Reversing the stack to find the original intent.
Capital Research Global Investors dumped $462 million of MSTR in Q2. That single line in the 13F filing is the canary. But the market cheerleaders focused on the headline: 12 out of 15 top institutions increased their positions. Net inflow: $700 million. A triumph, they say.
Let’s trace the actual execution path.
The headline is true. The narrative is misleading. A forensic read of the 13F data reveals a structural divergence: passive index funds are buying because the index says so. Active managers are selling because the model has changed. The flywheel that once turned BTC purchases into equity premium is now grinding in reverse.
Context: The Capital Structure as Smart Contract
MSTR is not a blockchain protocol. It is a financial engineering construct. The original design: issue equity or debt, buy Bitcoin, hold forever. The "never sell" promise was the sole invariant. Investors paid a premium to NAV for pure, leveraged BTC exposure with no forced liquidation.
Then came STRC, the preferred stock. Fixed dividends. No variable expense. When BTC price stagnates, those dividends must be paid from somewhere. MSTR started selling Bitcoin in Q2 to fund the payouts. The invariant is broken. The smart contract—if we analogize the capital structure—now has a mandatory sellBTC() function that triggers every quarter.
Based on my audit experience dissecting 0x’s fillOrder logic, I know that once a forced exit path is introduced, the system’s equilibrium shifts. The original intent was accumulation. The new intent is consumption. That is not a bug report—it is a fundamental redesign.
Core: Decomposing the Marginal Buyer
Let’s examine the Q2 13F data line by line.
- Vanguard added $147 million. BlackRock added $84 million. These are passive index funds. They rebalance quarterly, tracking the Russell 1000 or S&P 500. They do not make active bets on MSTR’s Bitcoin strategy. They buy because MSTR’s market cap weight increased.
- Capital Research Global Investors cut $462 million. UBS cut $142 million. Geode cut $5 million. These are active managers. They can choose to exit. They did.
Net change: +$700 million. But the composition is critical. Passive inflows are sticky but not signal. Active outflows are signal. The aggregate hides the divergence.
Now map this to the "tokenomics" of MSTR stock. The supply model is elastic: MSTR can issue new shares to buy more BTC. The flywheel requires that the market price stays above NAV, so that new equity raises net positive BTC per share. As long as passive funds keep buying, the premium can persist. But the active selling reduces the premium. If the premium collapses to zero or negative, MSTR cannot raise new capital without diluting existing holders. The flywheel stops.
Truth is not consensus; truth is verifiable code. The code here is the 13F filings. The verifiable fact: the marginal active buyer is gone. The marginal holder is now a passive robot that does not care about the "never sell" promise. That is a fragile foundation.
Contrarian: The "12 out of 15" Narrative is a Trap
The official Strategy tweet highlighted "12 of 15 top institutional holders increased their positions." This is selective disclosure. It omits that the three sellers accounted for a disproportionate share of the net change. Capital Research alone sold more than Vanguard and BlackRock combined bought.
Furthermore, the 12 holders that increased include firms like State Street and Geode, which are also predominantly passive index investors. The only active manager among the top 15 that increased was Capital International, which added $1.1 billion. But that is a separate entity from Capital Research—same parent, different mandate. The active managers as a group are net sellers.
Abstraction layers hide complexity, but not error. The headline abstraction is "institutional confidence." The underlying error is the conflation of passive allocation with active conviction. The signal is not the net number; it is the divergence.
Consider the incentive structure. MSTR now has a fixed cash outflow: STRC dividends. The only way to cover it without selling BTC is to generate operating income—which MSTR does not have. So the selling is structural. Every quarter, the model consumes a piece of its Bitcoin reserve. This is exactly the opposite of the "accumulation" that the original thesis relied upon.
If BTC price drops further, the selling will accelerate. The dividend is fixed in dollar terms, but the BTC needed to cover it increases as price falls. That is a negative convexity. The model is short volatility.
Takeaway: The Next 13F Will Be the Verdict
Q2’s 13F shows a system in transition. The passive funds are still buying, but the active money is already pricing in the capital consumption. The next 13F, due in November 2026, will reveal whether the divergence deepens.
If more active managers follow Capital Research out the door, MSTR’s premium will shrink. Equity financing becomes expensive. The flywheel reverses. The only way to sustain the model is to find a new source of non-dilutive capital—or to hope that BTC price rallies enough to make the selling painless.
Reversing the stack to find the original intent. The original intent was a Bitcoin treasury that never sells. The current reality is a Bitcoin treasury that sells every quarter. The code has changed. The market is still catching up.
Truth is not consensus; truth is verifiable code. Check the next 13F. Check the BTC balance. The signal is in the diff, not the headline.
