Strategy (MSTR) Q3 2026 — Fewer Coins a Share, Fewer Claims Ahead

Strategy's stock rose 76% in the third quarter against 43% for bitcoin, and the company sold $5.41 billion of common stock. It put 11% of that into bitcoin. The rest retired preferred stock and filled a dollar reserve. Each share now holds 11.8% less bitcoin than in June and has 34% fewer dollars of net senior claims ahead of it (debt and preferred, less its dollars and software). Rolled forward on July's method, fair value rises from $94 to $124.79, almost all of it from the bitcoin price. Four changes to the method, each shown with its own effect, net to −$0.18: the preferred and the convertibles at market value, a tax on the unrealized gain, and dilution from employee awards. Fair value is $124.61. The $141.62 close of 29 October is 13.6% above that, a premium of 1.14× to net asset value.

TON618 Capital Research · As of 29 October 2026 · Download PDF · Download model

In July we valued Strategy at its net asset value, about $94 a share: bitcoin at spot, plus cash and the software business, less the convertible notes and the preferred stock. We gave the preferred-stock franchise no value of its own, because the premium that would make new issuance pay had gone. The third quarter tested that call from both sides. The premium came back, from 1.04 times NAV at the end of June to 1.13 times at the end of September, and the company sold more common stock than in any quarter this year. But it bought little bitcoin with it. It used the proceeds to buy back its 12% preferred below par, build $5.07 billion of dollar assets by 25 October, and sell no new credit at all. This note asks what that trade did to each share, and what $141.62 pays for.

Terms used throughout, defined once. Net asset value (NAV): bitcoin at the market price, plus Strategy's dollar assets and the software business at a comparables value, less the convertible notes and the preferred stock at their market values, and a 21% federal tax on the unrealized gain on the bitcoin. Our fair value is NAV per share at a premium of 1.00×. Like-for-like basis: where an exhibit compares two dates (the quarter's bridge, coin per share, the trade, the funding paths, the strike and wipe-out levels), the preferred is at liquidation preference and the convertibles at principal on both dates, no tax is deducted, and the exhibit says so. mNAV: price divided by NAV per share. The company's own figure leaves out the software and uses the preferred's notional amount. Two bars for selling stock. Above NAV per share ($124.61), new common adds net asset value per share. Above coin value per share (bitcoin held times the bitcoin price, divided by basic shares, $157.10), it also adds bitcoin per share. The note names the bar each statement uses. Both are the company's own: its presentation says that at an mNAV above one, new stock that buys bitcoin "would be expected to increase Net Bitcoin Per Share", and its BTC Yield measures the gross bitcoin per share. Net coin per share: NAV divided by the bitcoin price and the share count, the bitcoin each share would own after every senior claim. USD assets: the company's USD Reserve plus USD Cash. BTC Yield: the company's percentage change in bitcoin per assumed-diluted share, a measure that, by the company's own description, does not deduct the senior claims. Shares: basic class A and B shares, including those sold after the balance-sheet date, plus employee options and stock units by the treasury-stock method (425.4 million in all). All six convertible notes are out of the money at this price, so none is counted as converted in the base case.

The release reports operating income of $20.91 billion for the quarter and diluted earnings of $44.75 a share. Almost all of it is the $20.91 billion fair-value gain on bitcoin. Software revenue was $125.4 million, down 3%, at a 71% gross margin. Bitcoin rose 43% between the quarter-end prices, and MSTR rose 76%. The gap between the two has two sources, and the filed balance sheets separate them.

NAV per share rose 62%, from $83.91 to $135.86: more than bitcoin's 43%, because the common is a geared claim on the coin. The premium, mNAV, rose 9%, from 1.04× to 1.13×. So most of the 76% came from NAV and a smaller part from the premium. The price did run ahead of net asset value, but by about a tenth, not by the gap to bitcoin.

The flywheel is a different question: did issuance at a premium turn into more bitcoin per share? It did not. The company sold $5.41 billion of class A stock and no preferred. Shares outstanding rose from 371.6 million to 422.0 million, or 13.6%, in one quarter. About 50.4 million shares were added in all, including a small number from equity awards, at an average $107.35. At the quarter's average bitcoin price, $71,171, NAV per share on June's balances was $112.70. The common was therefore sold at roughly 0.95× NAV, and at 0.66× its coin value per share. Below 1.0× coin value, a dollar of new stock buys less bitcoin than the share it creates already owns. Every coin bought with it lowers bitcoin per share. The proceeds went elsewhere:

The rows do not sum to the proceeds: bitcoin sold and interest earned came in, and dividends and interest went out.

At the 29 October close the stock trades at 1.14× NAV on our definition (1.16× on the company's, which leaves out the software and the tax and uses the preferred's notional amount; management cited 1.15× at the 23 October close). So the price clears the first bar: common sold at it adds net asset value per share. Market capitalization over the bitcoin alone is 0.90×. Coin value per share is $157.10, 11% above the price, so the price does not clear the second bar: common sold at it still lowers bitcoin per share. The answer to the first question: the price ran modestly ahead of net asset value (the quarter's figures above are on the like-for-like basis), and the flywheel did not restart. The quarter's issuance was not bitcoin accumulation. It was a balance-sheet repair paid for in common stock.

The company's own measure records the cost. BTC Yield was −7.8% from the second-quarter print to 25 October. Year to date it fell from +4.5% to −3.7%, at 187,753 sats a share on about 452 million assumed-diluted shares. On basic shares the fall is larger. Each share held 227,662 sats of bitcoin on 30 June and 200,910 on 25 October, −11.8%. Holdings grew 0.3% over those four months and basic shares 13.7%.

That is half of the trade. The other half is what sits ahead of each share. Convertible principal and preferred liquidation preference, less dollars and the software value, came to $49.33 a share on 30 June and $32.31 on 25 October, −34%. The preferred buybacks and the reserve did that. The common now owns fewer coins, with fewer dollars owed ahead of it. Whether the share came out ahead depends on the bitcoin price at which the claims are counted. A fixed-dollar claim is a smaller bite of the stack when bitcoin is high and a larger one when it is low.

At $78,193 the trade cost each share 3.0% of its net coin, 164,580 sats to 159,588. At $32,000, the bear case's bitcoin price, it added +36%. At $120,000, the bull case's, it cost 6.7%. The company bought insurance against a fall in bitcoin and paid for it in the coin that would compound in a rise. The Q2 note's question was whether the preferred stack could be carried through a drawdown. The company's answer is to shrink it and pre-fund it, and that is a defensible choice for the credit. The cost of the choice falls on the common, and the cost grows the further bitcoin rises.

The path from here, with no bitcoin return assumed. Hold the bitcoin price, the share price and the premium where they closed, and fund the $1.51 billion annual fixed charge four ways. From the reserve, as the company now does, bitcoin per share holds flat and NAV per share falls 2.9% a year, the charge leaving as dollars. With new common at $141.62, bitcoin per share falls 2.5% a year. With bitcoin sales it falls 2.3%. Repeat the third quarter's pattern for a year ($5.41 billion of common a quarter, 11% of it to coin) and NAV per share edges up +1.5%, because stock sold above NAV adds to it. Bitcoin per share falls 24%. No funding path raises gross bitcoin per share at this price; issuing above NAV does raise NAV per share, which is the other bar. Doing so takes common sold above coin value, $157.10 a share today, or preferred sold to buy coin. Preferred-funded coin raises the company's measure at once, but it adds no net coin per share when it is issued, because the new claim offsets the new coin. It adds net coin only if bitcoin then rises faster than the coupon, and that is a forecast of bitcoin. The company's objective of doubling BPS, about 10.4% a year, rests on the second route. Phong Le described the plan on the call: digital credit is the primary way to acquire bitcoin, and the idea is to issue common stock above 1x mNAV to pay the dividends and interest on that credit. At 1.14× on our basis (1.16× on the company's) that common would be sold above NAV but below coin value. On its own, the dividend leg is the "new common" path above: bitcoin per share falls 2.5% a year at today's prices. The credit leg's coin raises bitcoin per share when it is bought, and raises net coin per share only if that coin outruns its coupon. Since June the company has not issued any preferred.

The buybacks changed the cost of the stack as well as its size. Strategy has bought back 20.98 million STRC shares this year at an average $97.39, retiring $2.10 billion of notional at about a 2.6% discount. At a 12% coupon on $100, each buyback retires a claim yielding 12.3% at the purchase price, while the quarter's interest income earned about 4.5% annualized on the company's cash and short-term investments. Those buybacks removed $252 million a year of dividends. At the preferred's notional amounts on 25 October ($13.35 billion, the company's figures by series) and its stated rates, the preferred stack costs $1,007 million a year for STRC and $468 million for the other four series. With $35 million of convertible interest, the total fixed charge is $1.51 billion a year. That is $3.57 a share, against the $1.76 billion management gave in July and the $1.55 billion the June note used. In bitcoin it is 19,300 coins a year, 2.3% of the stack at today's price, from about 3.3% in July.

The wipe-out and strike use liquidation preference, the claim that binds in a liquidation. The wipe-out price counts the dollars and the software against the claims. The company's own floor, $17,662, counts the dollars but not the software. Third-quarter preferred dividends were $419.6 million, from $400.7 million in the second quarter: the July STRC rate increase arrived before most of the buybacks. The annualized charge above is the forward figure at today's share counts.

Two features of the stack changed without changing its cost. From 1 November STRC has a dividend record date on every calendar day, and STRF, STRK and STRD follow on 1 January 2027. The 29 October 8-K states that the change "does not increase or decrease the total amount of regular dividends payable". It changes when the reserve pays, not how much. The return-of-capital guidance rests on the company not expecting positive earnings and profits, a regular-tax measure, for ten years or more. Separately, the Q2 10-Q says that under Treasury's interim guidance of September 2025 the company plans to exclude unrealized bitcoin gains when computing income for the corporate alternative minimum tax, and does not expect to become subject to that tax because of them. Neither change enters the NAV.

Whether STRC stays at par at 12% is the question of The Last Moat Is the Coupon (17 August), and its tripwire is scored in §5. One point bears on this note's valuation. Par has been held this autumn by buybacks paid from USD Cash, as three of the four October 8-Ks state and the fourth's balances show. On the quarter's sources and uses, our inference is that the cash came from the common sold in the third quarter. No new STRC was sold. A preferred franchise has value to the common only when new preferred can be sold at a cost below what the bitcoin it buys earns. So far the franchise has been supported, not sold.

June and July valued Strategy at NAV at spot bitcoin, the software at $1.4 billion, the convertibles at principal and the preferred at liquidation preference, on basic shares, at a premium of 1.00×. That engine on its own inputs reproduces July's table value of $93.59, published as about $94. Rolling each input to its 29 October value, one at a time, gives $124.79. This update then changes four things in how the value is measured, each its own row below and each argued after the table. The house carries no presumption in favor of the earlier method.

The order matters for the capital rows, and the note states it rather than choosing the flattering one. Taken first, at July's bitcoin price, the capital activity since the Q2 note adds +$0.43 a share. Taken after the bitcoin move, as above, it subtracts $2.38. This is §2's crossover in dollar terms: the repair is worth more per share the lower bitcoin is. Both orders are on July's method.

What changed in the method, and why.

• The preferred at market value, not liquidation preference (+$2.46). A perpetual preferred never matures. Its value to its holders, and so the claim it takes from the common while the company runs, is the present value of its dividends at the market's required yield, which is its price. Liquidation preference binds only in a liquidation, and the common is worth nothing in one only if the claims exceed the assets. The company itself retires preferred at market prices, as the STRC buybacks at $97.39 show. On the 29 October closes (STRF $102.67, STRC $99.99, STRK $70.22, STRD $70.52) and STRE's indicative €78.88 in the company's presentation, the stack is worth $12.37 billion against $13.40 billion of liquidation preference. Liquidation preference stays where a liquidation is the question (the strike and wipe-out) and in the like-for-like exhibits. The scenarios keep it too, because no preferred prices exist for those bitcoin prices.
• A 21% tax on the unrealized gain (−$1.16). At spot the bitcoin is worth more than its $64.02 billion cost, and a sale would be taxed at the federal rate. The note deducts $0.49 billion at the anchor, and the same rule at each scenario's price: zero in the bear case, below cost, and $7.9 billion in the bull case. No credit is taken for a loss. This replaces the earlier sensitivity that deducted the 30 September deferred tax liability at the 29 October price. The rule is a liability on a sale; the minimum tax is a separate question (§3). On a going-concern view the tax would be discounted for coins the company does not expect to sell. The house deducts it in full because selling is now policy: the company sold bitcoin in each of the last two quarters (5,552 coins in the third), and its board has standing authority to sell bitcoin to fund dividends, buybacks and the reserve. Without the tax, holding the same three converted series, the bull case is $247.52.
• Employee awards (−$0.90). Options in the money at $141.62, by the treasury-stock method, and the stock units outstanding at 30 June add 3.04 million shares. The third-quarter 10-Q will refresh the count.
• The convertibles at market value (−$0.57). With the preferred at market, leaving the convertibles at principal would value the two claims on different bases. The company's presentation prices the six series at 23 October (from $90.88 to $125.16 per $100), which gives $6.96 billion against $6.71 billion of principal: four series trade above par on the value of their conversion option, two below. A model of each series, a bond floor of $6.01 billion (2-year Treasury 4.67% plus an assumed 3-point credit spread, to each put date) plus options worth $0.93 billion (Black-Scholes at the stock's one-year realized volatility, 81%), gives $6.95 billion, a close cross-check. The prices are from a day the stock closed $4.39 higher, which overstates the option value slightly.

One adjustment was checked and not made. The 30 September dividends payable ($136 million) were paid from the USD Reserve in October (8-Ks of 5 and 19 October), so the 25 October USD assets are already net of them. Dividends accrued since then sit in the preferred prices, which trade with the accrual.

Scenarios. The June note's three pairs of bitcoin price and premium are kept, applied to today's balances and share count, as brackets rather than forecasts. $32,000 is 74% below bitcoin's closing high of $124,824 (6 October 2025), a shallower fall than the 77% and 84% falls from the 2021 and 2017 highs on the same daily closes. $120,000 is 4% below that high. The bull is not set the same way. On the same closes each cycle's high has exceeded the one before, the last by 1.85× (and the one before by 3.4×), so $120,000 is a bracket near the last high, not a cycle-high case. Above the tax cost each further dollar of bitcoin adds 79 cents of NAV. 0.85× and 1.20× bracket the premium's readings since July: 1.00× at the Q2 note, 1.13× at 30 September on the like-for-like basis and 1.14× at the 29 October close. The tax is deducted at each scenario's bitcoin price, and the outer scenarios carry the preferred at liquidation preference and the convertibles at principal, because no prices exist for those bitcoin prices. In the bull case the three convertible series in the money at that value are treated as converted, which takes it from $226.35 to $225.87. Probabilities are left out, as before, because any weighting would be a bitcoin forecast.

The bear case is higher than in July because the claims ahead of the common are smaller. The bull case is lower because there are more shares, a tax on the larger gain and three series converting. That is the trade of §2 in scenario form. The bull case keeps June's 1.20× premium although the company's chart shows median premiums of 2.28× to 2.60× in past bull years (on its own basis, which reads about 3% above the house's today). At 2.28× (2021) applied to the house NAV (the bull case's basis), $120,000 bitcoin gives about $430 a share, and today's bitcoin price gives $284. The house keeps 1.20× because a premium has value to the common only through issuance above NAV that is turned into bitcoin. In the third quarter the common sold at about 0.95× NAV (like-for-like) and 11% of it bought bitcoin, and the company names credit, not equity, as its next engine. Tripwires 2 and 3 are the observable route back to those premiums. The scenarios otherwise hold today's share count. At the bull case's price ($226) the stock would still be below its coin value (about $241 a share at $120,000), so stock sold there would still lower bitcoin per share; the frozen share count does not understate the bull case on that count. At 1.20× it would be above NAV, though, so on the value bar stock sold there would add NAV per share, and the frozen share count understates the bull case's value. A 10% fall in bitcoin takes fair value down 11.6%, to $110.16, and a 10% rise lifts it 9.9%, to $136.93. The claims ahead of the common are fixed in dollars, which gears the fall. The tax takes 21% of each further dollar of gain, which damps the rise.

Sensitivities, one at a time (fair value a share against $124.61):

On the stack checker. The house DCF checker (check_dcf.py) has five sections. Its first three, discounting, terminal boundary and exit multiples, have nothing to test in a NAV at spot, which discounts no cash flows. Its fourth, growth stages, applies by the stack's own rule only where a model carries above-normal growth: "A business with no supernormal driver gets no supernormal stage". This note carries none. Bitcoin is marked at spot, and the software business, with revenue down 3%, is a $1.4 billion comparables value. The fifth, return on capital against its cost, applies to every valuation and is shown below.

Return on capital: a measure of bitcoin, not of the business. By the house's mechanical rule the regime is weakness: −22.5% against 16.9%: the spread was +4.6 points a year ago and is −39.4 now, so the return has gone from above its cost to well below it. Strategy's operating income is almost entirely the change in the value of its bitcoin. The quarter's $20.91 billion gain sat beside $89.6 million of software gross profit. So the trailing return is bitcoin's twelve-month move run through the balance sheet, not a spread the business earns over its cost of capital. The label repeats what the NAV already shows at spot. It is recorded because the house requires it for every valuation, and it adds nothing beyond the bitcoin price.

Why this hurdle. The cost of equity uses the stock's own measured risk: Strategy's equity is a geared claim on its own bitcoin, and 60 months of its own returns give the stock's sensitivity to the market (beta 3.31, standard error 0.70). The listed alternatives each import something else. Blending toward Coinbase, MARA and Riot (20.9%) mixes in an exchange's and two miners' operating businesses (trading volumes; power and hardware costs), though the miners also hold bitcoin, far less than Strategy's 848,517 coins. The usual adjustment toward a market-average beta (16.5%) assumes a drift that nothing in the balance sheet implies. A hurdle built from bitcoin itself would mean stating a required return on bitcoin, which the house does not do. The regime label is the same at every one of these rates: September's −22.5% is far below any of them.

What $141.62 requires. At today's balances, NAV equals the price when bitcoin is at $88,986, +13.8% above the close; above the tax cost each dollar of bitcoin gain adds 79 cents of NAV. At today's bitcoin price, the price pays 1.14× NAV: $7.2 billion, or $17.01 a share, above the coin, dollars and software net of every claim.

The credit franchise, still at zero. The July note left the preferred franchise near zero beyond NAV. Its value to the common depends on issuance at a premium that is turned into bitcoin. The premium is back at 1.14×, so the price clears the NAV bar but not the coin-value bar. The franchise has still sold nothing in volume: no preferred since June, and the common sold in the quarter went mostly to the reserve and STRC, not to bitcoin. The $7.2 billion premium is the market's price for the franchise restarting. A read that does not depend on that price: each dollar of common sold at $141.62 adds 12.0% of a dollar of NAV to the existing shares. At the third quarter's pace of $21.6 billion a year, all of it buying bitcoin and with no bitcoin return assumed, issuance would take about 2.8 years to add $7.2 billion. The house carries it at zero until it restarts, for the same reason as in July. A premium is a valuation of issuance not yet done, and the only issuance this quarter cost bitcoin per share and, at about 0.95× NAV, value per share. A franchise valued at the market's figure gives the price, $141.62, by construction. The house would change this on observable conditions, set out in §6 and in "What would change our view": issuance above NAV, in volume, with the proceeds buying bitcoin, whether common or preferred.

The June, July and August notes armed twenty triggers on Strategy. One, STRC back to par by 8 September, was scored a miss in September and is not repeated. The other nineteen are graded here against the release, the 8-Ks of 28 and 29 October, the company presentation and the market record; two of them could not be read from those sources and are carried. The third-quarter 10-Q was not filed at the time of writing, so its readings are tripwire 1 in §6.

Each tripwire has a basis, a threshold and what the model does with the answer, so the next update applies rules written now. The fourth-quarter release is expected in early February 2027.

1. The third-quarter 10-Q. Basis: the preferred rollforward, the share count on the cover and the tax note (due by about 9 November). If the preferred at 25 October differs from $13.40 billion by more than $0.2 billion, the NAV is restated on the filed figure and the difference reported. A tax note that changes the minimum-tax position, or a drop in the return-of-capital expectation below ten years, is recorded beside the preferred cost in the next update. Any reconciling gap in October's STRC settlement count is closed here. The convertibles are restated at the 10-Q's disclosed fair value (Level 2) in place of the 23 October prices, and the award count is refreshed.
2. Issuance above NAV, in volume, into bitcoin. Basis: the weekly 8-Ks' common sales and use of proceeds, against NAV per share on this note's definition ($124.61 today; it moves with bitcoin). $1 billion or more of common sold in a quarter above NAV per share, with the proceeds buying bitcoin, is the restart condition. The house then reopens the franchise value at the issuance actually done. Each update also reports the coin-value bar ($157.10 today): issuance above NAV but below coin value adds value per share while lowering bitcoin per share, and the update says which applied.
3. Preferred issuance resumes. Basis: weekly 8-K ATM tables. $500 million or more of preferred sold in a quarter at or above its stated amount, at a coupon of 12% or less, with the proceeds buying bitcoin, prices the franchise from the spread between the coupon and the company's own 10.6% hurdle on the issuance done, and adds the new coupon to the fixed charge. Preferred sold below the stated amount, or sold to fund dividends, counts as cost, not franchise.
4. The reserve. Basis: USD assets in weekly 8-Ks against the fixed charge. Coverage below 2.0 years (the low end of the company's stated two-to-three-year target), or a reserve draw that is refilled by bitcoin sales rather than common, moves the base case's funding path from "reserve" to "coin" (§2: bitcoin per share down 2.3% a year at today's prices). Coverage above 3.0 years with the ATM idle confirms the pre-funded path.
5. The premium through 1.0×. Basis: mNAV on this note's definition. Ten consecutive closes below 1.00× make the loop dilutive at NAV, and the update re-runs the funding paths at the sub-NAV price. The June and July mNAV triggers stay armed alongside.
6. Full-year BTC Yield. Basis: the fourth-quarter release (early February 2027). A full-year figure below −3.7% means the fourth quarter repeated the third quarter's pattern; the base case then carries the Q3-mix path. A positive fourth-quarter contribution, funded by preferred or by common sold above coin value, moves the base case to the reserve-funded path, with bitcoin per share flat, and is recorded against the company's 10.4%-a-year objective.
7. Finance leadership. Basis: the 10-Q and 10-K signatures, the auditor's report and any 8-K Item 4.01 or 4.02. The CFO change is recorded as not arising from a disagreement (8-K Item 5.02). A restatement, a material weakness or an auditor change within twelve months of it is recorded as a governance event in the next update.
8. The price. Fair value $124.61. A close at or below it resolves the gap. The bear case, $27, is the level at which the stock would price bitcoin near $32,000 at a discount to NAV.

Fair value $124.61 a share at net asset value (bear $27 / base $125 / bull $226), from $94 on 31 July: $124.79 on July's method, then −$0.18 from four method changes, each stated with its own effect. At the $141.62 close of 29 October the price is 13.6% above fair value, a premium of 1.14×. After the release the stock traded at $132.76 at 9:08 p.m. ET (−6.3%, 1.07× NAV), one data point and not an input. The bitcoin price accounts for the increase from $94. The quarter's capital activity was a trade: 11.8% of each share's bitcoin for 34% of the net claims ahead of it, a smaller fixed charge ($1.51 billion) and 3.4 years of dollar cover. That trade strengthens the credit. For the common, it was worth more coin below $63,600 bitcoin and less above. The flywheel did not restart: the common was sold below coin value and mostly not spent on bitcoin. The price therefore pays $7.2 billion for a franchise that has not issued since June, or it requires bitcoin at $88,986. The house values what has been done and lists in §6 the issuance that would change that.