The Filing Says Confidence. The Arithmetic Says the Ceiling Was Too Low.
Strategy's board approved doubling the aggregate authorization under its Digital Credit Securities Repurchase Program from $1 billion to $2 billion. The 8-K filed Tuesday. As of Sept. 7, $1.19 billion remained available.
That is the headline. Here is the ledger behind it.
Between Aug. 31 and Sept. 7, Strategy repurchased 1,810,885 shares of STRC, its variable rate perpetual preferred, for $176.3 million. Cash out, shares in. Divide one by the other and you land at $97.36 per share. STRC is engineered to hold $100 par. It did not.
Now perform the subtraction the press release declines to perform. If $1.19 billion of a $2 billion program remains available, cumulative utilization is $810 million. Of that, $176.3 million was deployed inside a single week. The residual $633.7 million was deployed between the program's June 29 origin and Aug. 30.
Strategy absorbed 81% of its original preferred authorization in roughly ten weeks. Then it doubled the ceiling. And the instrument still prints below par.
The board resolution is not a statement of conviction. It is a statement that the previous ceiling was insufficient. Those are two different documents that happen to share a cover page.
I have read enough capital-structure disclosures to separate an offensive doubling from a defensive one. Offensive doublings arrive with a shrinking discount. Defensive doublings arrive with a flat one. This is the second.
The rest of the 8-K is more instructive than the number at the top of it. Fund from cash, not equity. Buy the preferred, not the common. Hold the bitcoin. Say nothing about the discount.
Read the whole thing and a hierarchy of capital allocation falls out of it — an explicit ranking of which liabilities Strategy is willing to retire with scarce dollars, and at what price. That ranking is the actual disclosure. Everything else is formatting.
What June 29 Actually Authorized
To read September's filing you need the June architecture. On June 29, Strategy announced a financing overhaul with three separate authorizations buried inside one narrative.
The first authorized up to $1.25 billion in bitcoin sales. Not purchases. Sales. That authorization sits in a footnote and almost nobody quotes it.
The second created a $1 billion repurchase authorization for the preferred stack — the digital credit securities.
The third created a separate $1 billion repurchase authorization for common stock, meaning MSTR shares.
As of Sept. 7, the preferred side has drawn to $810 million and been re-ceilinged to $2 billion. The common side sits at zero. The full $1 billion remains untouched on the MSTR side, and the filing states it plainly.
That asymmetry is the first real signal in the document. A treasury vehicle that repurchases one wrapper and refuses to touch the other is telling you something about the relative price of its own liabilities. It is not making a values statement. It is pricing two instruments against one asset base.
Background that matters for the rest of this analysis: Strategy holds 845,050 BTC acquired for $63.73 billion, an average of $75,412 per coin. The USD Reserve stands at $5.10 billion and exists specifically to back preferred dividends and debt interest. General purpose USD cash adds another $1.44 billion. Total disclosed USD liquidity: $6.54 billion.
I rebuilt the treasury figures from the two disclosed fields before writing a word of this. 845,050 coins times $75,412 equals $63.7269 billion. The filing reports $63.73 billion. The two bind to within four dollars a coin, roughly 0.005% of the average cost. When a filing reconciles to itself that tightly, the numbers are internally generated from one spreadsheet rather than assembled from several. That tells me the treasury table is reliable. It tells me nothing about whether the treasury is comfortable.
The Arithmetic of Paying 97 for Par
Start with the trade itself, because the trade is clean.
STRC carries a $100 par and a variable rate designed to defend it. Strategy retired 1,810,885 shares of that par structure. Nominal liquidation preference retired: $181,088,500. Cash paid: $176,300,000. Discount captured: $4,788,500. Per share, $2.64. As a percentage of the par retired, 2.64%.
That is not a rounding error. It is a real transfer of value from the sellers to the remaining holders, and Strategy is one of those remaining holders. Retiring a $100 claim for $97.36 is the same species of trade as buying an asset below liquidation value. It is instantly accretive to everything left standing behind it.
Now extend it forward. Retiring $181.1 million of par also retires the future dividend obligation attached to that par. The filing does not disclose the current reset rate, so I ran a bracket. If the floating rate sits anywhere between 9% and 11%, the annual cash dividend eliminated on this single week's repurchase runs between $16.3 million and $19.9 million. Over five years, undiscounted, that is $81 million to $100 million of avoided distributions against a $4.8 million entry discount.
The discount is the headline. The dividend elimination is the business case.
The two effects compound in the same direction: you capture the spread once, and you stop paying the coupon forever. For a perpetual instrument with no maturity, the coupon is the entire cost of the liability. Removing it in perpetuity at a 2.6% discount to par is a spread trade, not a share buyback in any conventional sense.
This is where I start thinking about 2017. During the ICO boom I audited three Southeast Asian utility token launches by hand, tracing distribution logic through the Ethereum state tree. Two of the three promised decentralization while retaining admin keys they could exercise at will. The lesson was not that the tokens were bad. The lesson was that the contract said one thing and the team said another, and the contract always won.
An 8-K is not a whitepaper, but the discipline transfers. Read the numbers, not the framing. The framing says support. The numbers say accretion at 2.64% off par, on a schedule that ran out of authorization in ten weeks.
The Funding Decision Is the Real Signal
The buybacks were funded entirely from USD cash. No at-the-market sales. No bitcoin purchased. No bitcoin sold. Holdings flat at 845,050 coins.
For a company whose public thesis has been an equity flywheel — issue shares at a premium, convert the proceeds into bitcoin, watch the premium persist — the absence of ATM issuance is louder than any buyback.
The flywheel paused for a week. That matters because the flywheel is the mechanism that has historically funded the preferred stack. Issue common above net asset value, capture the spread, use the spread to service the dividend obligations that sit below it.
Strategy chose cash instead. Specifically, it chose to draw on USD liquidity that exists to cover preferred dividends and debt interest, and to convert some of that coverage into discount capture.
Run the numbers on capacity. Remaining buyback authorization is $1.19 billion. Total disclosed USD liquidity is $6.54 billion. The remaining authorization equals 18.2% of total liquidity, or 23.3% of the USD Reserve alone. That is not a trivial allocation. If the program is fully drawn, roughly a quarter of the dividend-coverage reserve is spent on retiring claims rather than servicing them.
There is a defensible logic here. Retiring a claim removes the need to service it, so spending reserve on buybacks reduces the reserve's future obligations in the same motion. Coverage ratios can improve even as the reserve shrinks. Whether they do depends entirely on the price paid relative to the coupon removed.
Which brings the whole transaction back to one variable. Everything works at $97.36. Less of it works at $99.50. None of it works at $101. The program is a bet on a discount, and the discount is the only assumption that has to hold.
The Discount That Refuses to Close
Here is the anomaly worth sitting with.
Strategy has now absorbed $810 million of its own STRC. That is a substantial buyer removing supply from a comparatively small instrument. In most markets, $810 million of persistent bid against a below-par instrument closes the gap. Flow imbalance is the standard explanation for a discount, and flow was delivered in size.
The gap survived. The instrument still cleared at roughly $97.36 in the week ending Sept. 7.
That is the finding. When a discount persists after the issuer has absorbed 81% of its own authorized capacity, flow is no longer a sufficient explanation. You have to move to the next hypothesis: the discount is not about supply. It is about the price the market assigns to the claim itself.
In 2020 I built Python scrapers against Uniswap and Curve, clustered more than 500 wallets, and found that roughly 60% of the apparent organic volume in early yearn.finance forks was insiders washing the tape. The volume was real and reported. The volume was also meaningless, because the wallets on both sides answered to the same owner. Raw tape without counterparty attribution is narrative, not evidence.
The same discipline applies here. Aggregate buyback dollars tell you what Strategy spent. They do not tell you who was on the other side, at what size, or whether the sellers were re-buying the instrument later through a different venue. The 8-K discloses one side of the trade and is silent on the other. Reading only the disclosed side is how you end up mistaking absorption for conviction.
A variable rate perpetual is designed to self-correct. Price falls below par, the reset mechanism lifts the distribution, the higher yield attracts buyers, the price walks back to par. That is the contract's own feedback loop. The loop did not close.
Three explanations survive contact with the data, and the filing distinguishes between none of them. The rate mechanism may be capped, so the self-correction has a hard ceiling the market has now reached. The market may be pricing issuer-level risk above any rate the instrument can legally pay. Or the float may be captive to holders who cannot mark the position and therefore cannot arbitrage the gap.
All three produce the same tape. They do not produce the same future. That is the analytical problem the 8-K hands you and declines to solve.
Why STRC and Not the Other Three
The filing states that no STRF, STRK, STRD, or MSTR shares were repurchased during the week. Four instruments sat untouched while one absorbed $176.3 million.
Read that as data rather than preference. A rules-based repurchase program buys whatever is cheapest relative to its contract value. If STRC is the only series trading below par, then STRC is the only series with a discount to capture, and the concentration is mechanical rather than editorial. Strategy is not expressing affection for one instrument. It is harvesting one dislocation.
If instead the other series are trading at or above par, the program is behaving exactly as a rational capital allocator would, and there is no story. The story only appears when you ask what happens when STRC is the only series below par. Then the concentration is not just mechanical — it is diagnostic. It says the market has singled out one claim in the stack.
I cannot settle this from the filing, and neither can anyone else. The disclosure covers what was bought, not what the alternatives were quoted at. What I can say is that the absence of activity in three of four series is a testable fact, and the absence of any price commentary about them is a gap in the disclosure that a serious allocator would want closed.
The common authorization is the more interesting museum piece. One billion dollars, untouched. Strategy will buy its preferred below par and will not buy its common at any price it currently faces. That is a two-sided spread trade on its own wrapper: extract value where the market pays below contract, refuse value where the market pays above net asset value.
It is the correct trade. It is also a confession about where the company believes its own securities are mispriced.
The Coverage Number the 8-K Withholds
The filing gives you the reserve size and the authorization size. It does not give you the ratio that matters.
$5.10 billion in the USD Reserve, explicitly designated to back preferred dividends and debt interest. What it does not disclose is the annualized cash obligation of the preferred stack. Without the numerator of that fraction, nobody outside the company can compute the coverage period.
Run the bracket. If the total preferred distribution burden is $500 million a year, the reserve covers roughly ten years of it. If it approaches $1 billion, coverage compresses to about five. If the stack has grown faster than the reserve, the number is lower still, and every dollar redeployed from the reserve into buybacks shortens it further.
A repurchase program funded from a coverage reserve is a duration decision, not a return decision. You are trading future dividend security for present discount capture. That can be the right trade. It is only the right trade if you know the coverage period you are shortening, and the filing does not tell you.
This is the specific disclosure gap I would press on if I had the call. Not the authorization size. Not the weekly deployment. The annualized preferred distribution obligation, and its ratio to the USD Reserve. Everything else in the document is downstream of that one number, and that one number is absent.
845,050 Bitcoin and the Sale Authorization Nobody Reads
Zero bitcoin purchased. Zero bitcoin sold. Holdings flat at 845,050 coins at an average cost of $75,412.
In a bull market, a treasury company that does not add to its treasury in a week is a treasury company that found a better use for the marginal dollar. This week that use was retiring preferred at a 2.6% discount to par. That is a defensible ranking, and it reveals the internal hierarchy clearly: discount capture on the preferred stack outranks spot bitcoin accumulation, and both outrank common buybacks at current prices.
That hierarchy only makes sense if the discount is temporary. If STRC's gap to par is transient, buying it at $97.36 and holding bitcoin simultaneously is the correct sequencing. You harvest a closing spread and keep the optionality.
If the discount is structural, the ranking inverts. You are spending coverage capital on an instrument the market has already repriced, and the bitcoin you did not buy in a bull market is bitcoin you will buy at a worse price or not at all.
Now look at the authorization nobody quotes. Up to $1.25 billion in bitcoin sales, approved June 29, standing, untouched. A bitcoin treasury company with a pre-authorized sale facility is holding a tail hedge it does not advertise. That is prudent. It is also an admission that the board models a scenario in which the treasury itself becomes the liquidity source.
The bear market doesn't arrive as a headline. It arrives pre-authorized, in a footnote, months in advance, waiting for a trigger that the people reading the headline never see coming.
Exchange flow taught me this in 2022. I tracked roughly 10,000 BTC moving out of exchange cold wallets into known deposit addresses weeks before the Celsius and Voyager failures became public, structured my own book into a 70/30 stablecoin ratio, and published the risk assessment while the tape still looked calm. The movement preceded the news. It always does. What precedes the movement is an authorization.
Correlation Is Not Causation — Until the Sample Repeats
The obvious counterargument deserves a fair hearing, because it is strong.
Companies buy back their own securities constantly. It is the most ordinary capital allocation act in existence. A company retiring a claim trading below its contract value is not distressed — it is efficient. Nothing in this filing proves stress. A discount to par is not a default signal. A doubling of an authorization is not a warning. Read as a set of independent facts, every element here is unremarkable.
That is the honest position, and I hold part of it. The buyback is not evidence of anything except that the discount existed. The existence of a discount is not itself an anomaly — instrument-level dislocations happen in every market, for reasons ranging from index mechanics to tax lot timing to a single holder needing liquidity.
The anomaly is the persistence. One week at $97.36 is noise. Ten weeks and $810 million of absorption without convergence is a sample. And the sample is drawn from the cleanest possible experiment: a single issuer, a single instrument, a disclosed and dated buying program, and a contract that mechanically defines fair value at $100.
The blind spots are real and I will not paper over them. Nobody outside the company can see the STRC order book depth, the float composition, or the reset cap that governs how far the rate can climb. Nobody knows whether the sellers are natural holders rotating out or arbitrageurs financing a basis trade against the next reset. Nobody knows whether the $633.7 million deployed between late June and late August executed at better or worse prices than this week's $97.36.
The correlation I can defend is narrow. Persistent discount plus heavy issuer absorption plus a doubling of authorization is not proof of a problem. It is proof that the original authorization was calibrated to a discount that did not close as fast as the board expected. That is a forecast error inside the company's own treasury model, and forecast errors are the most reliable leading indicators I know.
Where I look next is the ratio nobody publishes. Preferred buyback spend against reserve drawdown, disclosed weekly. If the reserve holds near $5.1 billion while the buyback consumes $1.19 billion, the coverage math is stable and this is efficient housekeeping. If the reserve falls in step with the program, the coverage period is compressing in real time, and the doubling was not a decision about discounts at all.
What to Watch Next Week
The forward signals are specific and dated, and none of them involve the bitcoin price.
STRC's print against $100 par. This is the only number in the entire complex that carries information this week. If the discount narrows after the authorization doubled, flow was the binding constraint and the trade is working. If it holds at $97 and change, flow was never the constraint, and the repricing is about the claim.
A second consecutive week of buybacks at a similar pace. One week is a decision. Two weeks is a program. The pace tells you whether the board is harvesting a window or defending a floor.
The next reserve disclosure. $5.10 billion in the USD Reserve plus $1.44 billion of general cash is the base. What that base does while $1.19 billion of authorization is live is the coverage question, answered with data instead of assumption.
And the $1.25 billion. Bitcoin purchase activity, ATM issuance activity, or any draw on the sale authorization. A treasury company that resumes buying coins has decided the discount is temporary. One that reaches for the sale facility has decided it is not. Both are possible. Only one is priced.
Everything else in the week's filing is accounting. Strategy bought 1,810,885 of its own shares for $176.3 million, and the only number in the document that describes the future rather than the past is the $2.64 per share it left on the table.
The question is not whether $97.36 is a good price. It is what the instrument is worth if the buyer who has been setting that price stops.
Liquidity didn't dry up in STRC. It was never deep enough to matter, and the tape has now said so for ten weeks in a row.
