Why Ethereum, and not Bitcoin

Bitcoin is a bearer asset. You hold it and it does nothing; the only return is the price. Ethereum pays. Stake it and the network pays you for securing it — currently 2.66%3 a year, settled in ETH, funded by consensus rewards, priority fees, and MEV.

That single difference changes what a corporate treasury can be. A bitcoin treasury is a holding pen: value accrues only if the price rises. An ether treasury is an operating business with a revenue line. Its ETH count can grow without issuing a single new share.

Ethereum is also where the dollars already live, and the distribution is lopsided enough to check. 49%49 of all stablecoin supply sits on Ethereum, against roughly a third on Tron. Split by issuer it is sharper still: 72% of USDC is on Ethereum or its Base rollup, against 37%53 of USDT. The regulated, US-domiciled token is the Ethereum-native one, the offshore token is the Tron-heavy one, and regulated payment flow has been migrating from the second toward the first. If tokenised dollars become the rails for a meaningful share of global settlement, the fee and staking income of the chain beneath them is not a speculative claim on the future — it is a toll on activity already happening.

34.7% of all ether is currently staked, and every staked ether is one that cannot be sold without first leaving the validator queue. Supply tightens as the network is used more heavily. That is the opposite of how most assets respond to demand.

The gap between what it owns and what it earns

BitMine Immersion holds 5.82M ETH4 — 4.8% of every ether in existence, and 74.3% of all ether held by publicly traded companies. On that second measure it is not the leader of a category so much as it is the category.

But holdings are not the interesting number. Staking rewards accrue strictly in proportion to stake weight, so what a treasury actually earns depends on how much of its stack is staked rather than how much of it exists. BitMine stakes 87.1% of its ether. The network as a whole stakes 34.7%.

That gap is the entire business. 4.8% of the supply converts into 12.1% of every ether staked on the network, and therefore roughly that share of everything Ethereum pays out to validators. Measured per ether owned, BitMine earns about 2.51× what the average holder earns. Current run rate: $250M8 a year.

The rewards arrive denominated in ETH, not dollars. They are not a dividend to be spent — they compound the ether count without diluting anybody. Layered on top, the capital structure works in both directions: when the shares trade above net asset value, issuing equity to buy more ether raises ether-per-share; when they trade below it, repurchasing stock does the same thing from the other side. BitMine has bought back 20.8M10 shares since July.

What a share is actually worth

A treasury company is worth its assets multiplied by whatever the market will pay for them. Written out, that is: implied share price = ether per share × ether price × mNAV. Three inputs, no hidden terms — and you can set two of them yourself below.

Ether per share is the number that matters and the one most easily misread. BitMine’s total holdings only ever go up, but total holdings are a press release, not a result. What matters is holdings divided by shares outstanding, and the denominator moves too. Right now that is 0.009640 ETH, from 5.82M ETH4 spread across 603.2M9 shares.

One caveat the model cannot paper over: holdings are published weekly, but the share count is only confirmed quarterly on the 10-Q cover page. Ether per share therefore lags both new issuance and buybacks by up to a quarter, and the figure above is dated to the older of its two inputs rather than the newer. Right now that lag runs in the reader’s favour: 20.8M10 have been repurchased since the last cover date, worth roughly 3.6% to ether per share once the next filing confirms it, so the figures here understate the position.

That claim is checkable, so here it is checked. Ether per share rose 5.39% between November 2025 and today, so the flywheel did work across the period as a whole. But not every quarter earned it: the table below measures each one directly, and the Dec–Feb quarter went backwards despite raising billions. Comparing an average issue price against quarter-end net asset value would have called that quarter accretive — it mixes an intra-quarter price with an end-of-quarter denominator, and when ether is falling that flatters the multiple. The honest test is simply whether ether per share went up.

The premium is neither free nor permanent. Someone paying 1.55× net asset value receives well under a dollar of assets for every dollar handed over; across the three raises roughly $3.5B passed from incoming shareholders to existing ones. That transfer is the fuel, it is disclosed and legal, and it is running out: 1.55×, then 1.41×, then 1.05×, with the market now below parity. Under one the machine reverses — issuing destroys ether per share and retiring stock creates it, to the point where selling coins to fund a buyback still leaves every remaining holder with more coins each. That is why the company stopped selling shares in July and began buying them back.

Two scoreboards run at once here and no honest account should show only one. Ether per share is up. Measured against what was paid, BitMine has spent $19.1B acquiring a crypto position now marked at $10.9B — -43.0% against cost. The flywheel optimises coins per share, not dollars per share, and the two diverge whenever the coin falls faster than the accretion accumulates. Both statements are true simultaneously, and anyone who tells you only the first is selling you something.

mNAV is the multiple the market applies to that asset value. Above 1.0 you are paying a premium for the treasury; below 1.0 you are buying the ether for less than it is worth, and collecting the staking yield on top. It is not a constant and it is not a forecast, so the model below lets you choose it rather than accepting a number from me.

Tom Lee’s case, and how it has aged

The clearest published argument for why a treasury company should trade above the value of the coins it holds belongs to Tom Lee, BitMine’s chairman. It is worth setting out properly, because the model above leaves mNAV as a free parameter and his framework is a serious attempt to fill it in.

His method runs opposite to the usual one. Rather than forecasting gas fees and stablecoin volumes to derive a price, he starts from ether held per share and asks what premium the vehicle itself deserves. The reasoning behind that inversion is that no spreadsheet has reliably explained a crypto price a year ahead, and that precision in the earnings line is worth far less than being directionally right about the multiple.

The build-up has four parts. Net asset value sets the floor. The staking income is then capitalised — treated as a recurring earnings stream and given a money-market multiple, which on his assumptions lifts that floor well above parity by itself. Three further premiums are argued on top: velocity, for how quickly ether per share is compounding; liquidity, because a deeply traded stock can issue instruments more cheaply; and scarcity, because the largest vehicle in a category earns a network effect the others cannot.

A strategic argument sits underneath it, and it is the part most easily missed. An entity holding a large enough share of the supply becomes difficult to route around: a buyer who wants size cannot take it from the open market without moving the price against itself, which can make buying the vehicle the cheaper path. That is why the balance sheet and the compliance posture are strategy rather than housekeeping. A plain capital structure carrying no senior claims, with staking run domestically under US rules, is what keeps a company straightforward to acquire — and what makes regulated institutions comfortable with who is doing the staking rather than merely that it is happening. On this reading the clean balance sheet is not conservatism. It is what preserves the option of being bought as a strategic asset.

That specific claim has weakened since he made it. In June the company issued $350M32 of 9.5%33 perpetual preferred stock, which ranks ahead of the common and carries a fixed annual cost. It is a modest amount against an eleven-figure treasury, and it is not debt — but a senior claim is precisely the kind of structure the argument says to avoid, and it is the first thing an acquirer would have to price. The capital structure is no longer quite the one the thesis describes.

The historical analogy he reaches for is Exxon, which spent close to thirty years among the five largest companies in the S&P 500 priced on proven reserves rather than on earnings. In his words, “crypto treasuries are just the new Exxon.”

Two things to hold on to while reading the ladder below. The multiple applied to staking income is an assumption rather than an observed number, which is why this page makes it an input you can change instead of a constant you have to accept. And the velocity premium is left out of the arithmetic altogether: the figure behind it described BitMine’s first month of buying, and the company has since spent a quarter repurchasing its own stock rather than issuing it.

How a government actually sheds debt

Start with what does not happen: the debt is never repaid. Issuing bonds is refinancing, not reduction — maturing paper is swapped for new paper and the interest is added on top. At this scale the stock falls only through a sustained primary surplus, which the United States has not run since 2001, or through default, which is not available to the issuer of the world’s reserve currency. What falls instead is the burden: the debt measured against the economy servicing it.

That burden turns on a single inequality. When the effective interest rate on the debt sits below nominal growth, the ratio deflates on its own and no repayment is required. It currently does: 3.36%38 against 5.70%39, a gap of -2.34%. That differential alone moves the debt ratio by -2.24% of GDP a year before a dollar of new borrowing is counted.

Running against it is the primary deficit — the shortfall excluding interest — at 2.73% of GDP. Net the two and the ratio is rising by 0.50% a year, roughly $156.2B short of standing still. The United States is far closer to stabilising its debt than the headline figures suggest, and the entire question is whether that rate stays beneath that growth.

The instinct is to inflate the debt away. It does not work cleanly, because inflation erodes a fixed claim only across the term of the debt, and this debt is short: bills are 21.81% of it and reprice within the year. Raise inflation and the rate follows before the erosion arrives, so the debt simply refinances at the new level. Nothing else in this problem carries the leverage that the rate-versus-growth gap does.

What has actually worked is repression — holding the rate below inflation so that bondholders fund the reduction through negative real returns. The United States did precisely this after the Second World War, cutting the ratio by roughly three quarters across three decades with essentially no repayment, while regulated institutions were required to hold the paper. The modern version of that captive buyer is written into statute: the GENIUS Act obliges stablecoin issuers to hold full reserves in dollars or short-dated Treasuries, which makes every newly minted token a bid for bills placed because the law requires it, not because the yield is attractive. That suppresses the one rate the arithmetic above depends on.

Two things follow that most accounts of this get wrong. The reserve income never reaches the Treasury — the issuer keeps it, around $12.3B a year on the current float, while the government goes on paying the market rate on every bill it sells. And the holder is paid nothing at all, which leaves a real return of -3.40%. The channel is a seigniorage machine for the issuers, and only incidentally a relief mechanism for the government.

What it does provide is placement. The float already represents $246.4B of Treasury demand against $414.4B of net new bill issuance a year — a majority of the marginal supply, from a standing start — and it comes from a buyer obliged to appear regardless of price, which matters as foreign central banks step back. Extended, a crypto market of $4.27T would cover every new bill the deficit requires. That figure is derived from the ones above rather than chosen for effect.

One mechanical constraint bounds all of it. An issuer cannot manufacture demand: every token exists because someone wired dollars in, so the Treasury bid is entirely derivative of demand for the token itself. Buying ether with a stablecoin does not destroy the stablecoin, it hands it to the seller — rotation inside crypto contributes nothing at all. Only fiat crossing the boundary moves the float, and that money is the least durable in the system. Through the last bear market the float fell -34%54, and USDC — the Ethereum-native one this argument leans on — fell -56%55. A captive buyer forced to sell bills during a crypto drawdown is the wrong kind of buyer for the short end, arriving with supply exactly when everything else is stressed.

Three things cut against it, and any version of this thesis that omits them is selling something. Stablecoins are 12.13% of the crypto market today and that share compresses in a rally, so the scenario this argues for is the one in which each dollar of crypto market capitalisation carries less Treasury bid. The channel pulls issuance shorter, which is exactly what makes inflating away fail. And the measured price effect is small: BIS work finds an inflow of $3.5B47 moves three-month bill yields by 0.04%48 at peak, with little spillover to longer maturities — a local effect, not a slope that can be ridden to a solution.

Strip it back and the argument needs no particular market capitalisation. It needs the rate held below growth for long enough, which is a policy choice with a captive buyer behind it. That policy has an obvious payer and an obvious beneficiary. The payer is whoever holds the numéraire — cash, bills, tokenised dollars, all earning -3.40% in real terms. The beneficiary is whoever holds something that cannot be issued.

It also clarifies what Washington is buying. The stated motive is dollar competitiveness rather than debt relief: tokenised dollars extend the currency into populations American banks will never serve. The revealed preference is sharper still — the executive order establishing a bitcoin reserve permits the government to acquire more bitcoin and explicitly forbids it from acquiring ether. So the tailwind reaches Ethereum through one door only, and it is not a sovereign bid for the asset. Tokenised dollars have to settle somewhere, and whoever secures the chain they settle on is paid for it.

What kills this

Every argument above can be correct and this position can still lose most of its value. These are the specific ways that happens, in rough order of how much they worry me.

• A discount to net asset value can persist for years. Trading below NAV is not automatically an opportunity — it may be the market correctly pricing in dilution risk, governance risk, or plain doubt that shareholders will ever see the underlying assets.
• Dilution is the central danger of the whole structure. Issuing equity to buy ether only helps when the shares trade above NAV. Issue below it and ether-per-share falls even as total holdings rise. Every press release announcing record holdings must be read against the share count — which is exactly why this page derives per-share figures rather than quoting the headline.
• This is a leveraged bet on a single asset. A single-asset holding company amplifies the underlying in both directions, and ether has drawn down more than seventy percent within a cycle more than once.
• Staking is not risk-free income. Slashing, validator downtime, smart-contract risk in any liquid-staking layer, and a regulatory reclassification of staking-as-a-service are all live possibilities — and the last of those has taken down larger balance sheets than this one.
• Key-person and governance risk. The strategy is closely identified with its leadership, and the treasury-company structure concentrates capital allocation decisions in very few hands with limited practical check from shareholders.
• The macro thesis can be right while the equity still fails. "The government needs crypto adoption" does not entail "this particular equity appreciates." Regulatory capture could route the entire stablecoin bid through bank-issued tokens on permissioned chains, leaving public-chain economics untouched.
• The failure mode Lee himself names is leverage — exotic instruments and debt structures rather than plain equity. That was a fair description of BitMine when he said it. It has since issued $350M32 of 9.5%33 perpetual preferred stock, which is not debt but ranks ahead of the common and carries a fixed annual cost that has to be paid before common shareholders see anything. Net asset value on this page is stated after deducting it. Separately, the buyback cuts against the velocity argument: a company repurchasing stock is shrinking the share count rather than growing the stack, and the velocity premium and the buyback cannot both be the reason to own it.
• Value leaks to layer twos. Rollups settle to Ethereum but pay it far less than mainnet activity does — that is what a rollup is for. As stablecoin volume scales it migrates to cheaper execution, so the toll accruing to whoever secures the base layer shrinks per unit of activity. Counting Base alongside Ethereum, as the settlement figures above do, flatters the case for exactly this reason. BitMine’s staking revenue is a claim on layer-one economics specifically.
• The concentration risk is mine, not the company’s. Most of my capital sits in one ticker deriving its value from one asset. That is the single most likely reason this ends badly, and no amount of thesis quality compensates for position sizing.