There is a number moving through crypto media right now that refuses to balance. On one side of the ledger, River's dataset claims individual holders net-sold 696,000 BTC across 2025. On the other side, the same body of data shows institutions, funds, and governments net-buying close to a million coins. The difference โ roughly 300,000 BTC โ has no named seller and no named buyer. It simply exists, an orphaned entry in a ledger that is supposed to be the most transparent in human history.
I found it the way I find most things: not in the headline, but in the margin. I was building a reconciliation model for a client's treasury exposure when the columns refused to close. Three hundred thousand coins, priced at the article's own cited figure of $83,000, is $24.9 billion of unaccounted flow. That is not a rounding error. That is a signal.
Logic blooms where silence meets code.
Context: The Story Everyone Is Telling
The narrative currently circulating โ and it is a narrative, dressed in the grammar of statistics โ is a clean binary. Wall Street versus ordinary people. Institutions versus the crowd. The framing goes like this: individual investors still hold roughly 66% of all bitcoin, some 13.9 million coins, while ETFs hold 1.48 million, corporations hold 1.66 million, and governments hold 518,500. Read quickly, it says the little guy still owns the game, and the suits are merely renting a seat.
I have spent a career reading balance sheets that were designed to be read quickly. My instinct, trained across six weeks of line-by-line review of the Ethlance crowdsale contract in 2017, is to distrust any number that arrives pre-aggregated, pre-summarized, and pre-digested. The Ethlance audit taught me something that has never left me: the vulnerability is rarely in the code you are shown. It is in the assumptions layered on top of it. That contract looked elegant until you traced the integer overflow through the token distribution logic. The beauty was real. So was the flaw.
The bug hides in the beauty.
So let me apply the same discipline to the ownership map. Not to the bitcoin protocol โ that is the most audited code in existence, running for sixteen years, secured by proof-of-work and the longest-chain rule with a trust assumption that approaches minimal. The protocol is not the question here. The question is the layer above it: the custody structure, the aggregation methodology, and the corporate treasury engineering that the headline number quietly smuggles into the word 'individual.'
Core: An Audit of the Numbers That Everyone Quotes
Let me begin where any competent audit begins โ with provenance. Where does the 66% figure come from? It comes from third-party aggregators. Bitbo. River. Platforms that are not audited themselves, that publish no reproducible methodology, and whose statistical definitions are opaque on the two questions that matter most: does 'individual holding' include exchange wallets, and does it exclude provably lost coins? The source notes for the underlying data are listed as 'none' or 'aggregator' โ which, in audit language, means the figure is unverified and, more importantly, unverifiable by anyone reading the report.
This is the first thing I would flag in any engagement. A number without a methodology is not data. It is testimony.
Now trace the definitions through. Start with the 1.6 million BTC estimated lost forever โ coins whose private keys are gone, whose owners are dead or forgetful, whose hardware was thrown into landfills. These coins are not held by anyone. They cannot be sold. Yet they almost certainly sit inside the 13.9 million 'individual' total, because aggregators generally cannot distinguish a lost coin from a dormant one. Add the 968,000 coins attributed to wallets associated with Satoshi Nakamoto โ never moved, possibly never movable. Add, most consequentially, the exchange custody wallets.
Here is the structural problem that the headline hides. When you deposit bitcoin on a centralized exchange, you do not hold a coin. You hold an IOU against a commingled pool. On-chain, that pool is a handful of very large addresses. A single exchange hot wallet holding customer assets for ten million users appears on the ledger as one address โ or a few โ controlled by a corporate entity. If the aggregator's heuristic counts that address as 'individual,' then ten million retail depositors are being represented by one corporate custodian. If it counts it as 'institutional,' then retail is being undercounted. Either way, the label is a fiction imposed on a chain that only knows addresses, not identities.
I trace the shadow before it casts.
The same distortion runs the other direction. The 1.48 million coins held by spot ETFs are, on-chain, a small number of enormous addresses controlled by custodians like Coinbase Custody. The article's own logic concedes that ETFs hold 1.48 million coins as 'passive money, redeemable in both directions.' But redeemable by whom? By the ETF issuer, on behalf of the shareholder. The shareholder does not own an address. The shareholder owns a share of a trust that owns an address. The chain sees the trust. The market sees the shareholder. The gap between the two is where every ownership statistic quietly dies.
So when the report claims individuals hold 66%, what it likely means is: addresses that are not tagged as ETF, corporate, or government. That is a residual. It is the number you get after you subtract what you can identify. It is not a measurement of retail self-custody. It is the absence of other labels.
Let me put a range on it. Strip out the 1.6 million lost, the 968,000 Satoshi coins, and the portion of exchange wallets that are genuinely custodied on behalf of identifiable institutional or ETF clients, and the true free float โ coins actually available to trade, held by entities who could actually move them โ lands somewhere between 10 and 12 million. That is 48% to 57% of maximum supply, not 66%. The narrative overstates retail ownership by somewhere between nine and eighteen percentage points. In a market where every basis point of float matters to price discovery, that is a material misstatement.
Now the second layer of the audit: the flows. The report states that in 2025 individuals net-sold 696,000 coins while institutions, funds, and governments net-bought close to a million. Even accepting the labels, the arithmetic does not close โ and the gap of roughly 300,000 coins cannot be explained by new issuance alone. Bitcoin's post-halving issuance runs at approximately 3.125 BTC per block, which annualizes to roughly 164,000 coins per year, about 0.78% of circulating supply. That accounts for less than two-thirds of the discrepancy. The remainder could be exchange internal transfers, reclassification between periods, or โ most likely โ the overlapping definitions I described above, where a coin is counted as sold by one cohort and bought by another within the same aggregation framework.
This is not a minor bookkeeping artifact. It means the central claim of the ownership narrative โ that we know who is selling to whom โ is built on a foundation that does not reconcile. And a ledger that does not reconcile is not a ledger. It is a story.
Core, Continued: The Corporate Treasury Flywheel and Its Reversal
Here the audit turns from arithmetic to engineering, and this is where my 2020 work on Curve's stableswap invariant becomes relevant. When I simulated ten thousand arbitrage attacks against the AMM model, I was not testing whether the math was elegant โ it was. I was testing whether the incentive structure held under adversarial conditions. The invariant survived because its geometry was self-correcting. The corporate treasury model is the opposite: its geometry is self-reinforcing in one direction and self-destructing in the other.
Understand the mechanism. A company like Strategy โ holding 848,000 BTC, roughly 51% of the entire corporate sector's holdings โ does not simply buy bitcoin with cash. It runs a flywheel. The company trades at a premium to its net asset value (mNAV greater than one). That premium lets it issue equity or debt at a favorable rate, convert the proceeds into bitcoin, expand the net asset value, and โ so long as the premium persists โ repeat. Each turn of the wheel raises the per-share bitcoin backing. It is elegant. It is also entirely dependent on one variable: the premium.
Now watch what the data shows. In a single week, Strategy spent $142.5 million servicing interest and dividends, and $73.7 million repurchasing preferred stock โ while buying only 334 BTC for $28.7 million. Read that again in audit terms: the cost of maintaining the capital structure ($216.2 million) was more than seven times the capital deployed into the asset ($28.7 million). The maintenance cost has exceeded the accumulation. That is not a growth posture. That is a defense.
The repurchase of preferred shares is the detail I would circle in red. When a company starts retiring its most expensive capital instruments, it is signaling that its cheapest capital โ common equity issuance at a premium โ is becoming harder to access. The market's appetite for the premium is thinning. And when the premium thins, the flywheel does not slow. It reverses. Equity issued below net asset value destroys per-share backing rather than creating it. The engine that pushed bitcoin higher becomes the engine that pushes it lower.
Metaplanet, holding 44,000 coins with 85% to 90% of its total assets in bitcoin, illustrates the fragility in miniature. It sold 10,000 coins and repurchased 11,000 in the same period, framing the maneuver as a demonstration of liquidity. I have seen this pattern before, in the 2021 NFT generator review, where a project's public framing of an internal decision concealed a defensive reality. The phrase 'demonstrating liquidity' is the language of a treasury testing whether it can still sell. A company with genuine liquidity does not need to prove it. A company that must prove it is asking a question it fears the answer to.
Vulnerability is just a question unasked.
Concentration compounds all of this. Strategy alone holds more than half the corporate sector's coins. The corporate sector holds 8% of supply. That means a single company's balance sheet decision is a systemic event for an entire ownership cohort โ and, because that cohort's buying has been a load-bearing pillar of the 2024-2025 price narrative, a systemic event for the market's story about itself.
Core, Continued: Custody, Accounting, and the Sell-Then-Rebuy Signal
The deeper structural risk lives in a place most analysts never look, because it is not on the chain. It is on the balance sheet. And the balance sheet answers to a different set of rules than the protocol does.
Bitcoin at the protocol layer has no team, no admin keys, no upgradeable proxy, no pause function. In audit terms, its attack surface is remarkably small. But the entities holding the largest concentrated positions are not protocol-native. They are public companies subject to disclosure law, and their bitcoin is subject to accounting standards that treat it as an asset on a quarterly cadence. This creates a category of risk that did not exist when bitcoin was a self-custodied retail asset: forced transparency that amplifies fragility.
Consider the mechanism. When Metaplanet sells 10,000 coins, it must disclose the transaction as a material event. The disclosure is public. The market reads it as a signal โ correctly or not โ and prices it in. This means the corporate treasury's every defensive move is broadcast in advance of its full execution. The company cannot quietly de-risk. It can only de-risk loudly. And a loud de-risk is a self-reinforcing cascade: disclosure triggers selling, selling depresses price, depressed price shrinks net asset value, shrunken net asset value impairs the financing that sustained the position, impaired financing forces more selling.
This is reflexivity with a legal obligation attached.
I built a version of this logic into the AI-agent security framework I co-authored in 2025 โ the one that identified how AI hallucinations could trigger unintended smart contract interactions, and that led us to design a 'code-stasis' verification layer requiring human approval for high-value autonomous actions. The lesson we encoded was that autonomy without a circuit breaker is not efficiency; it is fragility wearing efficiency's clothes. Corporate treasuries running levered bitcoin exposure have no circuit breaker. They have a disclosure calendar.
The accounting dimension is subtler still. When a company sells coins and repurchases a similar quantity, it may be managing not economics but optics โ confirming that the position is liquid, satisfying auditor or exchange requirements, or resetting a cost basis for reporting purposes. The transaction is real on-chain. The economic substance may be cosmetic. A reader who sees 'sold 10,000, bought 11,000' and concludes 'net accumulation' has been shown a number without its purpose.
Security is the shape of freedom, but only when you can see the whole shape.
Contrarian: The Blind Spot Is Not the Protocol โ It Is the Custody Layer
Here is where I diverge from nearly every analyst writing on this topic. The market's entire risk apparatus is pointed at the wrong layer.
When bitcoin wobbles, the community interrogates the protocol: the mining hashrate, the node count, the mempool, the difficulty adjustment, the theoretical threat of a 51% attack. This is a security posture aimed at a threat that has not materialized in sixteen years and, by the structure of proof-of-work, is economically irrational to attempt. Meanwhile, the actual concentration of ownership has migrated to a layer with none of the protocol's guarantees: custodians, ETF trusts, and corporate treasuries. These entities have admin keys, boards of directors, leverage, and disclosure obligations. They have the very properties bitcoin was engineered to eliminate.
And the data describing them is unaudited, aggregated by third parties with opaque definitions, and reconciled by no one. We have built the world's most transparent settlement layer and then placed the majority of its ownership behind custodial labels we cannot verify.
I listen to what the compiler ignores.
The report's 'Wall Street versus ordinary people' framing is itself the tell. It is a narrative optimized for engagement, not for accuracy. The binary is false on its own terms. The 'ordinary people' category, as I have shown, is a residual bucket contaminated by lost coins, dormant Satoshi coins, and commingled exchange custody that represents institutional flows under a retail label. The 'Wall Street' category is dominated by a single company whose sustainability depends on a premium that is now visibly thinning. Neither side of the binary is what the headline says it is.
The genuine risk is not that Wall Street is taking bitcoin from ordinary people. It is that the entire ownership map is drawn with a ruler that no one has calibrated โ and that the most concentrated, most levered, most fragile holder in that map is also the one whose distress would be most publicly, and therefore most dangerously, disclosed.
Takeaway: What the Next Disclosure Cycle Will Reveal
I am not making a price call. My work in the aftermath of the Terra collapse in 2022 taught me to strip speculation from analysis and leave only structure. So let me leave only structure.
Watch three things, in this order.
First, the mNAV premium of the largest corporate holder. This is the single most important number in the corporate treasury model and the one least discussed. So long as it holds above one, the flywheel turns. The week it slips toward parity, the engine that has been a structural bid becomes a structural offer. The preferred-share repurchases already hint that management sees the premium thinning before the market does.
Second, the disclosure calendar. Every quarterly and material-event filing from the treasury cohort is now a scheduled volatility catalyst. When the cost of servicing a capital structure exceeds the capital deployed into the asset โ as it did by a factor of seven in the week the report cites โ the next filing will either confirm a defense or reveal a retreat. Either outcome is a directional signal.
Third, the free float. If the true tradeable supply is closer to 10 or 12 million than the headline 13.9 million, then every institutional bid and every forced corporate sale moves price more than the market believes. Thin float amplifies both the up-move and the unwind.
In the void, the bytes whisper truth โ but only to those who reconcile the columns.
The ledger is public. The ownership of that ledger is not. We have spent a decade celebrating the transparency of the chain while quietly accepting the opacity of everything built on top of it. The next chapter of bitcoin's ownership story will not be decided by the protocol, which will continue to do exactly what it has always done. It will be decided by whether the entities holding the concentrated positions can survive the disclosure of their own fragility.
That is the question the market has not asked. And a question unasked is the only vulnerability that never gets patched.