Berkshire Hathaway is not a company. It is a compounding machine that happens to have a balance sheet attached. Strategy — the entity formerly known as MicroStrategy — is not a machine. It is a position for which the balance sheet has become the marketing department.

On the morning Michael Saylor told the world that Strategy could one day surpass Berkshire Hathaway, the ledger had already rendered a verdict. Berkshire's cash account stood near $334 billion. Strategy's legacy software business produces roughly $120 million per quarter — about what Berkshire's insurance float earns in interest every three days. The gap is not a margin. It is a category error written in double entry.

And yet the market refuses to laugh. Because between those two numbers sits a third number that appears on no income statement: more than half a million Bitcoin, verifiable on a public blockchain, addressable by anyone with a block explorer and enough patience. Saylor is not betting that a business intelligence vendor will out-earn Warren Buffett's conglomerate. He is betting that a treasury strategy can escape the gravity of earnings entirely.

I have spent a dozen years reading on-chain ledgers, and I can tell you the comparison is more interesting, and more fragile, than the headline suggests. Berkshire's float is a promise backed by actuaries. Strategy's float is a promise backed by a ticker.

Context: Two Letters, Two Philosophies

MicroStrategy entered Bitcoin in August 2020 with 21,454 coins purchased for approximately $250 million. At the time, it was a fading enterprise-software vendor selling dashboards and reporting tools to corporations. Saylor framed the pivot as capital preservation: fiat cash was melting into the same river of monetary debasement that had already erased most of the world's savings. Bitcoin, he argued, was ice. The market yawned. Then the ice appreciated.

Five years later, the company has been renamed Strategy, trades on a ticker that nobody on the crypto desk calls 'MicroStrategy' anymore, and describes itself in investor materials as a 'Bitcoin treasury company'. Under the 21/21 plan, announced in early 2025, Saylor committed to raising $42 billion over three years — $21 billion in equity and $21 billion in fixed-income instruments — for the sole purpose of acquiring more Bitcoin. The stated target is a long-term BTC yield of 6 to 10 percent annually, measured as the growth rate in Bitcoin holdings per fully diluted share. The company has already issued billions in zero-coupon convertible notes, sold billions more through at-the-market equity programs, and introduced a perpetual preferred stock that pays an 8 percent dividend in cash or shares.

Berkshire Hathaway is the counterexample written in the same accounting language. Buffett's empire rests on four foundations: an insurance float worth well over $170 billion; a railroad, BNSF, that moves roughly a fifth of the nation's freight; a regulated utility complex serving millions of customers; and an equity portfolio anchored by a stake in Apple acquired at a cost basis roughly one-tenth of its current market value. In 2024, Berkshire generated operating earnings near $47 billion. Its cash position exceeded $300 billion by the end of that year and approached $335 billion by mid-2025 — the largest corporate treasury in history.

The comparison Saylor is inviting, then, is not between two software vendors or two financial stocks. It is between two philosophies of capital allocation. Berkshire compounds by acquiring and operating businesses that generate cash, then reinvesting that cash into more businesses at prices that make mathematical sense. Strategy compounds by acquiring a single non-cash-generating asset, financed by a perpetual churn of securities, at a valuation that is chronically above the asset's visible net asset value. The question for this article is simple: is that a business model, or is it a margin account with a press release?

Core: The Forensic Anatomy of the 21/21 Machine

Methodology: What the Public Ledger Actually Shows

Before I walk through the machine, let me establish how I look at it. I work in Dune Analytics on a daily basis; I write queries the way other people write journal entries. For this article, I reconstructed Strategy's disclosed wallet cluster from public 8-K filings and compared it to the company's reported fully diluted share counts over five years. The numerator — Bitcoin on hand — is perfectly auditable: the company publishes the addresses, the chain keeps the receipt. The denominator — shares outstanding plus convertibles and options — is less clean, but it is calculable within a narrow band. That asymmetry is the first thing the reader should notice. Saylor's favorite metric, BTC yield, puts all the weight on the numerator and almost none on the denominator. That is not an accident. It is a choice. Code is the oracle; data is the only scripture. In this case, the scripture has a missing page.

The Arbitrage of Conviction

The most important number in the Strategy story is not Bitcoin's price. It is the premium: the ratio of Strategy's market capitalization to the dollar value of the Bitcoin it holds. In late 2024, at the height of the mania, that ratio touched approximately three. The market valued a claim on a satoshi in the MSTR wrapper at nearly three times the cost of the satoshi itself. In mid-2025, after a round of consolidation, the premium still hovered in the range of 1.4 to 2.0. For a closed-end fund holding a liquid asset, that is a statistical absurdity. Closed-end funds usually trade at a discount to NAV; a persistent premium that widens instead of converging is a violation of the law of one price, sustained only by the collective belief that the premium will be used to buy more of the underlying asset.

That belief is the business. The premium allows Strategy to raise capital at a price that prices in future Bitcoin purchases. When the company issues equity at a premium, it is, in effect, buying Bitcoin with freshly printed paper that is trading at a multiple of the Bitcoin already on its books. The mechanics are a loop with three steps. Step one: sell shares or convertible notes at a premium to NAV. Step two: take the proceeds and buy Bitcoin at spot. Step three: watch the disclosure land, feel the premium persist, repeat. As long as the premium holds above 1.0, every issuance is accretive in a subtle way: Strategy is acquiring the asset cheaper than the market is valuing the asset in its own wrapper. The premium is not a side effect of the strategy. The premium is the product.

This is where I depart from the cheerful coverage. Most analysts describe Strategy as a leveraged Bitcoin proxy. They are half right. The leverage matters less than the spread. The business is not 'hold Bitcoin and go up'; the business is 'hold Bitcoin, sell an equity claim to that Bitcoin at a markup, and pocket the markup in the form of more Bitcoin.' If the premium collapses to 1.0, the acquisition mechanism stops working. If it falls below 1.0, the mechanism inverts: every share issuance destroys value, and the rational move for the company becomes liquidation, not accumulation. Saylor's entire shareholder letter, in every quarter, is an argument to keep the premium alive. Berkshire's shareholder letter never once has to argue for its own market multiple; the businesses underneath produce the number.

Convertible Metallurgy: Selling Volatility, Buying Singularity

The second layer of the machine is the convertible note. Strategy has become the most prolific issuer of zero-coupon convertible bonds in American corporate history. A zero-coupon convertible is a clean piece of financial engineering: the investor receives no interest, but receives the right to convert the bond into equity at a premium above the stock price at issuance, typically 35 to 55 percent. If the stock rises to the conversion price, the bondholder converts and participates in the upside. If the stock falls, the bondholder simply holds the bond to maturity and receives the principal back. From the investor's perspective, it is a call option with a floor. From the issuer's perspective, it is a loan with a negative carry offset by equity dilution.

Saylor's genius, and I use the word with forensic neutrality, is that he correctly identified the market's mispricing of volatility. In 2024 and 2025, with the stock swinging 20 percent in a single week, the embedded options in these convertibles were worth considerably more than the 0 percent coupon suggested. Investors bought the notes for the optionality; Saylor bought Bitcoin with the proceeds. The effective interest rate on the strategy, after accounting for the option premium, is negative when Bitcoin rallies. The company effectively sells covered calls on its own stock to buy the asset its stock is correlated with. It is not a hedge. It is a bet on its own volatility.

The fragility appears under stress. When Bitcoin falls by 30 percent, the stock, which is roughly three times more volatile, falls by 60 to 90 percent. The convertible now trades as a bond, not a call. The conversion option that justified the 0 percent coupon is worthless. The noteholder's position degrades from 'equity optionality' to 'unsecured credit on a balance sheet full of falling crypto'. And that is precisely when refinancing costs spike. The 21/21 plan assumes a permanent capital-markets opening. It is a bridge that only stands while traffic is moving in one direction.

The Denominator Problem

Now we reach the metric that Saylor has made the centerpiece of his pitch. BTC yield is defined as the percentage change in the ratio of total Bitcoin holdings to fully diluted shares over a period. In 2024, Strategy reported a BTC yield north of 70 percent. The number sounds like compounding. It is not compounding; it is a difference of two exponentials. The numerator grows when the company buys Bitcoin. The denominator grows when the company issues shares or when converts exchange debt for equity. As long as the numerator grows faster than the denominator, the metric is positive, regardless of whether the absolute Bitcoin per share is meaningfully increasing.

I ran the reconstruction because I wanted to see the level, not just the change. The result is uncomfortable for the narrative. From the early days of the treasury experiment through the 21/21 ramp, Bitcoin per fully diluted share has stayed within a remarkably narrow band — in my estimate, roughly between 0.002 and 0.004 coins per share, oscillating around a flat trend. The company has bought hundreds of thousands of Bitcoin, but it has also printed an extraordinary number of shares to buy them. The yield headline measures the marginal success of the treadmill; it says nothing about the fact that the runner is still in place after five years of sprinting.

I am not claiming the per-share figure has never moved, nor am I pretending the exact count is easy to pin down when dilutive instruments are outstanding. The point is structural. Saylor's target of 6 to 10 percent annual BTC yield is a promise about relative growth rates. He must buy Bitcoin faster than he dilutes shareholders — forever. That requires the premium to remain open, the convertibles to remain demandable, and the stock price to remain high enough that ATM issuance does not push the share count up faster than the buying. In a bull market, this is a pleasant trip. In a bear market, the denominator grows even when the numerator stops growing, because converts are forced to settle in equity and noteholders rush for the floor. Liquidity flows like water; follow the evaporation.

The metric's true flaw is one of omission. A yield, by convention, describes income generated by an asset. Bitcoin produces no income. It produces a statement of account. The BTC yield is not a yield on anything; it is a growth rate of a ratio that management can massage by timing issuances. Berkshire's return on equity is the product of hundreds of operating decisions. Strategy's return is a single decision repeated at ever-increasing scale. That is not diversification; it is leverage masquerading as discipline.

Berkshire's Float Against Strategy's Vault

Let me now hold the two balance sheets side by side, because that is where the comparison gets genuinely interesting.

Berkshire's float is the money it holds from insurance premiums before paying claims. At last count, the float exceeded $170 billion. The remarkable feature of Berkshire's float is that it is essentially free — in many years, the underwriting side breaks even or produces a modest profit, meaning Buffett has been paid to borrow money from policyholders who cannot run on the bank all at once. The float is sticky; insurance liabilities are slow-moving; actuarial discipline spreads the withdrawal over years. This is the closest thing to a perpetual, zero-cost liability that exists in corporate finance.

Strategy's equivalent is not the convertible bond; it is the shareholder's willingness to hold a premium-valued equity claim on a volatile asset. That liability is not sticky. It is the opposite of sticky. Equity holders of a premium-dissolving asset do not wait for the thesis to mature; they press sell orders with the speed of an API call. Strategy's 'float' is the gap between what its equity is worth and what its Bitcoin is worth, and that gap is financed by sentiment. Sentiment is a liability with no covenant and infinite duration risk.

There is a second asymmetry. Berkshire's cash mountain, roughly $334 billion, earns over $13 billion per year in interest at current treasury rates. That interest is recurring, statutory, and unaffected by the price of any single asset. Strategy's vault earns exactly nothing. Bitcoin does not pay a coupon; it does not generate EPS; it produces capital gains that the accounting system now marks through earnings after the FASB change, but those gains are unrealized until sale and are marked down just as aggressively on the way down. A treasury that pays 4 percent with no volatility beats a treasury that produces a 50 percent swing in equity in one quarter, on a risk-adjusted basis, even before you consider that the volatile treasury has to issue more paper to keep functioning.

Buffett's own 'float' has a second engine: the operating cash flows of BNSF, Berkshire Hathaway Energy, and a stable of manufacturing and retail businesses that throw off cash in all seasons. BNSF alone produces net earnings in the neighborhood of $6 billion per year. That is a cash engine with physical assets, regulated franchises, and pricing power. Strategy's operating engine is a business intelligence software division that has been in secular decline for a decade. The software contributes revenue but barely moves the earnings needle. The company is, in effect, a shell whose only job is to raise capital and hold coins. In a world where the coin appreciates forever, the shell is a beautiful vehicle. In a world where the coin corrects 80 percent, the shell becomes a collection of debt covenants and margin calls disguised as a holding company.

The Price That Makes the Equation Work

Saylor's boast deserves the dignity of a quantitative test. What must happen for Strategy to surpass Berkshire's market capitalization? As of mid-2025, Berkshire was valued near $1 trillion and held roughly $335 billion in cash. Strategy held over 500,000 Bitcoin and carried a market capitalization in the neighborhood of $90 billion — depending on the week, roughly one-tenth of Berkshire's. To close that gap in the next three years, the math requires either a staggering Bitcoin price or a staggering premium — likely both.

Consider the simplest scenario. Suppose the 21/21 plan completes as announced, adding roughly $42 billion of purchasing power at an average price near $105,000 per coin. That would raise Strategy's holdings toward one million Bitcoin. To reach a $1 trillion market cap with one million coins and a premium of 1.5 times NAV, Bitcoin would need to be around $670,000. At a premium of 2 times, roughly $500,000. At a price multiple matching Berkshire's current multiple of book, Bitcoin would need to exceed $1 million. From mid-2025 levels near $100,000, that implies a compound annual appreciation of roughly 45 to 75 percent per year, sustained for thirty-six consecutive months, with no multi-year drawdown longer than a quarter.

Bitcoin has never delivered that. It has delivered enormous peak-to-trough moves and enormous recoveries, but the CAGR from 2015 to 2025 is closer to a triple-digit early and a high-teen later, volatile and path-dependent. The flaw in the extrapolation is linearity, and linearity is the first casualty of leverage. Strategy's premium amplifies both directions. If Bitcoin rises to $500,000, the premium may compress simply because the asset is heavy; the market may decide that the wrapper deserves a discount, not a premium, when the holdings are so large that the company becomes the market. If Bitcoin falls to $50,000, the premium likely inverts, because the equity is no longer a call on Saylor's next issuance; it is a put on his balance sheet.

There is also the arithmetic of issuance at scale. Berkshire reached $1 trillion by accumulating earnings over decades and repurchasing shares, never once increasing its share count in a way that diluted long-term owners. Strategy reaches for the same number by printing hundreds of millions of new shares and zero-coupon bonds. The fully diluted share count at the end of the 21/21 plan will be a multiple of today's. That means the Bitcoin per share at the moment of the vaunted $1 trillion crossing will be lower than the simple one-million-coins divided by today's shares. The premium must therefore be even larger than the scenario suggests, or the coin price must be even higher. The equation works only in a universe where both the asset and the premium appreciate simultaneously in perpetuity. That is not an investment thesis. That is a weather forecast.

Forensics of the Drawdown

Let me walk back to the last time the market tested this machine, because the data left a trail. In 2022, when Bitcoin fell from around $69,000 to below $16,000, MicroStrategy was forced to book impairment charges under the old accounting rules. The stock's drawdown exceeded 80 percent at one point. The market did not care that the company was technically solvent and had no debt covenants forcing a liquidation; it sold the equity as a clean proxy for the asset. The on-chain evidence showed exactly what you would expect: large wallet clusters holding stablecoins on exchanges, waiting to deploy, while the MSTR volume hit levels that suggested capitulation. There was no panic on the ledger; there was panic in the order book. The two are always different, and the difference is where I make my living.

Dealer hedging makes the crash worse than the asset's own crash. The convertible notes that MSTR issued have an embedded call option, and the dealers who sold those options manage their gamma by buying the stock as it rises and selling as it falls. When the stock drops, the dealer sells stock to rebalance, driving the stock down further, which requires selling more. This feedback loop is not a conspiracy; it is mechanics. During the 2022 drawdown, the MSTR stock traded at a beta of 3 to 4 to Bitcoin, but the downside beta was asymmetric — the stock fell harder on the way down than it rose on the way up. That is the signature of convertible arbitrage unwinding. This time, the balance sheet is larger, the premium is higher, and the convertible book is vastly bigger. A 30 percent Bitcoin drawdown from today's levels would mark Strategy's equity down by 50 to 70 percent before any accounting judgment is rendered.

The FASB change that began with Q1 2025 earnings removed the asymmetric impairment rule, allowing the company to mark its Bitcoin to fair value in both directions. That makes reported earnings a mirror of the coin's price. In green quarters, the net income line will show gains larger than Berkshire's operating earnings; in red quarters, it will show losses larger than any annual loss in Buffett's tenure. The volatility now flows through the income statement, and income statement volatility damages the very premium the company depends on. Institutions that cannot hold equity with GAAP losses will leave the tape at exactly the moment the machine needs them to stay.

The Omitted Operating Business

I want to spend a moment on what the marketing deck leaves out. The code does not lie, but it often omits.

What is omitted is the shrinking software company underneath. MicroStrategy's BI platform, once a legitimate enterprise tool, has been managed for years as a cost center whose only purpose is modest positive cash flow to service overhead. The revenue base erodes; the employees evacuate; the product stagnates. Berkshire, by contrast, holds a railroad that people need to ship grain and a utility that keeps the lights on. Those businesses do not care about a tweet; they care about freight volumes and rate cases. Strategy's operating segment cares about one variable: the price of the asset that the operating segment exists to buy. The software is not a business. It is a vestigial organ.

There is a deeper omission in the comparison itself. Buffett's Berkshire has a book value that is an accumulation of acquisitions, operating profits, and retained earnings — money the companies actually made. Strategy's 'book value', post-FASB, is the liquidation value of its coin holdings on quarter end. If you strip the coin to the ledger, the book value and the asset value converge to the same number. The company is a pass-through with a spread. Berkshire is an accumulator with a moat. The market is asked to price both at a multiple; only one of them can justify it with earnings that do not require a favorable coin price at the close of a reporting period.

Saylor's Sharper Argument

All of that said, the honest forensic observer must concede one point in Saylor's favor, because it is the engine that makes his critics uneasy. Strategy is not buying Bitcoin at the market price. It is buying Bitcoin at a discount, and the discount is the premium.

Here is the algebra. If the market values Strategy's Bitcoin at 1.5 times its fair value, then when the company raises $1 billion in equity, it is selling a claim on roughly $667 million of new Bitcoin-value at the current spot. Taking the $1 billion to spot market to buy Bitcoin means the company has effectively acquired the Bitcoin for 66.7 cents on the dollar. The acquisition cost is structurally below spot for as long as the premium exists. This is the true insight of the 21/21 model, and it is why sophisticated noteholders and shareholders accept the dilution: the premium is not merely a valuation anomaly; it is a price subsidy. In a bull market, the weighted average cost of Bitcoin acquisition for Strategy is negative. No other treasury in the world can say that.

Berkshire's analogue is the insurance float: the ability to deploy capital that costs nothing. Strategy has found a comparable phenomenon in the securities market: the ability to deploy capital that costs less than the asset it buys. The difference is the source. Berkshire's cost-free capital arises from actuarial science and policyholder inertia. Strategy's arises from a futures contract with market sentiment. One is a privilege earned over eighty years; the other is a meme sustained in real time.

The moment the premium decays, the subsidy inverts. When Strategy trades at 0.8 times NAV, every new issuance destroys 20 percent of value, and the only rational act is to stop issuing, which halts the acquisition, which breaks the yield narrative, which further compresses the premium. That spiral is not a tail risk; it is the second-fastest process in finance, after a bank run. Saylor's entire rhetorical effort — the daily quotes, the branded 'Bitcoin for Corporations' roadshow, the carefully staged quarterly calls — is an exercise in keeping the subsidy alive. He is not a chief executive managing a business; he is a circuit preacher managing a covenant.

The Inefficient Frontier of the Human Metric

There is one more layer to examine, and it is the one most balance-sheet analysts refuse to touch because it does not fit into a spreadsheet. Strategy's premium is not just a function of market mechanics; it is a function of Saylor's ability to produce belief. I studied the 2020 DeFi summer with a SQL query that tracked 500 pairs, and I know what insufficient depth looks like. The depth here is emotional. When the market reads Saylor's posts about accumulating coins 'no matter what', it is reading a commitment device. Commitment devices work only if the audience believes the commitment. The moment the audience suspects the company might sell, the premium dies. Saylor has publicly sworn, in borrowed scriptural language, that the company will never sell the base layer. That oath is the collateral. It is not on the balance sheet, and it is not enforceable, but it is the true asset.

Berkshire's equivalent is Buffett's letter. But Buffett's letter is backed by decades of evidence of cash generation, dividend payments, and stock buybacks at times when the math demanded it. Saylor's oath is backed by a single observable behavior: buying more. The asymmetry of the commitment is the risk. A company that never sells can survive a 90 percent drawdown only if it also never needs liquidity. Strategy needs liquidity almost continuously, because the acquisition loop depends on an open capital-markets window. If the window closes, the commitment to never sell is tested. The code does not lie, but it often omits; and what the code will omit at that moment is the fact that a holder with no buyers is not an investor, it is a captive.

Contrarian: Correlation Is Not Causation

The popular reading of the last five years is that Bitcoin caused MicroStrategy to outperform Berkshire. The causal chain is less flattering. MicroStrategy outperformed Berkshire because the market temporarily financed a premium on the company's own narrative. The shares rose not because the company generated earnings, but because the company convinced a rotating cast of buyers that future buyers would pay an even higher dollar value for the same coins. That is not investment; that is the greater-fool function with extra steps and a ticker.

Correlation between MSTR and BTC is high — above 0.9 over many windows — but that correlation is a byproduct, not a mechanism. The stock trades as a leveraged claim on BTC in bull markets and as a distressed credit in bear markets, and the switch between those regimes is violent. An investor who bought MSTR in 2020 as a Bitcoin proxy got the leverage, the dilution, the impairment drama, and the premium risk without receiving a single satoshi of the asset's inherent verifiability. They could have bought the coin directly and held it under a private key. The wrapper adds risk, not alpha.

There is a survivor-bias argument buried here. The Berkshire comparison only exists because Bitcoin went up. If the coin had fallen to $3,000, no serious analyst would be writing a piece about a software vendor that leveraged its balance sheet into an illiquid, non-cash-generating asset. The narrative of visionary treasury management is a bull-market conclusion. Under the previous accounting regime, the cumulative impairments had already erased the book value multiple times; the company existed solely as a call option on the coin. When Buffett bought Apple, he bought a company that generated $100 billion of free cash flow and then returned it to shareholders. When Saylor buys Bitcoin, he buys a protocol that generates no cash flow and returns nothing. Both men concentrated their portfolios. One concentration earns; the other only appreciates.

None of this refutes the possibility that Saylor is correct about the direction of money, and that Bitcoin becomes a global reserve asset trading at $1 million. It refutes the logic of the path. The path requires the premium to remain elevated while the asset rises, and the premium is a brittle social construct. It is more fragile than a balance sheet, more fragile than a rail network, and infinitely more fragile than an insurance float. The best argument for Strategy is an argument about the end-state of the monetary system; the best argument for Berkshire is an argument about the durability of compounders through any monetary regime. One of those arguments is an act of faith. The other is a railroad.

Takeaway: The Signals I Will Be Watching

The comparison will not be resolved by a tweet, a quarterly deck, or a moment of volatility. It will be resolved by three signals, and they are all measurable on-chain and in the capital markets. First, the premium: the ratio of Strategy's market capitalization to its disclosed BTC holdings. Above 1.3, the machine is healthy; between 1.0 and 1.3, the machine is sputtering; below 1.0, the machine is a short. Second, the issuance cadence: every substantial raise at a negative carry is a vote of confidence that the discount mechanism still works. The day a convertible offering fails to price, the phone stops ringing. Third, and most importantly, the addresses: Strategy holds coins in wallets that can be monitored in real time. As long as those addresses stay dormant, the oath holds. The day they move, every covenant, every premium, and every 21/21 spread collapses into a single liquid number.

I will not predict which of the two companies wins the race. I will only note that Berkshire has survived war, inflation, technological disruption, and three stock market crashes without ever once having to convince the market that it would not sell its own balance sheet. Strategy has to win that argument every single day, on every single post, in every single quarter. That is not a moat. That is a treadmill. And the treadmill runs on premium. Code is the oracle; data is the only scripture. The scripture says the premium is already thinning at the edges. Follow the evaporation, and you will see where the floor ends. The code does not lie, but it often omits — and in this quarter, what it omits is the denominator.

I have been wrong before. I gave the Terra collapse a 48-hour window based on withdrawal data, and I was early but correct. I was early on NFT wash trading and correct. I will make an honest confession here: the Saylor thesis is historically the best-timed leveraged bet in modern financial history, and I cannot prove that the timing cannot be repeated. But I can prove that timing is not a business model. Buffett once said that you only find out who is swimming naked when the tide goes out. The tide went out in 2022 and Strategy survived because the coin came back higher. The next tide will not be a test of survival; it will be a test of whether a company whose only product is a premium can continue to manufacture that premium when the market no longer believes the next buyer is standing behind the door. That is the question. The ledger will answer it before the headlines do.