MicroStrategy’s Bitcoin Treasury Sells to Pay Dividends

In six years an American software company turned itself into the largest listed holder of bitcoin on the planet, with 845,050 coins on its balance sheet. That Bitcoin Treasury is no longer run as a dormant reserve but as an asset financed by liabilities that come due in dollars. This file walks through the full mechanism, from the founding purchase in 2020 to the 2026 sales that fund preferred dividends. It also explains why management had to rewrite its own reporting metrics to stay legible.

Key Takeaways

• 845,050 BTC held for $63.73B, an average cost of $75,412 per coin
• Preferred dividends are paid in cash, which means selling bitcoin to fund them
• New metrics were introduced to account for claims that rank ahead of shareholders

Contents

From $250M in 2020 to 845,050 Coins
A Funding Machine With Three Floors
When Dividends Start Dictating Sales
Why the Reporting Metrics Had to Change
What the Exposure Loads Onto the Stock

From $250M in 2020 to 845,050 Coins

Everything starts with one decision taken in August 2020. Facing near-zero rates and an idle cash pile losing value, management put $250M into 21,454 bitcoin. At the time the move was framed as balance sheet protection rather than a change of business.

That first lot works out to roughly $11,650 a coin, which says a lot about the scale of what followed. No listed company held a comparable position then, and accounting standards had nothing ready for the line item. Every framework that exists today was built after the fact.

The size of that first purchase also explains the shape of everything that followed. A company that commits its entire excess cash to one asset has no second reserve to fall back on, so any further acquisition has to be funded externally. The capital structure that dominates the story today was, in effect, decided by that opening move.

What came next was five years of continuous accumulation. Price rallies funded capital raises, drawdowns funded purchases, and the cadence never tracked the level of bitcoin itself. In 2024 alone, three billion dollars came in through convertible notes to fund further acquisitions.

The 2025 rebrand made the shift official. A software publisher became a company whose main activity reads on the balance sheet rather than the income statement. Software barely covers the running costs now, and the equity moves with an asset the company does not control.

The position as of September 7, 2026 shows the distance travelled. The company holds 845,050 bitcoin bought for an aggregate $63.73B, around $75,412 a coin including fees, alongside a $5.10B dollar reserve and $1.44B in available cash.

Holdings are still up roughly 30% since January despite the disposals, and total assets approach $72B. At the September 7 price near $77,760, the unrealised gain on the whole position comes to about $2B, or 3% after six years of accumulation. That thin margin is the number most commentary leaves out.

A Funding Machine With Three Floors

The Bitcoin Treasury model rests on one principle: raise capital below the expected return of the asset it buys. Three channels feed that machine, and each one brings its own constraint to the balance sheet.

The first floor is at-the-market issuance of common stock. As long as the share trades above the market value of the bitcoin it represents, every issuance raises bitcoin per share despite the dilution. It is the most elegant piece of the structure and the one most dependent on sentiment.

The dependency runs deeper than it looks. If the share trades below the value of the bitcoin behind it, the same issuance destroys bitcoin per share instead of creating it. The tool does not simply become less effective in that scenario, it turns actively harmful, which is why management stops issuing rather than push through a discount.

The second floor is convertible debt. It costs little in interest, converts into equity if the price climbs, and drains cash if the price stalls. The company has shown it can handle those maturities without touching the reserve, retiring $1.5 billion without selling a single coin.

The third floor is the newest and the tightest. Perpetual preferred shares, the heaviest series trading under the STRC ticker, carry a recurring dividend that has to be paid in cash. They allowed large raises without diluting common holders, at the cost of a permanent payment obligation.

Regulatory filings let anyone trace that circuit directly. The late-August filing with the US regulator confirms that proceeds from equity sales funded the bitcoin purchases made in the same week. Selling stock at a premium to buy the underlying asset stays the core mechanism, and it only works while the premium exists.

That third floor is what changed the nature of the whole structure. A convertible is repaid once, on a known date. A perpetual dividend comes back every quarter, whatever bitcoin does and whatever management would prefer.

Also on Cryptonomic:

• Hunter Biden Launches LAPTOP Memecoin, Founders Take 30%
• Liquid Network Recovers 3,400 of 4,000 Stolen Bitcoin
• Bitcoin Drops Under $79K Two Days Before US Inflation

When Dividends Start Dictating Sales

2026 exposed the logical consequence of that design. In July, $216M of bitcoin was sold to meet preferred dividends. In early August, another 1,638 coins went out for the same reason.

The summer sequence made the mechanism impossible to miss. Between June and August, 6,948 bitcoin were sold at a weighted average near $62,200 to fund those dividends. In late August, 4,603 coins came back at $80,318 each, an $18,000 gap between the exit and the return.

Management defends the trade on different ground entirely. Chief executive Phong Le frames decisions around the cost of capital raised rather than the price of bitcoin. Across the ten weeks separating the two operations, net debt went from roughly $7B to zero, which made equity issuance cheaper on the other side.

The argument holds over the life of the balance sheet. What stays true is that the cash constraint is now structural, and the bitcoin reserve is the adjustable variable when cash is needed. Michael Saylor dropped the never-sell doctrine publicly back in May, owning the idea of selling to steady the market.

The week of September 1 to 7 showed the other face of this management. Zero bitcoin bought, zero sold, but 1,810,885 preferred shares repurchased for $176.3M out of the dollar cash pool. The board simultaneously doubled the digital credit repurchase authorisation from one to two billion dollars, with $1.19B still available.

That quiet week followed an immediate return to buying. The previous week saw $370M go into bitcoin after two full months of inactivity, so the comeback lasted exactly seven days. Alternating between resumption and pause has settled in as the operating regime, where 2021 through 2024 offered near-continuous accumulation.

Buying back its own preferred stock reduces the future dividend load. That is liability management rather than a bitcoin decision, and it signals that the current priority sits on shrinking recurring commitments instead of growing the pile.

The 6,948 coins sold over the summer amount to less than 1% of total holdings, a proportion management describes as marginal against a stack of 845,050. The objection is not about scale, it is about direction. A treasury that was presented for five years as a one-way accumulation now moves in both directions on a schedule set by cash obligations rather than conviction.

Why the Reporting Metrics Had to Change

As the liability stack behind the Bitcoin Treasury grew more complex, the old reference points stopped working. Comparing market capitalisation to the value of the coins ignores debt and preferred stock, both of which rank ahead of common holders in any liquidation. In July 2026 management introduced two metrics built to correct that bias.

The first, enterprise mNAV, adds market capitalisation, total debt and perpetual preferred stock, then subtracts dollar reserves. It measures the full economic cost of the capital structure instead of the equity valuation alone.

The second, net reserve, starts from total bitcoin plus cash, then subtracts preferred stock and out-of-the-money convertibles. It isolates the share of the reserve that genuinely belongs to common shareholders once senior claims are served. Against a $55.6B bitcoin reserve, the gap between the two readings runs into billions.

Neither metric is standardised, which is the honest caveat to attach to both. They were designed in-house, they are published by the company they measure, and no auditor certifies the framing. They remain more informative than a raw coin count, but a reader should treat them as management’s own map of its balance sheet rather than as an independent measurement.

Both answer a question shareholders had been asking for months. When a company funds a volatile asset with instruments that outrank its own equity, the headline value of the stack stops matching what a shareholder owns. Part of the reserve already backs other creditors, and that share grows with every preferred raise.

The primary record sits with the regulator rather than with any commentary. The 8-K filed with the SEC in late August confirms how equity proceeds were deployed, and the first quarter 2026 results release lays out the cash position and the liability structure behind it.

What the Exposure Loads Onto the Stock

An investor buying the share is not buying bitcoin. They are buying exposure to bitcoin plus a balance sheet, an issuance policy and a dividend calendar they have no control over. The premium on the vehicle pays for that extra risk, and it compresses the moment the management story gets harder to read.

Analysts have put a number on the compression. Bernstein keeps an Outperform rating while cutting its price target from $450 to $350, and still models bitcoin at $150,000 by mid-2027. The cut lands on the vehicle premium, not on the thesis about the underlying asset.

One session gave that gap a concrete measure. The stock closed September 1 at $124.88, down 6.1% on the day, while bitcoin gave back only 1.1%. Five points of divergence between an asset and its main listed proxy means the market was trading the financial structure that day, not the crypto.

Index pressure sits on top of it. MSCI has opened a process to remove the stock from its indices on the basis that the real activity has drifted from its original classification. An index exit forces passive funds to sell regardless of any view on the company.

The mechanical nature of that flow is what makes it dangerous. Passive vehicles do not weigh a thesis, they track a list, and a removal triggers selling into whatever liquidity happens to exist that day. For a stock whose premium already carries the funding model, an index-driven drawdown feeds straight back into the cost of raising the next dollar.

Purchase pauses are the other signal worth tracking. The company had already suspended acquisitions ahead of a quarterly earnings release, and the week ending September 7 closed with nothing bought and nothing sold. Those still windows say more about funding conditions than any acquisition press release.

Management, for its part, rules out further disposals in the near term. Phong Le describes himself as a buyer at $80,000, then at $90,000 and $100,000, and up to $130,000 on a new all-time high. Those intentions carry the weight of intentions inside a model where the dividend obligation stays contractual and quarterly.

The deeper question is whether the model travels. Dozens of listed companies have copied it since 2024 without the same access to debt markets or the same shareholder base. A structure that works because of a valuation premium stops working when that premium goes, and the company that walked this path first is also the furthest into discovering where it ends.

For anyone tracking this Bitcoin Treasury, three numbers carry the file. Coin count gives the scale of the exposure, net reserve gives what actually belongs to common shareholders, and the preferred dividend bill gives how many dollars have to leave the building every quarter no matter what. Read together, those three lines describe a position that is neither a simple bitcoin proxy nor a conventional balance sheet, and that distinction is the whole point of tracking it closely.

Follow the story on Cryptonomic.