Capital B's $78 Million Loss Is 97% Accounting Artifact. The Other 3% Is Where the Risk Lives.

Ninety-seven percent of Capital B's first-half net loss never touched a bank account.

The company reported a net loss of $78 million. Inside that number sits an impairment charge of €67.5 million — $75.9 million at the implied cross. Subtract one from the other and you get roughly $2.1 million of everything else. Staff. Audit. Listing fees. Interest. Whatever the operating business actually did.

So the headline is a $78 million loss. The reality is a $2.1 million operating result wrapped in a $75.9 million non-cash accounting entry.

I have seen this movie before. In late 2017 I wrote a Python scraper against the Ethereum mempool during the Tezos ICO because I did not trust the Telegram channels. Everyone was quoting the raise. Nobody had read the vesting schedule. The number that mattered was buried in the contract, not in the announcement. Same structure here. The number that matters is not the size of the loss. It is the accounting regime that produced it.

Let me be precise about what changed and what did not.

Capital B's cash: unchanged by the impairment. Its Bitcoin count: unchanged. Its debt: unchanged. Its obligations to counterparties: unchanged. What changed is the carrying value of an intangible asset on a balance sheet, marked down to a recoverable amount as of some period-end date the source material does not specify.

That is not a loss. That is a label.

And labels are where the money gets made and lost, because most people price the label and never open the box.

Context: What A Bitcoin Treasury Company Actually Is

A Bitcoin treasury company is a simple machine with a complicated reputation.

It raises capital — equity, convertibles, preferred, private placements, sometimes debt — and converts that capital into Bitcoin. It does not build software. It does not sell a product. Its output is one number: BTC per share.

That is the whole thesis. Everything else is financing detail.

The category has a hierarchy. Strategy sits at the top with the largest book, the deepest toolkit, index inclusion, and enough liquidity to absorb institutional size without shattering the spread. Metaplanet built the Asian franchise on retail distribution and a favorable tax treatment. The Blockchain Group took the European route through French listings. Below them sits a long tail of small-cap wrappers, several of which are former shell companies that discovered a narrative and re-rated overnight.

Capital B belongs to that tail. I want to flag the entity-identity question explicitly, because the French treasury sector has enough name overlap — The Blockchain Group, Capital B, TOBAM-adjacent vehicles, various placements and rebrands — that serious diligence should confirm the listing, the ISIN, and the actual filing before treating any figure as a fact. The source material for this piece does not provide the reporting period, the Bitcoin holdings, the average cost basis, the share count, or the mNAV. That is five load-bearing numbers missing out of five. I will mark confidence levels throughout, because the alternative — pretending one line item is a full picture — is how people lose money in this asset class.

What the source material does give us is the split. $78 million net loss. $75.9 million impairment. That ratio is not a coincidence. It is a structural signature.

The Two Accounting Regimes

Two regimes govern this asset class, and they are not symmetric.

Under IFRS — which a French issuer would typically apply — crypto assets held for long-term appreciation fall under IAS 38, Intangible Assets. The default measurement model is cost. You carry Bitcoin at what you paid, less any accumulated impairment. You test for impairment at least annually and whenever there is an indicator of a decline. Impairment flows to profit and loss. Reversal, under IAS 36, is permitted — capped at the carrying amount that would have existed had no impairment ever been recognized. For a non-amortized asset, that cap is original cost.

There is a theoretical alternative. IAS 38 permits a revaluation model, but only where an active market exists. The IFRS Interpretations Committee's 2019 agenda decision on holdings of cryptocurrencies effectively closed that door for most issuers by making clear that the cost model is the practical path for intangibles in this class. A handful of companies have argued an active market exists. Almost none have successfully applied it in a way that survived an audit committee.

There is also a classification fork. If you hold crypto for sale in the ordinary course of business, you are a broker-trader and IAS 2, Inventories applies — measured at the lower of cost and net realizable value. Still asymmetric. Still a ratchet. Still no mark-up until sale.

Now the other side.

Under US GAAP, ASU 2023-08 changed the game for fiscal years beginning after December 15, 2024. Crypto assets in scope are measured at fair value every reporting period, with changes flowing through net income. Both directions. Up and down. No impairment test. No reversal mechanics. Just a continuous mark.

| Dimension | IFRS (IAS 38 cost model) | US GAAP (ASU 2023-08) | |---|---|---| | Measurement basis | Cost less impairment | Fair value | | Upward price moves | Not recognized until disposal or reversal event | Recognized every period | | Downward price moves | Impairment to recoverable amount | Recognized every period | | Reversal permitted | Yes, capped at original cost | No — not needed, already marked | | Reporting frequency | Test at period end or on indicator | Every reporting period | | P&L character | Ratchet — down, then conditionally up | Continuous two-way mark |

That table is the entire story. Everything else in this article is a consequence of it.

The ratchet is the thing to internalize. Under cost-plus-impairment, a Bitcoin treasury company's income statement is structurally biased toward recognizing losses during drawdowns and toward silence during rallies. Price falls — impairment hits the P&L immediately. Price recovers — nothing happens until either a formal reversal test or an actual disposal.

So a "widened net loss" at an IFRS filer is, to first order, a statement about where the period-end price sat relative to the historical cost basis. It is not a statement about whether management is competent. It is not a statement about whether the coins are still there. It is a statement about a date on a calendar and a number on a spreadsheet.

And in a bear market — which is where we are — that ratchet bites hardest, because every reporting period re-tests the same underwater position against a lower or flat price.

Core: Opening The Box

The Arithmetic Of The €67.5 Million

Start with the cross.

The impairment is €67.5 million, presented as $75.9 million. Divide. 75.9 / 67.5 = 1.1244.

That implied EUR/USD rate pins the translation to a window when the euro traded in the 1.12 handle. It does not give us the period-end date, but it narrows it. This is a small forensic trick that costs nothing and pays for itself constantly: when someone hands you a converted number, back out the cross. It tells you which quarter you are actually looking at, and it tells you whether the number was converted at the period-end rate or at some other rate entirely. Both are informative.

Now the more interesting back-solve. What does a $75.9 million impairment imply about the size of the book?

Under IAS 36, impairment equals carrying amount minus recoverable amount. For Bitcoin, recoverable amount is essentially market price. There are no cash flows generated by holding a coin, so value in use collapses to the exit price. The write-down is therefore the gap between what the company paid for the impaired tranches and what those tranches were worth at the test date.

If the write-down is x percent of the impaired carrying amount, then impaired carrying amount = $75.9M / x.

| Depth of markdown | Implied impaired carrying amount | |---|---| | 10% | ~$759M | | 15% | ~$506M | | 20% | ~$380M | | 25% | ~$304M | | 30% | ~$253M |

Three caveats, and they matter more than the table.

First, only the underwater tranches get impaired. If management bought at $40k, $70k, and $95k, and the test price was $80k, only the last cohort is written down. The impairment measures the underwater slice, not the whole book. Total holdings could be materially larger than the implied figure.

Second, the testing unit matters. IAS 36 impairment testing for a fungible asset class is typically done at the level of the asset as a whole, not lot by lot. Under a whole-position test, the write-down is the gap between average carrying cost and market. Under a lot-level test, it is the sum of individual gaps. The disclosed number does not tell us which method was applied, and the two methods can produce very different figures on the same portfolio.

Third, timing. If the test date is a half-year end, the impairment reflects one specific day's close. Bitcoin does not respect reporting calendars.

Confidence on all of this: low. I am reverse-engineering a balance sheet from a single line item. But the direction of the inference is sound. An impairment of this size implies a book that is not trivial, and a cost basis that sits above the test-date price. That is real information. It tells you management bought into strength.

The 97/3 Split

Now the number that actually surprised me.

Net loss $78 million. Impairment $75.9 million. Residual $2.1 million.

A listed company on a European exchange — statutory audit, legal counsel, D&O insurance, listing maintenance, at least a skeleton staff, and custody fees on a nine-figure Bitcoin position — typically burns more than $2.1 million in a half year. Institutional custody alone runs in the tens of basis points annually. On a $400 million book, that is $1–2 million a year before you pay a single employee.

So either Capital B is extraordinarily lean, or the residual is being flattered by offsetting items — interest income on cash, FX translation gains, a lending yield on part of the Bitcoin, a one-off disposal. The source material does not say, and I am not going to invent it. Anyone who fills that gap with a confident explanation is guessing and calling it analysis.

But the structural conclusion holds regardless: this entity is functionally a Bitcoin exposure vehicle with a corporate wrapper attached. Its P&L is the P&L of an unhedged long Bitcoin position, filtered through an accounting regime that only lets the losses through.

Which raises the question every holder should be asking. If the operating business is a rounding error, what exactly are you buying?

Answer: you are buying BTC per share, manufactured through capital markets. And that brings us to the only metric that matters.

The Flywheel, Written Out

Most commentary on treasury companies hand-waves the mechanics. Let me do the algebra instead.

Let N = shares outstanding. Let B = Bitcoin held. Let P = share price. Let Q = Bitcoin price.

BTC per share = B / N.

Market cap = P × N.

Net asset value = Q × B.

mNAV = (P × N) / (Q × B).

Now the company issues ΔN new shares at price P, raising ΔN × P dollars, and buys Bitcoin at Q. New Bitcoin purchased = ΔN × P / Q.

New BTC per share = (B + ΔN·P/Q) / (N + ΔN).

Define m = P/Q — the share price expressed in Bitcoin per share, which is just mNAV × (B/N). Substitute:

New BTC per share = (B/N) × (N + ΔN·m) / (N + ΔN).

Accretion requires (N + ΔN·m)/(N + ΔN) > 1, which reduces to m > 1, which reduces to mNAV > 1.

That is it. That is the whole machine. When the market values the wrapper above the thing inside it, issuing shares creates Bitcoin per share out of nothing but sentiment. When the market values the wrapper below the thing inside it, issuing shares destroys Bitcoin per share.

No operating skill. No moat. No proprietary technology. Just a multiple.

| mNAV at issuance | Effect on BTC/share | Rational management action | |---|---|---| | 2.0 | Strongly accretive | Issue aggressively | | 1.5 | Accretive | Issue | | 1.0 | Neutral | Indifferent | | 0.8 | Dilutive | Buy back, do not issue | | 0.5 | Severely dilutive | Model is broken |

Now the reflexivity. mNAV is not a fundamental. It is a sentiment multiple applied to a commodity position. It expands when the narrative is hot and capital is cheap. It compresses when the narrative cools, when the equity gets de-rated, when the convertible market shuts, when the index committee removes you, when a large holder decides to rotate.

And here is the asymmetry nobody prices: the flywheel runs on issuance, but the impairment runs on price. They are not the same variable. A company can have mNAV below 1 — flywheel off, no more accretive raises — while still holding a book that keeps generating impairment charges every reporting period. That is the slow-motion failure mode. It does not require fraud, leverage, or a bank run. It requires a persistent discount to NAV plus a volatile underlying asset.

The floor is a suggestion, not a law.

What The Impairment Does Not Tell You

Three things the number cannot answer, and each one is load-bearing.

One: did BTC per share fall? The impairment is a markdown of carrying value. It does not change the number of coins. It does not change the number of shares. If no equity was issued during the period, BTC per share is flat — the shareholder owns exactly the same claim on exactly the same pile. If equity was issued at a discount to NAV, BTC per share fell, and the impairment line tells you nothing about that. The dilution lives in the share count, not the P&L. Anyone reading the loss as a proxy for shareholder damage is reading the wrong statement.

Two: what is mNAV? Not disclosed. If it is above 1, this is a non-event and the company can raise again on accretive terms. If it is below 1, the impairment headline is a distraction from the actual problem, which is that the capital-raising engine has stalled and the equity is trading below the sum of its parts.

Three: what do the debt documents say? This is the one I would want first. Impairment is non-cash, but covenants are written on reported figures. If a convertible indenture or a credit facility carries a net-asset or leverage covenant keyed to the balance sheet, a $75.9 million write-down can trip it without a single satoshi moving. And if the company has any Bitcoin-backed lending, the loan-to-value is marked against the same price that drove the impairment. A margin call is a cash event triggered by an accounting one.

Liquidity vanishes the moment you need it most.

I have watched this exact transmission channel before. In May 2022 I was running a delta-neutral short against the UST/LUNA pair, funded by stablecoin lending on Aave. The trade worked — up roughly 150% — but the interesting part came afterward. Influencers who had correctly called the collapse immediately rotated into new "safe" assets. I pulled SOL's validator set and found roughly 30% of stake sitting with one exchange. The lesson was not that Terra was bad. The lesson was that the thing everyone was calling decentralized had a single point of failure that nobody had bothered to look up.

Same method here. The reported loss is the headline. The load-bearing question is who holds the paper, on what terms, and at what mark.

The Reversal Option

Here is where the IFRS/GAAP split stops being an academic footnote and becomes a tradeable asymmetry.

IAS 36 permits reversal of an impairment loss for an intangible asset when the recoverable amount increases, capped at the carrying amount that would have been determined had no impairment been recognized. For a non-amortized Bitcoin position, that cap is original cost.

US GAAP under ASU 2023-08 prohibits reversal entirely, because it does not need one. The asset is marked to fair value every period in both directions. There is nothing to reverse.

Translate that into market terms. An IFRS treasury company that has taken a large impairment is carrying a free, embedded, non-cash call option on its own income statement. If Bitcoin recovers above the impaired carrying amount, the next reporting period books a reversal gain. Not a sale. Not a cash inflow. A line item that flips from negative to positive, mechanically, because a price moved.

And the market will read it as a turnaround.

That is the mispricing. The reversal is arithmetic, not performance. It is the same spreadsheet run in the other direction. A trader who understands the accounting regime knows the reversal is coming before the headline does, and knows it is not information. A trader who does not will bid the "return to profit" press release and get run over when the next impairment lands.

This is the same class of error I exploited in early 2024. Ahead of the spot Bitcoin ETF approvals, implied volatility was priced artificially low because institutional models were importing equity-vol assumptions into an asset with crypto-specific liquidity risk. I bought both legs of a straddle for a combined premium around $1.2 million. Approval came, price spiked, then corrected hard on miner supply, and the vol expansion let me exit both legs for roughly 65%. The trade was not a directional call. It was a bet that the market had mispriced a structural feature — a regime change — as a sentiment event.

Accounting regime differences are the same kind of feature. They are not opinions. They are rules, published in advance, that determine which direction a P&L can move and when.

Options give you the right to walk away. Accounting regimes give companies the right to re-mark. Know which one you are holding.

The Stale Balance Sheet

One more mechanical point, because this is where the timing edge lives.

Under the cost-plus-impairment model, the balance sheet is only marked at test dates. Between test dates, it is frozen. If Bitcoin rallies 30% after the period end, the reported NAV understates the position by roughly that much on the impaired tranche. If Bitcoin falls 30%, it overstates it.

That means IFRS treasury companies systematically publish stale NAVs, and the staleness has a sign. After a drawdown, the reported book overstates. After a rally, it understates. Anyone valuing these equities off the last filed balance sheet is valuing a photograph of a market that no longer exists.

Under fair value accounting, that lag disappears. Which means — and this is the part I would flag to any allocator running a basket of these names — a portfolio of IFRS and GAAP treasury companies is not a portfolio of the same asset. It is a portfolio of the same asset with two different reporting lags and two different convexity profiles. The GAAP names mark continuously. The IFRS names mark in steps. In a fast market, those are not the same instrument.

Volatility is just noise waiting to be priced. But a reporting lag is not noise. A reporting lag is a structural feature, and structural features get priced eventually — usually by whoever notices them last, at the worst possible moment.

The Narrative Turn

Media framing is data.

The headline used "wider net loss." Not "impairment." Not "non-cash write-down." Not "carrying value adjustment." The word choice moves the reader from accounting to operations. "Wider loss" implies the business is deteriorating. An impairment charge implies an asset was re-marked.

Someone chose that frame. Possibly the company's own release. Possibly the aggregator. Either way the choice is informative. When a treasury company wants to signal strength, the release leads with BTC per share and holdings. When it wants to bury something, it leads with the GAAP-shaped headline and puts the mechanics in paragraph four.

Watch which number gets the headline. It tells you what the issuer thinks the audience will not read.

I will also note the gap in the source material itself. No reporting period. No holdings. No average cost. No share count. No mNAV. That is not a small omission — those five numbers are the entire analytical frame. Any conclusion drawn without them is inference wearing the costume of measurement. I have marked confidence throughout for that reason.

Chaos is just data with no label yet. But an unlabeled number is not the same thing as a measured one, and confusing the two is how people talk themselves into positions they cannot defend.

The Sector, On One Page

| Entity | Jurisdiction | Scale | Financing toolkit | What to check first | |---|---|---|---|---| | Strategy | US | Largest | Convertibles, preferred, ATM, index inclusion | mNAV, convertible strike stack | | Metaplanet | Japan | Large regional | Retail equity, warrants | mNAV, dilution cadence | | The Blockchain Group | France | Mid | European placements | IFRS treatment, holdings disclosure | | Capital B | France (unconfirmed) | Small | Unclear | Everything — see above | | Long tail (Semler, Nakamoto, Twenty One, others) | Various | Small | Shell plus narrative | Whether an operating business exists at all |

I will not pretend this table is precise. The only row I would defend without a filing in front of me is Strategy's, and even that changes quarterly. The Capital B row is deliberately thin because the source material is thin. Treat the blank cells as the finding.

The category's economics are not competitive in the normal sense. They are competitive in the financing sense. The winner is whoever can raise the most capital at the highest multiple to NAV and convert it into coins fastest. That is a capital markets game dressed as a technology bet. There is nothing wrong with that. But price it as what it is.

And there is a concentration point here that mirrors what I found in validator sets. A treasury company is a single-asset, single-custodian, single-jurisdiction, single-management-team exposure. If the custodian fails, the coins are gone. If the multi-sig has a race condition, the coins are gone. I have audited contracts where the entire "security model" was a comment in the README and a promise in a blog post. When you buy BTC per share from a wrapper, you are buying the wrapper's operational competence too — and that is almost never in the deck.

Contrarian: The "Non-Cash, Ignore It" Take Is Correct And Also The Trap

The consensus reading of this headline is already forming: impairment is non-cash, ignore it, the coins are still there.

That reading is correct. It is also exactly where the blind spot lives.

Non-cash is not the same as non-consequential. Follow the chain. Reported NAV falls. Loan-to-value on any BTC-backed facility worsens. Collateral calls or haircuts follow. Forced sale or discounted equity raise follows that. BTC per share falls. mNAV compresses. The issuance window closes. Every link in that chain is triggered by a number that did not involve a cash movement at the start. Accounting is not decoration. Accounting is the interface between the asset side and the liability side, and that interface is where solvency actually lives.

Second: the impairment is a disclosure about capital allocation. It says management's cost basis sits above the test-date price. That is not an accounting artifact. That is a statement about when they bought. Buying strength and marking it down is a strategy, and it is not automatically a bad one — but it is information about entry quality, and the market usually throws that information away together with the "non-cash" label.

Third: the reversal will be misread in the opposite direction. When the next IFRS reporting period shows a non-cash gain, the headline will say "returns to profit." It will not say "reverses a prior write-down." The same traders who correctly ignored the impairment will incorrectly celebrate the reversal. Both reactions are wrong for the same reason: they are reading a mark, not a business.

Fourth, and this is the one I would underline in red. Nobody in the coverage is asking what mNAV is. That is the only number that determines whether this company has a future in its current form. Above 1, the impairment is noise on the way to the next raise. Below 1, the impairment is the sound of a machine that has already stopped and nobody has said so out loud yet.

The floor is a suggestion, not a law — and so is the premium.

There is a fifth angle that almost never gets discussed, and it is the one that keeps me out of these names. The wrapper's value depends on a multiple that depends on sentiment that depends on the wrapper's own recent performance. That is a feedback loop with no external anchor. In a bull market it looks like genius. In a bear market — the one we are in — it looks like a slow bleed where the reported loss keeps widening and the coins keep sitting there and the equity keeps de-rating for reasons that have nothing to do with Bitcoin's price and everything to do with the multiple applied to it.

I do not trade narratives. I trade structures. And the structure here is: a small-cap wrapper holding a volatile asset, reporting under a ratcheted accounting regime, with an undisclosed mNAV and an undisclosed debt stack, in a market where the marginal buyer has stopped showing up.

That is not a thesis. That is a set of open questions. And open questions get answered by the tape, not by the press release.

Takeaway: What To Track Instead Of Net Income

If you hold this name, or any name in the category, stop reading net income. It is the least informative number in the filing.

Track five things instead.

BTC per share, quarter over quarter. If it is flat or falling, the machine is not working, regardless of what the P&L says. This is the only metric that measures whether the wrapper is adding anything to the underlying.

mNAV. Above 1 is a going concern in this business model. Below 1 is a liquidation candidate in slow motion. It is not disclosed here, and that omission is itself a signal worth weighing.

The reversal line. When it appears — and under IFRS it can appear — do not confuse it with performance. It is the same spreadsheet running backwards.

The debt documents. Covenant definitions matter more than the impairment. Read the indenture before you read the press release.

The custodian and the auditor. Unsexy. Load-bearing. Almost never in the pitch deck.

And a question to carry forward into the next reporting season. The next time a Bitcoin treasury company reports a profit, ask one thing before you buy the headline: did it sell anything, or did it just re-mark a spreadsheet?

Because in this asset class, the difference between a $78 million loss and a $78 million gain can be a single date on a calendar. The accounting regime that decides which one you see is not the same for everyone. And the companies that understand their own regime best are the ones most likely to tell you about it last.