A common divided by Net Bitcoin Per Share in dollars. Amplification is now BTC Reserve divided by Net Reserve, computed on an assumed conversion of the in-the-money instruments followed by a hypothetical liquidation in which every preferred liquidation preference is assumed equal to its notional amount. Strategy notes that under this construction a rise in the bitcoin price lowers Amplification and a fall raises it.
Both redefinitions ship with matching, dated notes. Strategy states of each that prior to July 24th, 2026 the term referred to a different metric, and that "references to the Company's mNAV calculated prior to that date are not comparable to the Company's mNAV calculated after that date." The Amplification note repeats the sentence with its own name in place of mNAV.
Alongside these, the same release introduced Net Leverage, a BTC Hurdle Rate and a BTC Floor Rate, and rebuilt the MSTR, Bitcoin, and Credit tabs of the analytics pages.
The definitional detail, each new object set against its prior form in Strategy's language and the tracker's, sits on the tracker's credit mapping page at cebetracker.io/credit/ and is not reproduced here. What this section establishes is only that the objects examined below are Strategy's, published by Strategy, on dates Strategy set, and checkable at the source.
The agreements come before the divides, and they are not small.
02 / The Agreements
Start with what the two constructions do the same way, because the divides that follow only matter against this.
Claims subtract from the reserve. Debt and preferred stand ahead of the common, and both constructions remove them before stating what the common owns. Neither construction states a gross bitcoin figure as the shareholder's position.
Cash nets against those claims. Strategy adds the USD Reserve back after deducting debt and preferred. CEBE subtracts cash from senior claims before converting them to bitcoin. It is the same operation written in two directions, resting on the same reasoning, that a dollar of cash and a dollar of claim cancel until one of them is paid.
The residual is expressed per share. Both state it in satoshis. CEBE reports sats per basic share, and Net BPS reports in sats and in dollars.
The residual is allowed to reach zero. Strategy's notes state that Net BTC and Net BPS may be reduced to, or below, zero if senior claims net of the USD Reserve approach or exceed the value of the bitcoin holdings. CEBE carries the same condition as break-even, the bitcoin price at which claims consume the whole treasury and the common's residual reaches zero. A metric built on gross holdings cannot state that condition, because there is no arrangement of senior claims that makes a gross number go to zero. These two can state it, and both do.
Issuance accretes above one. Strategy's language sets the test directly: when mNAV is above one, issuing common stock to acquire bitcoin is expected to increase Net Bitcoin Per Share and therefore to be accretive, and when mNAV is below one such an issuance is expected to decrease it and therefore to be dilutive. That is the accretion threshold, fixed at 1.0, set on the net metric rather than floated against a gross one.
That last agreement is larger than it looks. What it replaced shows why. Under the prior construction, parity and accretion were two different points. Strategy's own first-quarter earnings presentation, published May 5, 2026, carries the figure on slide 59: issuing equity was accretive to bitcoin per share only above roughly 1.22 times mNAV rather than the 1.0 the market assumed (as of May 3, assuming a bitcoin price of $78,350). The slide states the breakeven condition in its own terms, "BPS accretion occurs when ADSO Market Cap > BTC Reserves," and its subtitle says the threshold "changes with capital structure changes." That is a threshold that drifts, moved by every new issuance. When the claims are subtracted out of the asset rather than added into the numerator, parity and accretion collapse onto the same point and no amount of preferred issuance moves it. The number a share has to beat before selling it adds anything to the holder behind it is now the same number in both systems.
So the netting agrees, the subtraction agrees, the unit agrees, the floor agrees, and the threshold agrees. What separates the two constructions sits upstream of all of it, in the question of when an instrument counts as a claim at all.
03 / The Switch
The two constructions sort instruments by the same rule at the level of principle. An instrument is a claim or it is shares. It is never counted twice, and it is never dropped. Strategy's notes make the rule explicit: apart from options, only in-the-money instruments convert into Fully Diluted Shares Outstanding, and the notional amount of out-of-the-money convertibles and preferred is deducted in determining Net BTC instead. Nothing is subtracted as debt while its conversion shares sit in the denominator. That is the same discipline CEBE runs on. The divide in this section is not about whether their construction is coherent, because it is.
The divide is about what decides the state.
Under Strategy's definitions the deciding fact is the market price of the class A common at the measurement time. Conversion price at or below market, the instrument is shares. Conversion price above market, the instrument is a claim carried at notional. The classification is a reading of where the stock trades. CEBE's reads the standing state, the state the instrument is in until something is converted or repaid, and holds it across the price. One construction asks the market what an instrument is. The other reads the contract.
At $93.63 the question is not live. Every convertible note sits above the market, and STRK's effective conversion price of $1,000, a tenth of a class A share for every hundred dollars of preference, sits far above it. Every senior instrument is a claim under both readings, which is why the two numerators are the same arithmetic on the same inputs, and why they returned the same figure on July 23.
A rally makes the question live, one instrument at a time, in ascending order of conversion price. The thresholds are filed and fixed, and they are published on Strategy's own shares page where each one ties to its issuance indenture. The 2030 A converts at $149.77, the 2028 at $183.19, the 2032 at $204.33, the 2031 at $232.72, the 2030 B at $433.43, the 2029 at $672.40, and STRK at $1,000. Every published share count reproduces exactly as principal times the indenture rate, which means the table can be recomputed from the filings by anyone. The three earliest were issued before the August 2024 ten-for-one split and their rates were multiplied by ten at it, which the page discloses in its own lead footnote, and the split is the only anti-dilution event in the history because the company has never paid a common dividend.
Each crossing is a step, and the steps run in a sequence that can be published in advance.
Take STRK, the highest and cleanest of them, in isolation against the July 23 figures. In sequence, the crossings run in the threshold order above. $1,402M of notional at a $100 liquidation preference is 14.02 million STRK shares, converting at a tenth of a class A share each, so 1.402 million shares enter the count at the moment $1,402M leaves the claim line. Hold bitcoin flat at the July 23 figures and run that single crossing. Net Reserve moves from $35,912M to $37,314M, a gain of about 3.9 percent. The share count moves from roughly 383.6 million to roughly 385.0 million, a gain of about 0.37 percent. Net BPS steps up by roughly 3.5 percent.
A holder reading that print sees the bitcoin behind each share rise 3.5 percent in an afternoon, while owning exactly what they owned that morning.
Nothing in the capital structure changed. No note was converted, no dividend was paid, no bitcoin was bought or sold. A share price crossed a number, and a measurement moved.
The step is also discontinuous. A cent below the threshold the notional is in the claim line, and a cent above it the notional is gone and the shares are in the denominator. A stock that crosses $1,000 and settles back below it returns the notional to the claim line, takes the shares back out, and returns the measurement to where it was, on a capital structure that again did not change.
Reversibility is the tell. A fact about a company does not un-happen when the price retreats. Bitcoin bought stays bought, a note repaid stays repaid, shares issued stay issued, and a measurement built on those facts moves once per event and holds between them. A figure that can step up and step back down without a single corporate action in between is reading something other than the company. It is reading the market's current opinion about the company, which is a valuation-layer object, and Strategy has placed it in the Net BPS measurement.
The term of that reading is intraday, by their own footnote: in-the-money status is determined by the most recent price during market hours, which may differ from the closing price on the date presented. So the classification is taken intraday. Two readers pulling the same dashboard on the same afternoon can be handed different denominators, and a figure printed against a date can reflect a state the stock was not in when it closed.
The lowest conversion threshold sits about 60 percent above the July 23 close. Strategy's own dashboard prints the volatility of the class A common: 85 percent over the trailing 30 days, 75 percent over the trailing year, implied volatility at 86 percent, securities as of July 28, 2026, 4:00 PM ET. At that volatility, a move of that size falls inside one standard deviation of a single year, and a stock at that volatility trading anywhere near a threshold does not cross it once. It crosses, returns, and crosses again.
Strategy's limitations note states that in-the-money and out-of-the-money treatment is determined by market prices as of the measurement date and "can change materially as those prices change." The volatility figures on the same site quantify how materially, and how often.
So the volatility of the class A common is what the capital structure is built to monetize on the claim side, and it is now also the input the denominator is keyed to. The measurement inherits equity volatility. A numerator of standing claims inherits none.
That limitation covers the market moving the price. It does not reach a second fact, which is that the company's capital program is among the forces that move it. Strategy's capital formation runs through equity issuance at scale, issuance at scale is not price-neutral, and near a threshold the discontinuity converts a small price effect into a large measurement effect.
The lever runs in both directions. Supply pressure that holds the price below a crossing keeps the notional in the claim line. Supply that carries the price through it moves the notional out and the shares in. Nothing here requires a claim about intent and none is made. The point is independence. Once the measured party's ordinary activity is among the inputs to its own measurement, the figure alone cannot tell a reader whether the balance sheet changed or the share price did. Nothing Strategy does short of an actual corporate action moves CEBE's measurement side, so the issuance test is computed on ground the issuance cannot shift.
The interaction reaches that test directly. mNAV places the class A price in its numerator, and the same price drives its denominator through Fully Diluted Shares Outstanding. And the ratio against 1.0 is the company's own published test for when issuing is accretive. The signal that governs the capital program is partly an output of the capital program. Run the lowest crossing at $149.77. Net Reserve steps up by roughly 2.2 percent as $800M of notional leaves the claim line. The share count steps up by roughly 1.4 percent as the conversion shares enter it. Net BPS steps up by roughly 0.8 percent, so mNAV steps down by roughly 0.8 percent, discontinuously, on a rising price. At the STRK crossing the step down is roughly 3.4 percent.
Strategy's accretion rule keys on mNAV above 1.0. A rally through a threshold can therefore move the issuance test toward dilutive while the price is rising, with no change in the capital structure. That is a property of the design and is stated as one.
The principle underneath all of it is one sentence. A measurement should be computable from facts the measured party cannot reach except by acting.
That is the first divide, and it operates on claims that have to be valued somehow before they can be subtracted at all. How they are valued is the second.
04 / The Ratchet
CEBE carries preferred at accrued liquidation preference. The reason is positional. The metric is taken from the common shareholder's seat, and what stands in front of the common is not what a preferred trades for, it is what the preferred is contractually owed before the common receives anything. A discount in the market does not reduce that entitlement by a dollar. Where a preference accrues, the accrued figure is the claim, because deferral of a dividend does not shrink an obligation, it enlarges one.
Strategy's construction deducts at notional, and their pages say so in two places.
The same number in section 00 had a second condition, and this is it. The two numerators returned the same figure on July 23 not only because every instrument classified the same way, but because notional and accrued preference are the same number today. No tracked series is in arrears, so outside ordinary payment-cycle timing, accrued preference and face coincide across the stack, and STRK sits at its $100 stated preference. The classification condition is the one the market is watching. This one is quieter and it moves on different inputs.
STRK is the live case. Under the July 2025 amendment to its Certificate of Designations, the liquidation preference is the greatest of the $100 stated amount, the last reported sale price of the STRK shares, and a ten-day average of those prices. Below par the stated amount governs and the preference rests at $100. Above par the preference follows the market up.
That is an upward ratchet, and it runs against the common in a recovery. STRK's $1,402M of notional at $100 stated is 14.02 million shares, so every $10 of STRK price above par adds roughly $140M to the claim standing ahead of the common. No issuance, no filing, no corporate action. A construction reading notional holds the line at $1,402M through all of it.
That asymmetry is the entire divide. When the preferred trades below par, notional and preference agree, and the two constructions agree with them. When the preferred trades above par, the preference moves and notional does not. The measurement can only understate the claim, never overstate it, and it understates exactly in the conditions where the issuer's position is improving and the common is being told it owns more.
There is a second edge to the same mechanism. A preferred trading below its stated amount can be retired below the claim it represents. This is a real and quantifiable benefit to the common if the issuer acts on it. That discount is perishable. It closes as the price recovers, and past par it inverts into the ratchet above. A figure carried at notional shows the same $1,402M whether the discount is wide, narrow, or disappears entirely. It cannot show the opportunity while it exists and cannot show it expiring.
Neither divide so far touches the denominator. Claims valued one way or another are still divided by a count of shares, and the two constructions do not use the same count, or the same kind of object. One of them is a fact. The other is a model.
05 / Fact and Model
Three objects travel under one word, and this piece keeps them apart.
Fully Diluted Shares Outstanding is a model. Its membership is conditional: basic shares, plus all options and all restricted and performance stock units, plus the conversion shares of whichever convertible notes and STRK shares happen to be in the money at the measurement time. The convertible and preferred portion of the count changes when the price changes, on the same switch section 03 describes. Two counts taken a week apart can differ because a price moved and for no other reason.
Basic shares outstanding is a fact. There is a number of shares in existence on any given date. It is countable, it does not depend on what the stock did that afternoon, and it is the count CEBE divides by.
The audited diluted count from the earnings-per-share footnote is a different fact. It is governed by accounting rules rather than by either construction. It currently stands at 333,913,000, with the anti-dilution floor active in a net-loss period, which holds diluted equal to basic because adding conversion shares would reduce the reported loss per share. It is also a weighted average across the reporting period rather than a count on a date, which is why it sits below a point-in-time basic count during a period of issuance. It is reported as the issuer's figure and never computed in house.
None of this makes a model illegitimate. A model is a model. Fully Diluted Shares Outstanding is internally consistent, its rule is stated, and Strategy names it distinctly from Assumed Diluted Shares Outstanding precisely so the two are not confused. The claim here is narrower: a measurement should say which kind of object sits in its denominator, because the reader is entitled to know whether the number under the line was counted or constructed.
That it is constructed can be shown from Strategy's own page rather than argued. Their mNAV printed 1.00x at $93.63, and mNAV is the class A price divided by Net Bitcoin Per Share in dollars. Setting Net BPS to $93.63 against a Net Reserve of $35,912M back-solves the count to roughly 383.6 million, with a band of about 381.6 to 385.5 million on the rounding of the printed ratio. Market capitalization printed $35,501M at the same price, implying about 379.2 million shares in the count the market cap is set on. Strategy's shares page, dated four days earlier, publishes basic shares outstanding of 379,160 thousand, which lands inside a rounding digit of the market-cap implication and confirms the method. The spread between the two counts is roughly four million shares, and at $93.63 no convertible note and no STRK share is in the money, so the entire spread is options and restricted and performance units.
The construction also states what it declines to consider, which is the most direct evidence it is a construction. Strategy's shares page carries a footnote defining Fully Diluted Shares Outstanding, and the definition is partly a list of exclusions. The count is not calculated using the treasury method, the accounting standard for diluted counts. It does not take into account vesting conditions on equity awards, the exercise price of any stock option award, or contractual conditions limiting convertibility. In-the-money status is set on the most recent price during market hours, which the footnote notes may differ from the closing price on the date presented.
Read those in order and the count's rule is fully specified. Every option enters whether or not the stock has passed its strike. Every restricted and performance unit enters whether or not its conditions have been met, which means a performance award is counted as earned before the performance is measured. Because the treasury method is declined by name, the cash an exercise would deliver never reaches the reserve the shares are being divided into, so the count carries the dilution without the proceeds that produce it. And the moneyness test that governs the convert portion reads an intraday print rather than a close, so the denominator attached to a given date is a function of when it was taken.
None of that is concealed and none of it is an error: diluted earnings per share under the accounting standard uses the treasury method because it is solving a different problem under a different rule, and an issuer building a dashboard is entitled to state its own.
The two counts also differ in where they live. Assumed Diluted Shares Outstanding appears in Strategy's filings, with its definition carried in a filed key-performance-indicator table. Fully Diluted Shares Outstanding, as of this writing, does not appear in any Strategy filing. Its definition, as of this writing, appears only on the website, dated to the July 2026 metrics release, and in no Strategy filing. The next quarterly report is its first opportunity to acquire a filed definition. Whether it does is checkable.
There is one term the two constructions do not divide on, and it belongs here because it is easy to assume otherwise. Strategy's own footnote states that basic shares outstanding is deemed to include shares sold under at-the-market programs, and shares to be issued on exercised options, vested units, and received conversion requests, that were pending issuance at the date presented. That is a rule rather than a census. The tracker's count, derived from the same weekly filings that disclose those at-the-market sales, is built on the same basis. Two parties had to write down what basic shares means, and they wrote down the same thing. Even the term everyone calls a fact turns out to have a construction inside it, and this is the one place both constructions built it identically.
Basic shares are the one input that is a count rather than a model. CEBE's own construction lives in its claim line, and it is stated in one sentence on a public page. The divide this piece tracks is never construction against purity; it is stated constructions, checked against each other.
A claims line carried at notional, divided by a modeled count, produces a figure. What Strategy's notes then do with that figure is the last divide.
06 / Supplement and Headline
The last divide is not in the arithmetic. Both constructions can now produce the same residual on the same inputs. The divide is where each party files it.
Strategy's Important Information section describes the net metrics as "merely a supplement." The construction is published, defined, dated, and placed on the dashboard, and it arrives with a stated rank beneath the figures it accompanies. The gross reserve remains the number the pages are built around. Total Reserve, the bitcoin count, the holdings figure the market quotes back: those are the headline, and the net construction sits under them as additional detail for the reader who wants it.
CEBE inverts the order. The residual after senior claims is the headline, and the gross holding is context for it. A treasury figure that has not had claims removed describes the company. The figure that has describes the share.
Neither ordering is arithmetically wrong, and the two systems are not in dispute about any number when they rank differently. A supplement and a headline can hold identical values. What differs is which figure a reader encounters first, which one anchors the impression, and which one has to be sought out.
That choice decides what a reader encounters first. The gross number is the larger one and it is the one that moves with bitcoin alone, so it reads cleaner. The net number carries the capital structure inside it, which makes it noisier and makes it fall when claims are added. An issuer building a page has reasons to lead with the first. A metric built from the common shareholder's seat has reasons to lead with the second. Both sets of reasons are legible and neither requires bad faith to explain.
The observation this piece makes is only this: the same arithmetic now exists in both systems at different altitudes, and altitude is a decision. It is made once, by whoever builds the page, and it is rarely stated out loud. Strategy stated theirs, in three words, in a section most readers will not open. CEBE stated its own in the name of the measure: the residual is the headline, and everything else is context.
Rank is where the constructions differ in posture. Inputs are where they still differ in fact. Both parties publish figures for the same instruments on the same dates, the figures do not entirely agree, and that ledger runs in both directions.
07 / The Ledger
If the constructions are what the preceding sections say, their published figures should re-derive from the filings by someone outside the building. This section does that, and reports where it could not.
The ledger runs both ways, and the first item is Strategy's.
Two Strategy surfaces print different debt figures on the same day. The credit tab shows $6,714M and the MSTR tab shows $6,754M. The gap resolves as composition. The credit tab carries convertible principal only, the MSTR tab adds roughly $40.2M of other secured principal, and each figure is correct for what it counts. The dashboard also carries debt at face on its own, which is the convention the framework uses and one fewer difference between the two constructions. A reader moving between two tabs sees two numbers and no note explaining the difference, which is an argument for one engine and not an argument about measurement.
The second item is a different category.
What Is Proven
Strategy's dashboard figure for STRC is $227,519,200 above the figure their own filings produce. The decisive arithmetic is theirs. Their quarterly report shows 29,587,063 shares outstanding at December 31 and first-quarter sales of 20,659,450, and those sum to the 50,246,513 the same report states as outstanding at March 31, exactly. The stub week of March 30 to 31, 2,275,972 shares, is therefore already inside the quarter-end base. That chain uses no data from outside the filing.
Two further chains support it. Building forward from the audited base with the weekly sales disclosures gives 102,618,733 shares, which is the verified figure. And those weekly disclosures check independently against the capacity ladder: consumption of $5,237.4M against summed notionals of $5,237.222M, a gap inside the rounding of figures reported to one decimal place.
The alternative explanation was issuance, and the filing covering July 20 to 26 forecloses it. Shares sold, notional, and net proceeds all print as dashes, and available capacity is unchanged at $17,510.8M. That is the tenth consecutive week without a STRC sale, the last one having been the week ended May 17. Nothing moved the line between then and July 20, so the print is the same figure whichever as-of date it carries.
What Is Inferred
That the excess is specifically the stub week counted a second time rests on the double-count model reproducing their figure to within 780 shares. That is a reconstruction of what happened. It is not a disclosure of it, and the residual is unexplained.
What the Repurchase Does
A buyback was in progress during the same week, and it discriminates rather than confounds. The whole retirement is $28,893,000, an eighth of the delta, so it cannot account for the figure at any lag. And had the July 23 print reflected retirement pro rata across the window's four trading days, the delta would have widened to roughly $250.6M. It reads $227,519,200, which is the stub and nothing else.
What Is Predicted
If the method is unchanged, the next print reads $10,460,499,500 against a filing-derived $10,232,980,300, and the delta persists at exactly $227,519,200. If it closes, the correction happened in cycle and this becomes a dated correction rather than a live item. I am publishing this prediction before the print, which is the only order in which it is a test.
Structurally this is a dashboard maintained on an inception-to-date ledger diverging from an audited quarter-end base. Any issuer running parallel surfaces can reach that failure mode, and a single-source engine is what prevents it.
The lens turns here, and it has to. Before my own audit, my sheet carried Strategy's figure to the share, because I had taken the figure from the dashboard rather than from the filing. That is how I found the trail. Their surface diverged from their filings. Mine diverged by inheriting theirs, and the audit that traced it is what produced the chains above. It is the strongest argument I have for the rule that a figure enters only from a primary source, and I learned the rule the same way I found this exhibit.
That is the whole ledger with this issuer, in both directions. It is not the whole of the sector.
08 / The Market's Opinion
Section 03 described the moneyness switch as a valuation object placed in the measurement layer. That is precise about where it sits and imprecise about what it reads, and the distinction matters here.
Strategy's switch consults one price, the market price of the class A common. It asks whether the common has risen above a claim's conversion price, and it answers in two states. What it never asks is what the claim itself trades for. Net BPS deducts the out-of-the-money preferred at notional whether that preferred changes hands at par, above it, or well below it. The toggle is binary and it is keyed to the wrong security to be a market read of the claims.
STRK is where the gap is largest and easiest to check. It carries at $1,402M of notional on Strategy's credit tab, and it trades near 60 cents on the dollar of its $100 stated amount. No figure on their pages reflects that. The claim enters Net BTC at full notional, exits at full notional when the switch flips, and is not carried at the price the market has been willing to pay for it.
The same instrument supplies the section's sharpest illustration. STRK's contractual liquidation preference is the greatest of three figures, one of which is its own last sale price, so the claim's legal size is defined by reference to a traded market. The one instrument whose contractual claim reads its own market price is the one place Strategy's measurement declines to consult one.
A second threshold lives in that same gap. Conversion becomes economically live for a STRK holder when a tenth of the class A price exceeds what STRK itself fetches, which at current prices is roughly $615 on the common. That figure moves whenever STRK moves. It is distinct in kind from the $1,000 contractual switch in section 03, which is fixed by the conversion terms and does not move at all. One threshold is filed and the other floats.
CEBE's does not consult those prices either, and for a stated reason. Preferred is carried at accrued liquidation preference because that is what stands ahead of the common equity contractually, and a measurement that marked claims to market would report a distressed issuer's obligations as shrinking exactly when they are hardest to meet. The market's read on the claims is real, and it belongs one layer up, in valuation, where the Adjusted Claims specification, public and dated at cebetracker.io/claims/spec/, defines it as a continuous, moneyness-weighted object rather than a two-state toggle.
A creditor-lens analysis published July 26 extended the issuer's construction to market-priced claims, reaching the same class of adjustment on a different share count; the difference is its own story. On July 28, the day before this piece published, Strategy's chief executive stated publicly, in a post on X, that the upgraded mNAV "accounts for all securities senior to MSTR" and "establishes a 1.0x threshold for accretive MSTR issuance." The market's opinion of the claims is still only published in one place.
09 / The Principle
The divides share a shape. The switch reads a price the issuer's own program helps move. The count is rebuilt when that price crosses a line. The claims line holds still while the contractual claim ratchets past it. Each one admits something into the measurement that is not a fact about the company.
A measurement should be computable from facts the measured party cannot reach except by acting. That is the net effect.
A note on scope. Strategy is not alone. A second issuer shipped its own net construction weeks earlier, netted differently, and that is its own story.
A note on sources. Every Strategy definition, footnote, and figure quoted or cited in this piece is from Strategy's public dashboard, shares page, credit tab, and filings, as of the dates given in the text. The Adjusted Claims specification is public and dated at cebetracker.io/claims/spec/.