Which companies hold crypto on their balance sheet? The 2026 map

What this page is, and what it is not

This is a dated snapshot of listed companies that hold cryptocurrency as a treasury asset, compiled on 26 August 2026 from the documents named against each figure. It is not a league table, not a ranking of quality, and explicitly not a buy list. Naming a company here is not an endorsement of it, of its shares, of its preferred stock or of the token it holds. We publish no price views, no targets and no allocations.

Treat every number as perishable. Several of these companies disclose holdings weekly by Form 8-K, so a figure can be out of date within days of publication. Any list you find without an "as at" date on every row — including this one, if you are reading it in six months — is telling you what was true on some unknown day.

How we grade every figure on this page

Holdings numbers in this sector are not all the same kind of claim, and the difference matters more than the number. Every figure below carries a grade:

• Filed fact — taken from a company's own regulatory filing (SEC Form 10-Q, 10-K, 8-K or 6-K; an HKEX announcement; a Euronext regulatory release). This is the strongest grade available.
• Company statement — the company said it, in a presentation, dashboard or press release, but we could not tie it to a filed document.
• Market data — a price or market capitalisation observed on a date.
• Aggregator — a third-party tracker. See why you should not trust a holdings aggregator; these are demonstrably unreliable.
• Press report — reported by a news outlet, not seen by us in the original document.
• Our inference — arithmetic or reasoning of ours, clearly flagged as such and never presented as disclosure.

Dollar figures come from US filings and market reports and are left in dollars. Yen, euro, Hong Kong dollar, Canadian dollar and Swedish krona figures come from the filings of issuers reporting in those currencies. We have converted nothing, because a silent conversion hides the exchange rate and the date it was struck.

The names and tickers keep moving

This is the sector's defining reporting hazard. At least a dozen companies in this cohort changed identity between mid-2025 and mid-2026, and in most cases the ticker did not change with the name, or changed without the name changing:

• KindlyMD became Nakamoto Inc on 21 January 2026 (NAKA).
• Asset Entities became Strive, Inc. (still ASST).
• LM Funding America became PowerCompute on 22 July 2026 (LMFA became PWCM).
• Volcon, an electric powersports manufacturer, became Empery Digital (EMPD).
• The Blockchain Group became Capital B (ALCPB).
• WeTrade Group became Next Technology Holding (NXTT).
• Bitfarms Ltd became Keel Infrastructure Corp and redomiciled from Ontario to Delaware effective 1 April 2026; BITF became KEEL, and the old Ontario subsidiary was renamed Backbone Hosting Solutions Inc on 15 June 2026.
• Cipher Mining Inc became Cipher Digital Inc — same ticker, CIFR.
• Helius Medical Technologies became Solana Company but kept HSDT, a ticker that reflects a neurostimulation-device past.
• SRM Entertainment, a maker of theme-park souvenirs, became Tron Inc (TRON) — and still makes the souvenirs.
• Block's ticker changed from SQ to XYZ.
• 180 Life Sciences became ETHZilla in August 2025 and then Forum Markets in February 2026 — three identities in nineteen months.

Any article about "Bitfarms", "Cipher Mining" or "Asset Entities" written before 2026 describes an entity that no longer carries that name. Start from the registrant, not the brand: the section on how to verify any of this yourself shows how.

Where the sector stood on 26 August 2026

Start here, because it reframes everything below. Every holdings figure on this page sits inside a bear phase, and the financing mechanism that built this sector had stopped working for most of it by the snapshot date.

The price path behind every figure

• Bitcoin peaked around $126,000 in October 2025. Market data.
• It closed 30 June 2026 at $58,532. That is not a quote we looked up — it is the price Nakamoto Inc used to mark its own balance sheet in its Form 10-Q filed 13 August 2026, so it is a filed fact.
• It traded at about $78,986 on 26 August 2026, the date of this snapshot. Market data.
• Core Scientific's Form 10-Q filed 28 July 2026 gives the fullest filed context: bitcoin ranged from a low of $58,097 to a high of $97,877 in the six months to 30 June 2026, averaging $74,142. Filed fact.

Even the "same" 30 June 2026 price is not identical across filings: MARA and CleanSpark both marked at $58,524 per bitcoin, HIVE at $58,237, Keel at approximately $58,600, and Nakamoto at $58,532. Those are all filed facts, from four different pricing sources. If four audited filers cannot agree on a price to the nearest hundred dollars, treat any single sector-wide dollar total with suspicion.

That round trip — roughly $126,000 down to $58,532 and back to about $79,000 — is the single most important fact for reading anything below. It is what turned the treasury model from a self-reinforcing machine into a forced-seller problem, and almost every wind-down on this page is downstream of it.

What the market was paying for the coin pile

mNAV is the ratio of a company's market value to the value of the tokens it holds. Above 1.0 the market pays a premium for the wrapper; below 1.0 you can buy the coins inside the company for less than they are worth on an exchange — assuming you could ever get at them, which you cannot. We explain the arithmetic and its traps in mNAV explained.

The table below is aggregator data on an enterprise-value basis — the bitcointreasuries.net leaderboard, retrieved 26 August 2026. It is not filed, it is not audited, and different aggregators compute it differently — particularly where preferred stock sits ahead of the common shares and should be deducted first. Read it as a picture of a distribution, not as ten precise numbers.

The basis moves the number more than the date does. Strategy's 1.01 above is the enterprise-value figure. The same tracker showed Strategy at 0.74x on a basic basis on 25 August 2026 — 27 points apart, one day earlier, from one source: on one basis the company sits at a premium, on the other at a discount. Every ratio in the table carries the same ambiguity, so a bare "mNAV" number with no basis attached is not a fact you can use. We set out which basis does what, and why they diverge, in mNAV explained.

Two non-bitcoin comparators, for scale: Sharplink showed an mNAV of about 0.81 in August 2026 (aggregator), and Forward Industries disclosed its own fully diluted mNAV of 0.908x at 30 June 2026 (company statement — it publishes the figure itself, which is unusually transparent, and it published it at a discount).

Why below 1.0 changes everything. When the shares trade above the value of the coins per share, selling new shares and buying coins leaves every existing holder with more coins per share. That is the flywheel, and it genuinely worked in 2024 and 2025. When the shares trade below that value, the same action destroys coins per share: you are selling a claim on coins for less than the coins are worth in order to buy coins. The accumulation engine does not merely slow down — running it actively harms the people it is supposed to benefit.

How does the flywheel run in reverse?

The model these companies copied has three legs: issue equity above the value of the coins per share; use the proceeds to buy more coins than the dilution cost; repeat. The mechanism in full, including how it is supposed to work, is set out in how digital asset treasury companies actually work. What follows is the reverse gear, which promotional material routinely omits.

Leg one is load-bearing

Remove the premium and only three levers remain: borrow, sell coins, or stop. All three are visible in the filings.

Stop. Twenty One Capital held 43,514 bitcoin at 30 June 2026 against 43,515 at 31 December 2025 — it acquired net minus one bitcoin in the first half of 2026, despite being marketed as a bitcoin accumulation vehicle (filed fact, Form 10-Q filed 11 August 2026). The company's own key-performance table shows bitcoin per share falling from 12,557 to 12,547 satoshis over the same period. The engine did not just stall; the metric it exists to grow went backwards.

Sell. Strategy, which spent five years telling the market it would not sell, sold 1,690 bitcoin in the week of 3–9 August 2026 at an average of $64,262 — below its own stated average purchase price of $75,385 — and used the net proceeds to fund repurchases of its STRC preferred stock, per the filing's own footnote (Form 8-K, 10 August 2026). Filed fact. That is a company selling the asset below cost to defend the price of the paper it issued to buy the asset.

Buy back the shares instead. Once the discount opens, the accretive move flips from issuing stock to retiring it — and buybacks need cash, which in a company with no operating profit means selling tokens. Forum Markets (then ETHZilla) retired about 7.1 million shares, roughly 35% of its share count, at a $4.42 average in the second quarter of 2026, having funded earlier buybacks by selling ether; its chairman said the firm would keep selling ether "until the discount to NAV is normalized" (press report of a company statement). Empery Digital repurchased 26.2 million shares for $149.7m, cutting shares outstanding from 33,800,951 at 31 December 2025 to 28,110,111 at 30 June 2026 — funded by selling the bitcoin it was created to accumulate (filed fact, Form 10-Q filed 7 August 2026). Sharplink did both inside one quarter: a $75.0m registered direct offering at $7.49 on 23 June 2026, which the company stated was at a premium to net asset value, and about 2.1 million shares repurchased at roughly $4.70 (filed fact, Form 8-K Exhibit 99.1 filed 10 August 2026).

A company trading at a discount to its own tokens is structurally a seller of those tokens. That is the reverse gear, and it operates regardless of what anybody believes about the long-run price of the asset.

Dilution without selling a single coin

The most under-reported failure mode in this sector is not selling the asset. It is holding the asset while the share count explodes.

Next Technology Holding held approximately 5,833 bitcoin at 31 December 2025 and the identical 5,833 bitcoin at 30 June 2026 — nothing bought, nothing sold. Over the same six months, shares outstanding went from 4,882,556 to 147,296,192. Both figures come from the same Form 10-Q, filed 24 July 2026 (filed fact). A 1-for-100 reverse split followed on 10 August 2026 (press report), on top of a 1-for-200 reverse split in September 2025. Published analysis puts bitcoin per share falling from roughly 204,000 satoshis to roughly 4,000 (press report) — a destruction of about 98% of a holder's claim on the treasury, with the treasury untouched.

Reverse splits cluster in this sector because Nasdaq requires a $1.00 minimum bid price, and they cluster hardest around the companies whose coins per share have already been destroyed. Nakamoto executed a 1-for-40 reverse split on 22 May 2026 after a 10 December 2025 non-compliance notice, regaining compliance on 9 June 2026, and had only 17,894,943 shares outstanding at 10 August 2026 (filed fact, Form 10-Q filed 13 August 2026). PowerCompute executed a 1-for-25 effective 13 July 2026, leaving 1,794,135 shares outstanding at 11 August 2026 (filed fact, Form 10-Q filed 14 August 2026). American Bitcoin did a 1-for-15 effective 2 July 2026 (filed fact).

Two reverse splits inside twelve months is, by itself, the signal. You do not need a view on bitcoin to read it.

Collateral mechanics in a drawdown

Borrowing against the coins converts a price fall into a forced sale, on a schedule set by somebody else. Two filings disclose the thresholds precisely enough to be worth memorising.

Empery Digital's borrowing arrangements, permitting up to $150.0m: initial collateral of 250% of borrowings, a collateral call at 175%, excess returned above 345%, and a liquidation level at 150% at which the company must top up within 24 hours or the lender may liquidate. The Form 10-Q filed 7 August 2026 records those thresholds actually being breached and the terms renegotiated twice in 2026, after bitcoin fell on 4 February and again on 3 June. Filed fact.

Hut 8's FalconX term loan: $200.0m principal at 7.0% fixed, maturing 30 April 2027, secured on bitcoin in a segregated custody account, with an initial collateral ratio of 143%, a margin call at 130% and liquidation at 105% (filed fact, Form 10-Q filed 4 August 2026). Show a reader those three numbers and the risk stops being abstract: below a certain price, a third party sells 4,850 bitcoin, at the worst possible moment, and the shareholder has no vote.

Nakamoto's Kraken facility is the same shape in different words. Its Form 10-Q filed 13 August 2026 discloses that approximately $222.7m of its bitcoin is subject to a contractual sale restriction running to the loan maturities, and that below the liquidation threshold Kraken "may liquidate the pledged Bitcoin". The risk factors say it plainly: margin or collateral calls "could require the forced liquidation of Bitcoin holdings at unfavorable prices". Filed fact.

How do these companies pay for the coins?

The funding method decides how a company behaves when the price falls, so it tells you more than the headline holding does. Five distinct forms appear in this cohort, in roughly ascending order of how quickly somebody else can force the company's hand.

1. At-the-market equity — the default engine

Continuous issuance of ordinary shares straight into the market. Strategy sold $2,006.5m of stock in the week to 23 August 2026 (aggregator roll-up of its weekly Form 8-K disclosures — open the filing itself before quoting it). Strive funds its buying through an at-the-market sales agreement, PIPE warrants and its STRC preferred instrument (filed fact, Form 10-Q filed 10 August 2026). Sharplink raised $75.0m in a registered direct offering at $7.49 on 23 June 2026, which it stated was struck at a premium to net asset value, and its share count rose from 196,707,797 at 31 December 2025 to 216,983,308 at 30 June 2026 — about 10% dilution in six months (filed facts).

An at-the-market programme carries no forced-sale trigger, which is why it is the least dangerous form. But it is only accretive while the shares trade above the value of the coins per share, it dilutes relentlessly, and the share count does not shrink when the premium disappears.

2. Preferred stock — a cash obligation that does not care about the coin price

Strategy's stack of perpetual preferreds (STRK, STRF, STRD, STRC and STRE) is the template, and others have copied it. Metaplanet disclosed ¥23.6bn notional of MERCURY Class B preferred at 30 June 2026, issued but unlisted with no public trading market, alongside a MARS Class A preferred that is shelf-registered and not yet issued (company statement, Q2 2026 earnings presentation, 13 August 2026). Strive has a separate preferred line, SATA; BitMine's Series A preferred trades separately as BMNP.

The mechanism to understand is that preferred dividends are a cash cost, and bitcoin sitting in cold storage generates nothing to pay them. If the operating business does not generate that cash — and in most of these companies there barely is an operating business — the money has to come from issuing more securities or from selling coins. Strategy's August 2026 sale of bitcoin below its own average purchase price to fund STRC repurchases, described in the section above, is that obligation reaching back into the coin pile. It also matters for anyone reading an mNAV figure: preferred stock sits ahead of the ordinary shares and must be deducted before any net-asset-value-per-share calculation, which is one reason different trackers publish different mNAVs for the same company.

3. Convertible notes and coin-backed loans — where the margin call lives

This is the form that converts a price fall into a forced sale on somebody else's schedule.

• Twenty One Capital: $486.5m of convertible notes issued at closing, conversion price $10 at a 130% premium, effective interest 1.09%, secured on approximately 16,116 bitcoin (filed fact, Form 10-Q filed 11 August 2026).
• Nakamoto: the Kraken "June Loan" of 5 June 2026 — a fee of 7.75% a year while at least 2,000 bitcoin sits in a designated yield-strategy trading wallet and 8.00% below that, initial collateral of 3,805 bitcoin, partial maturity 4 December 2026 with the remainder on 30 June 2027, and a liquidation threshold at which Kraken may sell the pledged coin (filed fact, Form 10-Q filed 13 August 2026).
• DeFi Development Corp: repurchased $3.5m of principal of its July 2030 convertible notes for $2.3m cash — a 35% discount to par, which tells you where its own paper trades (company statement, 12 August 2026).
• Sequans: the completed version. Convertibles issued in July 2025 and collateralised by the bitcoin itself; when bitcoin fell, the collateral position became the binding constraint and roughly 80% of the stack was sold to redeem the debt (filed fact).

The thresholds disclosed by Empery Digital and Hut 8 — set out under collateral mechanics in a drawdown — are what these arrangements look like from the inside.

4. Debt with nothing pledged — a different risk, not a smaller one

Metaplanet's "BitBonds" are yen-denominated senior unsecured bonds: an inaugural ¥200m across series 21–24 at 4.0–4.3%, roughly three-year, by private placement (company statement, disclosed 13 August 2026). No bitcoin is pledged, so there is no margin mechanic and no liquidation trigger — buyers take Metaplanet credit risk instead. Its wider stack at 30 June 2026: ¥77.3bn of debt (a ¥67.5bn senior credit facility, ¥8.0bn of unsecured senior notes and ¥1.8bn of other), total debt plus preferred of ¥100.9bn, an equity ratio of 81.4%, and a self-calculated "coverage" of 4.1x which the company states is not a credit rating. The distribution arrangement — the issuer selling its own paper through its own broker subsidiary — is described under two rows that are not what they look like.

5. Coins contributed in kind, rather than bought

The least understood form, and the one most likely to be misread as an achievement. No cash changes hands: the company issues securities and receives tokens.

• Twenty One Capital was assembled this way. Tether contributed 24,500 bitcoin and Bitfinex 7,000 under a contribution agreement, with further PIPE bitcoin on top — including 4,812.22 BTC of "Initial PIPE Bitcoin" at $458.7m and a 1,381.16 BTC "June PIPE" at about $147.5m (filed fact).
• Next Technology Holding issued 135,171,078 shares plus 294,117,647 nominal-exercise-price warrants on 12 March 2025 as consideration for 5,000 bitcoin — a share-for-bitcoin swap, which is also why the acquisition value is hard to verify independently (filed fact).
• H100 acquired Norwegian bitcoin-holding companies holding 2,455 BTC by issuing 790.5 million new shares, with no cash consideration (press report). That is accretive to total coins and dilutive per share at the same time, and whether existing holders gain depends entirely on the exchange ratio against the mNAV of both sides — the calculation almost nobody does.
• Empery Digital paid $28.0m of a roughly $501.0m raise in bitcoin — 235.8 coins (filed fact). GameSquare bought a CryptoPunk for $5.15m in preferred shares rather than cash (press report).
• And the version with a liability attached. Alpha Compute's TON treasury was seeded by Animoca Brands affiliates with put options attached — options the contributors could exercise against the company, which makes an apparently equity-funded treasury a debt-like obligation. On the July 2026 wind-down the tokens went back to the contributors and the put liabilities were extinguished. Outside shareholders received nothing (company statement, 10 July 2026).

The question to ask of any treasury announcement is not how many coins, but whether they were bought or contributed, and on what terms. A contribution can bring a counterparty with a claim, a control block, or an option to hand the tokens back at a price of their choosing.

The running cost nobody advertises

Holding is not free. Strive's general and administrative expenses rose 268% year on year in the first half of 2026 on custody fees, insurance and listing costs (filed fact, Form 10-Q filed 10 August 2026). Sol Strategies held cash of just $0.37m against payables of $1.77m at 31 March 2026, with continued operation explicitly reliant on its crypto holdings, credit facilities and capital markets access (filed fact). PowerCompute's Form 10-Q filed 14 August 2026 carries a going-concern discussion against approximately $854 thousand of available cash. When the equity window closes, those bills are still due, and the only asset that can pay them is the one the company exists to hold.

What does "coins held" actually mean?

A headline holdings figure is not a claim on unencumbered assets. "Holds X bitcoin" and "owns X bitcoin free and clear" are different statements, and the filings distinguish them where the press releases do not. As at the dates shown:

The convertible notes secured on Twenty One's bitcoin are described in its own filing as leaving those coins "not available for general corporate purposes or liquidity needs while serving as collateral". That is the point. Encumbered coins are pre-committed to a lender under a margin mechanic; they are not a liquidity buffer, and in a stress they are the first thing to go.

When your bitcoin leaves your own balance sheet

There is a worse version, and almost nobody reports it. If the lender is permitted to rehypothecate — to re-lend or sell the collateral — then under US accounting the borrower has given up control, must derecognise the bitcoin altogether, and instead books a receivable from the lender. The asset stops being bitcoin and becomes a claim on a counterparty.

Cipher Digital states the mechanic in its own words in its Form 10-Q filed 4 August 2026: borrowings under its Coinbase overnight facility are "collateralized by bitcoin transferred to the Lending Service Provider's platform. Since the lender has the right to rehypothecate the bitcoin held as collateral, the Company derecognizes the bitcoin transferred." Filed fact. Cipher happened to be undrawn at 30 June 2026, so the risk was dormant at the balance sheet date — but the facility remains available.

PowerCompute is the live version: 174 bitcoin (145 at 31 December 2025) pledged under its Galaxy loan facility are derecognised from the bitcoin balance and sit in "Digital assets receivable", with a fair value of approximately $10.2m securing about $11m of borrowings. The headline balance-sheet bitcoin number is 144.6; total economic exposure is roughly 319 (our inference from the two disclosed lines). CleanSpark runs the same mechanic in reverse direction: its bitcoin line says 12,205 BTC, but a separate "Receivable from bitcoin collateral" line holds another 1,719 BTC ($100.6m) posted to derivative counterparties where control transferred, so economic exposure is 13,924 (filed facts, Form 10-Q filed 6 August 2026; the 13,924 total is our inference).

The lesson is blunt: your bitcoin can leave your balance sheet without you selling it. When you read a holdings figure, find out whether it counts the receivable line, and whether the coins behind it can be re-lent by somebody else.

Global bitcoin treasury companies outside the UK

The table covers the significant non-UK names, with the holding, the date it was struck, the document it came from, and the funding method — because the funding method determines how the company behaves in a fall, and it matters more than the headline holding. Companies whose coin is mined rather than bought sit in the miners section, and operating companies that hold alongside a real business sit in corporates are not treasury companies.

How big is the sector, really?

The most-cited aggregate, retrieved 26 August 2026, is 197 public companies holding 1,265,440 BTC, roughly 6.0% of the 21 million cap, with Strategy alone accounting for 840,447 BTC — 66.4% of the public-company total. Strategy's own number is a filed fact (Form 8-K filed 24 August 2026, holdings as at 23 August 2026); the total it sits inside is not. A dated cross-check from Bitwise put it at 174 public companies and 1,187,898 BTC as at 12 May 2026 (press report).

Both figures are aggregator roll-ups, and the larger one is demonstrably defective. We could not reconcile it to primary sources because it is not built from any. It includes:

• BSTR at 30,021 BTC — a company that never listed, whose merger died on 20 August 2026. That is 2.4% of the entire reported public-company total.
• SpaceX at 18,712 BTC — a private company, not a public company at all.
• Exchanges and custodians whose coin is a different animal entirely: Bullish 23,800, Coinbase 17,311, Gemini 5,528, BitGo 2,523.
• Miners' working inventory, which is produced and sold rather than accumulated.
• At least one stale entity: "LM Funding America, LMFA, 334 BTC" — an entity that has been called PowerCompute since 22 July 2026 and reported 144.6 BTC on balance sheet at 30 June 2026.

Strip out the non-listed, the exchanges and the miners and the genuine treasury-company universe is materially smaller than 197. There is also a consolidation problem that nobody has quantified, and it needs stating in two parts rather than one. The filed part: American Bitcoin's 8,002 coins are consolidated into Hut 8's 17,316 (filed fact, Hut 8 Form 10-Q filed 4 Aug 2026), so anyone who adds the two published headline totals together counts those coins twice. Whether the leaderboard itself does that is unresolved: it carries Hut 8 at 10,278, which is close to the 9,314 corporate-level residual we derive in the miners section — consistent with the tracker already showing Hut 8 excluding American Bitcoin rather than double-counting it. We could not confirm which, so we are not asserting a double-count here. Separately, Metaplanet's 2,100 BTC contribution creates a second listed entity holding coin Metaplanet still controls; Nakamoto holds an investment in Treasury B.V., another treasury vehicle. Our inference: any "corporate bitcoin" total you see is an editorial choice about consolidation dressed up as a measurement.

Non-bitcoin treasuries, and what staking actually does

A bitcoin treasury holds an inert asset. Its only levers are issuing stock above net asset value, borrowing, and selling. A proof-of-stake treasury has a fourth lever: the asset produces a protocol-level return. That single difference is the entire pitch for this half of the sector, and it is the half most often misdescribed.

The yield is revenue, and the company keeps it

This is the central point of this whole section and it is routinely misreported. Across every company we verified, staking rewards are recognised as corporate revenue and retained. Not one distributes staking yield to common shareholders.

The filed evidence is unambiguous. Sharplink booked "Revenue from staking" of $11.161m in the second quarter of 2026 and $22.662m in the first half (filed fact, Form 8-K Exhibit 99.1, 10 August 2026). That sits on the income statement above selling, general and administrative expenses of $9.06m for the same quarter. In other words, the staking line very nearly pays the corporate overhead. It functions closer to an operating subsidy for the company than to a return for the shareholder, and describing it as "yield you receive" is simply wrong.

Tron Inc books its equivalent as "unrealized income from staking activities" — $3.0m in the first quarter of 2026 and $6.33m across the first half, including $3.35m in Q2 (filed fact). Note the word "unrealized": that is accrued token value, not cash received, and it is retained.

The only mechanism we found that actually connects staking yield to shareholders is a buyback: GameSquare approved a repurchase programme "funded by Ethereum yield proceeds" in 2025 (its 2026 status is unverified). That is the exception, not the rule. Hyperliquid Strategies is explicit about the intended direction of travel, promising "compounding shareholder returns... through staking, yield optimization, and active ecosystem engagement" — note that this promises compounding inside the company, not distribution out of it.

The scale mismatch between yield and price

Ether staking yields roughly 2.7–2.9% — BitMine's own disclosures show a seven-day yield of 2.91% annualised against the CESR benchmark of 2.84% as at 8 March 2026, and 2.67% as at 23 August 2026. Solana is higher: Forward Industries discloses 6.73% gross validator APY before fees. Bittensor root-network staking has been claimed at around 10%, which mostly reflects emission-driven dilution rather than fee revenue — a different thing from a yield, and readers should understand the difference.

Now put those percentages against the price moves in the same filings:

• Sharplink, first half of 2026: staking revenue $22.7m against an unrealised loss of $827.7m and a net loss of $1,079.8m. Filed facts.
• FG Nexus, the definitive illustration: over the entire life of its roughly eleven-month ether programme it earned $144,000 in staking revenue, and recorded a $45.2m operating loss on discontinued digital-asset operations in the first half of 2026 alone. Filed facts, Form 10-Q for the quarter ended 30 June 2026. The staking yield was about three-tenths of one percent of the loss.

A 3% return on an asset that fell far more than 3% is not a return. Any explainer that presents "these ones pay you yield" without that ratio is misleading you by omission.

Slashing, lock-ups and validator failure

Slashing is asymmetric across chains, and it is usually the least of it. Ethereum has genuine slashing: a validator that double-signs or surrounds a vote loses stake and is forcibly exited. Solana's official documentation states slashing "is not automatic" and is applied only on a network restart after an attacker halts the chain; native SOL delegation has carried no slashing loss to date. Hyperliquid uses delegated proof-of-stake. TRX uses a freeze-and-vote model with 27 Super Representatives, not slashing at all. The bigger operational risks are validator downtime, custodian failure, and smart-contract or wrapper failure — the one that has actually cost money.

Unbonding lock-ups are the constraint nobody prices. Ethereum's queues are asymmetric and stress-correlated: as at 17 August 2026 the exit queue was trivial (2 validators, about 64 ETH, minutes) while the entry queue held 2,229,411 ETH — about 39 days. But in September 2025 the exit queue peaked at around 2.67 million ETH: weeks to unstake, precisely when everybody wanted out. The queue is short when nobody needs it.

Solana deactivation completes only at an epoch boundary, roughly two days, and no more than 25% of total network stake can activate or deactivate in one epoch — a system-wide throttle that binds exactly during a stampede. Hyperliquid imposes a one-day delegation lock-up plus a hard seven-day unstaking queue that cannot be accelerated, roughly eight days end to end. A company that states an objective of staking 100% of its treasury is therefore, at any moment, about eight days away from being able to sell anything.

Our inference, clearly flagged as such: BitMine reported exactly 5,067,309 ETH staked on both 9 and 23 August 2026 despite buying ether every week in between. That is consistent with newly bought ether sitting in Ethereum's roughly 39-day entry queue. The company has not said so.

Counterparty failure is not theoretical. Reliz Ltd, the operating entity of crypto brokerage BlockFills, filed for Chapter 11 on 16 March 2026. Keel Infrastructure had closed its outstanding positions with BlockFills before the filing and recognised a net gain of $1.9m, but still booked a credit loss expense (filed fact). "Yield on your tokens" frequently means lending them to somebody who can go bankrupt.

The wrapper problem

Headline "ETH held" numbers frequently are not ether. Sharplink's own footnotes disclose that its 888,938 ETH as at 3 August 2026 comprises 634,255 native ETH plus 181,748 ETH "as-if redeemed" from LsETH (Liquid Collective) plus 72,935 ETH "as-if redeemed" from weETH (ether.fi restaking). Roughly 29% of the headline figure is a liquid-staking or liquid-restaking token valued at a theoretical redemption rate.

That matters twice over. First, redemption depends on a protocol's queue and on its peg holding. Second, US accounting treats the two differently: Sharplink carries native ether at fair value but carries LsETH and weETH at cost less impairment, and the company states plainly that impairment charges "are not reversed for subsequent market recoveries". It took $76.1m of such charges in Q2 2026 and $267.8m across the half — losses that can never be written back even if ether recovers fully. Filed facts.

Tron Inc is the concentrated version of the same risk: published analysis of its filings indicates over 90% of assets sit in JustLend sTRX, a single uninsured DeFi protocol (press report). That is smart-contract risk on the entire balance sheet, not slashing risk. FG Nexus took a $2.8m impairment specifically on wrapped staked ether, separate from and additional to its $41.2m of losses on ether itself — the wrapper cost extra (filed fact).

"Yield-bearing" and "holds the asset" are not the same claim.

Not all "yield" is staking

GameSquare targeted 8–14% annualised through Dialectic's on-chain platform from 1 August 2025. That is DeFi lending and looping — liquidation, oracle and smart-contract risk — not consensus rewards. Any comparison table that places "8–14%" next to BitMine's 2.67% without that distinction is actively misleading.

XRP has no native staking at all. The XRP Ledger uses a federated consensus protocol with no block rewards and no proof-of-stake issuance, so an XRP treasury cannot run the ether or solana yield argument. Evernorth's stated plan is to lend XRP, provide liquidity, run options strategies and deploy Ripple's RLUSD stablecoin in DeFi — every one of which is counterparty, market or smart-contract risk rather than protocol yield. VivoPower's stated plan to "stake" XRP on the Flare Network means bridging it as FXRP and deploying it into lending-style products. Any coverage describing "XRP staking" is wrong. Around eight companies had publicly disclosed XRP treasury strategies as at January 2026 with aggregate stated commitments above $2bn — commitments, not verified holdings, and we would publish no figure for Trident Digital, Webus International or VivoPower without a filing.

One genuine second-order business does exist. Validator commission — SOL Strategies' 3.8 million SOL of assets under delegation, Forward's and DeFi Development's own validators, BitMine's MAVAN network, Tron's move to seek Super Representative status — is real, recurring and not directly token-directional. It is also small relative to treasury mark-to-market, which is why it never rescues a quarter.

Why should you not trust a holdings aggregator?

Because they are demonstrably wrong, and we can show you exactly where.

The bitcoin side of the sector shows the same failure in four different flavours, all listed under how big is the sector, really: a company that never listed still ranked fifth by holdings; a private company counted as public; exchanges and custodians whose coin belongs to their customers; and an entity listed under a name it stopped using on 22 July 2026, at a holdings figure more than double what its own filing reported.

There is a fifth flavour that catches even careful readers: gross figures presented as ownership. Galaxy Digital's Form 10-Q filed 10 August 2026 shows 21,084 bitcoin in its significant-holdings table. That is not Galaxy's treasury. The filing says the balance "includes digital assets borrowed from counterparties, deposited by counterparties as collateral for which Galaxy has control, and any digital assets which Galaxy has pledged as collateral and retains control." Set against it: $1,514.5m of digital assets borrowed and $983.1m of collateral payable. Galaxy does not disclose a clean net own-account figure, and we will not invent one. Any list ranking Galaxy among the largest corporate bitcoin holders on that number is measuring the wrong thing.

Which filing to open instead

• For a weekly holdings figure: the company's latest Form 8-K, usually under Item 8.01, with the holding, the average price and the share count in one table. Strategy, Strive and BitMine all disclose this way.
• For what is actually owned: the digital assets note in the Form 10-Q. This is where cost basis, fair value and the split between unrestricted, restricted, pledged, loaned and derecognised live. For Sharplink's ether, it is Exhibit 99.1 to the 8-K of 10 August 2026 and the footnote splitting native ether from LsETH and weETH.
• For a foreign private issuer — Bitdeer, Sequans, Prenetics, Genius Group, SOL Strategies — the equivalent is Form 6-K, and there is no quarterly 10-Q obligation, so the cadence is slower and cross-company comparison is weaker than it looks.
• For Hong Kong, Tokyo, Paris and Canada: the HKEXnews announcement, the TDnet filing, the Euronext regulatory release, and SEDAR+ respectively.

Corporates are not treasury companies

Tesla holds bitcoin. So does Block. So does Coinbase. None of them is a treasury company, and the distinction is teachable in one sentence each.

A treasury company's purpose is holding. Holding is the business. It funds purchases by issuing securities — at-the-market equity, convertible notes, preferred stock — so the share is a claim on a coin pile whose coins per share moves every time management taps the market. There is no operating cash flow to fall back on. You are exposed to two things at once: the coin price, and management's issuance decisions.

An operating company holds a treasury position alongside a real business, and does not issue equity to buy more of it. The coin sits on the balance sheet as one line among many. The share price is a claim on the business.

Four things that change when holding is the business

1. Dilution runs the other way. In a treasury company the share count is a policy lever pulled in order to buy coins; dilution is the engine. In an operating company share issuance is incidental. CleanSpark went the opposite way entirely, spending $460.0m buying back its own stock in the nine months to 30 June 2026 while selling coins (filed fact). No treasury company does that; it is the model inverted.
2. You are valuing a different thing. Treasury company: coins per share, times coin price, times whatever premium or discount the market applies. Operating company: discounted cash flows from a business, plus a small asset on the side. Applying either framework to the other produces nonsense.
3. Forced-seller risk exists in only one of them. A treasury company that has borrowed against its coins can be made to sell into a falling market by a lender. An operating company with a modest unlevered position has no such mechanism.
4. The accounting now bites both. Since ASU 2023-08, crypto is marked to fair value through net income every quarter. An operating company's reported earnings therefore swing on the coin price even though its business has not changed. Separate "the business got worse" from "the coin went down" before you conclude anything. KULR's own chief financial officer made exactly this point in August 2026, saying bitcoin volatility made "the underlying performance of our battery business more difficult for shareholders to assess" — and then exited the position.

Miners are different again

A treasury company's only real cost is financing. A miner is an industrial business that happens to be paid in bitcoin, and it carries four risks a treasury company simply does not have.

1. Operating costs that do not fall when the coin does. MARA's Form 10-Q states energy alone was 54.9% of owned mining revenues in Q2 2026, up from 34.1% a year earlier — the cost did not rise, the revenue fell. A treasury company holding through a drawdown simply holds. A miner keeps paying the power bill.
2. Energy exposure, including collateral calls. Power and transmission contracts demand security. TeraWulf posted $360.7m of letters of credit in July 2026 for one campus, cash-collateralising $286.7m of it with a $295.3m deposit (103%). That is nearly $300m of cash locked up before a single server is powered.
3. Hardware depreciation — the risk most readers miss entirely. Rigs are wasting assets. CleanSpark booked $111.0m of depreciation in a quarter with $138.0m of revenue. HIVE recorded $52.2m against $79.1m of quarterly revenue — 66%. Core Scientific took a $266.5m impairment on mining property and equipment in Q1 2026, $151.6m of it because rigs were carried above secondary-market value, then a further $19.5m loss remeasuring machines held for sale and a $41.9m loss to terminate a mining-equipment purchase contract with Block, Inc. Bitcoin in a wallet does not depreciate; the machines that produce it lose value relentlessly.
4. The halving. The block subsidy halves roughly every four years — cut to 3.125 BTC per block in April 2024, next expected around 2028. Revenue per unit of work is contractually programmed to fall by half on a schedule no company controls. There is no equivalent for a holder.

Net effect: a miner's equity is a leveraged, decaying claim on the bitcoin price minus a large and sticky cost base, with a scheduled revenue cut built in. It is not a proxy for holding bitcoin.

What the miners actually hold, and what is pledged

All figures below are as at 30 June 2026 and come from each company's own Form 10-Q for that quarter, with the filing date shown. The encumbrance column is the part usually buried in the notes, and it is the part that matters.

Read the encumbrance column before the headline column. MARA is the flagship case: 35,577 coins reported, 26,307 genuinely free at the balance sheet date, and then more than half the position pledged within five weeks of it. A reader relying on the 30 June number alone would badly misjudge the risk. MARA also explicitly rejects passive holding, describing bitcoin as "a productive asset" to be activated "through lending, structured trading arrangements, and collateralized financing" — which earns income and converts a bitcoin holding into a portfolio of counterparty credit exposures, complete with an allowance for credit losses.

The miners are also no longer accumulating. MARA sold approximately 23,093 bitcoin in the first half of 2026, cutting holdings from 53,822 to 35,577, having expanded its policy in 2026 "to allow for sales of bitcoin held on our balance sheet". Riot sold 9,665 bitcoin for $732.5m and realised a $113.2m loss, funding itself by selling coins rather than issuing shares. Cipher sold $123.4m of bitcoin at a realised loss of $47.7m. Core Scientific sold $208.2m of self-mined coin. HIVE mined 1,004 bitcoin in the June quarter and ended it holding 190. Our inference: the miners-to-AI rotation is one of the largest sources of corporate bitcoin supply in 2026, and it is the counterweight rarely set against the "companies are accumulating" narrative.

One more trap: American Bitcoin's 8,002 coins are consolidated into Hut 8's 17,316 (filed fact, Hut 8 Form 10-Q filed 4 Aug 2026). Adding the two published headline totals together overstates the sector by 8,002. By arithmetic (our inference, not a disclosed figure), Hut 8 at corporate level holds the residual 9,314, of which 4,850 are the FalconX pledge. The bitcointreasuries.net leaderboard carries Hut 8 at 10,278, close to that residual, so it may already be showing the company excluding American Bitcoin rather than double-counting it — unresolved, and the reason we do not assert a tracker double-count. American Bitcoin's management is also unusually clear that it is not a treasury company: "we generate bitcoin through scaled infrastructure rather than simply holding it on a balance sheet." Its accumulation is funded by production economics rather than by an equity premium, so it does not stop working when mNAV falls below 1.0 — but it is exposed to hashprice, energy costs and the halving instead.

The pivot to AI and HPC data centres

Several miners have stopped accumulating and rebuilt as data-centre landlords. The hardest number is Core Scientific's revenue mix: in Q2 2026, colocation revenue was $136.7m against digital-asset self-mining of just $21.5m — colocation is 83% of revenue, against $10.6m of $78.6m (13%) a year earlier. Its CoreWeave contract covers roughly 590 MW across five sites with about 395 MW billing, and it controls around 2.1 GW of gross utility power. Its own filing describes "the wind-down of our remaining mining obligations". Filed facts.

TeraWulf is the extreme case, and the most powerful single fact in this whole section: a company routinely listed as a "bitcoin miner" holds two bitcoin. Its property, plant and equipment is $3.60bn, of which $2,371.4m is construction in process, while miners are just $182.1m of gross cost — construction in process is thirteen times the value of the mining rigs. Anyone buying WULF for bitcoin exposure owns essentially none.

The rest of the cohort is going the same way. Cipher renamed itself from Cipher Mining to Cipher Digital, spent $964.3m on property and equipment in H1 2026 and raised $2.77bn of notes. Hut 8 raised $7.70bn of project debt in H1 2026 held in restricted construction accounts and now calls itself "an energy infrastructure platform", with mining third on the list. Riot signed a data-centre lease with AMD (50 MW, expandable to 200 MW) and, in August 2026, a lease with an unnamed "leading frontier AI lab" for an initial 191 MW. MARA restructured at a cost of $47.6m explicitly "to reallocate resources toward AI initiatives" and now describes itself as "a digital infrastructure company". Buying one of these shares in 2026 is increasingly a bet on data-centre construction execution and single-tenant credit risk, not on bitcoin — and Core Scientific's colocation revenue is concentrated in a single customer.

One correction worth making because a great deal of coverage still gets it wrong: CoreWeave does not own Core Scientific. CoreWeave agreed a roughly $9bn all-stock acquisition on 7 July 2025, and Core Scientific's own shareholders voted it down on 30 October 2025, with the merger agreement terminated the same day. Core Scientific trades independently on Nasdaq.

Failures, wind-downs and deals that died

This is the most under-published part of the sector, so we give it the space it deserves. Read it against the marketing language of 2025, when coin on a corporate balance sheet was routinely described as a permanent reserve asset that would never be sold.

Governance failure belongs on this list too. Twenty One Capital's chief executive Jack Mallers resigned as CEO and director effective 20 July 2026; the Form 8-K of 21 July 2026 states the departure was "not related to any disagreements with the Company", though Mallers publicly cited differences with the board over strategy. Raphael Zagury was appointed CEO, a proposed three-way merger with Strike and Elektron Energy was abandoned, and a two-way combination was reported as under consideration (press report, not filed). CEA Industries had to regain compliance with Nasdaq Listing Rule 5620(a) — the requirement to hold an annual meeting — by convening a special meeting on 22 July 2026. These are basic corporate-governance failures at companies asking the market for capital.

Two patterns are worth naming. First, marketing language survives the balance sheet by a considerable margin: Genius Group still described itself as "a leading AI-powered, Bitcoin-first education group" in the very release that disclosed it held no bitcoin, and said it "will recommence building its Bitcoin Treasury when it believes market conditions are more favourable". Check filings, not slogans. Second, two failed ether treasuries independently pivoted to physical cash-flowing assets — Forum Markets to tokenised jet engines and manufactured-home loans, FG Nexus to manufactured housing. When the wrapper fails, the capital does not go back to the token.

The full anatomy of what goes wrong, and the order in which it goes wrong, is set out in what actually goes wrong at a bitcoin treasury company. UK-listed wind-downs — including a company whose shareholders voted to liquidate its entire bitcoin position, return the capital and delist from the London Stock Exchange in July 2026 — are covered on the UK page linked at the top of this one.

How to verify any of this yourself

A snapshot goes stale. A method does not. Everything on this page came from documents you can open, and this is the order to open them in.

1. Find the registrant, not the brand. Search the company on EDGAR full-text search or the EDGAR company browser and note its CIK number. The CIK never changes, even when the name and the ticker do. Semler's CIK still exists; its tickers and exchanges do not.
2. Read the former names. The SEC's submissions record for a CIK carries a formerNames field — for example https://data.sec.gov/submissions/CIK0001920406.json for Strive. That is the fastest honest history of what a company used to be, and it is where the identity chains on this page came from.
3. For the current holding, read the weekly Form 8-K. Strategy, Strive and BitMine disclose under Item 8.01. The holding, the average price, the share counts and the cash balances are usually in one table.
4. For the real balance sheet, read the digital assets note in the 10-Q. Search the document for "pledged", "collateral", "restricted", "rehypothecate" and "receivable". Those five words separate what a company holds from what it owns. If a "Receivable from bitcoin collateral" or "Digital assets receivable" line exists, the headline holdings number is incomplete in one direction or the other.
5. Check the share count on the cover page. Holdings flat and shares up means fewer coins per share. This is precisely what happened at Next Technology Holding, and no holdings table would have shown it.
6. Read the subsequent-events note. MARA's 30 June encumbrance was 26%; its subsequent-events note discloses a further 18,750 coins pledged on 4 August. The most important fact in a filing is often dated after the balance sheet.
7. For a non-bitcoin treasury, find the wrapper split. Ask how much of the headline token count is native and how much is a liquid-staking or restaking token valued "as if redeemed". Then find whether the wrapped portion is carried at cost less impairment, and whether those impairments can be reversed. Usually they cannot.
8. Know who does not file quarterly. Foreign private issuers — Bitdeer, Sequans, Prenetics, Genius Group, Boyaa, SOL Strategies — file Form 6-K and have no 10-Q obligation, so cross-company comparison is weaker than it looks.
9. Use the right venue's feed. HKEXnews for Hong Kong, TDnet for Japan, the Euronext regulatory feed for Paris, SEDAR+ for Canada, and the FCA's National Storage Mechanism for UK RNS announcements.
10. Check corporate actions. Reverse splits, ticker changes and delistings are published by the exchanges. This is how you catch that a ticker in an old list now points at something else entirely.
11. Treat aggregators as a starting point only, and always click through to the filing. On the Sharplink example alone, the gap was 20,239 ether.

What this means if you are a UK investor

Four practical points, none of which are recommendations.

These are shares, and they are taxed as shares. Buying MSTR, ASST, BMNR or 3350 is buying a foreign equity, not a cryptoasset. The disposal is a share disposal for Capital Gains Tax and the share pooling rules apply — not the cryptoasset guidance you would use for holding bitcoin directly. That difference is precisely why this sector attracts UK retail interest: a listed share can sit inside an ISA or a SIPP where the coin itself cannot. If you are new to the terminology, start with what are digital assets, a UK primer.

The listing venue decides whether an ISA is even possible, and it splits three ways. HMRC maintains a list of recognised stock exchanges. NYSE, Nasdaq, NYSE American, the Tokyo Stock Exchange, the Stock Exchange of Hong Kong and the Canadian Securities Exchange are long-standing entries, so the US, Japanese, Hong Kong and Canadian lines above are ordinarily eligible in principle. A growth market is not disqualified for being a growth market — the test is whether HMRC has designated that specific market. Some are and some are not. In the UK, both AIM and the AQSE Growth Market are HMRC-designated, and shares admitted to them can be held in an ISA, which is why the London names on our UK-listed bitcoin treasury companies page are not ruled out on venue grounds. That designation does not travel with a brand: HMRC's recognition of Nasdaq does not extend to the Nasdaq First North markets, so the Stockholm First North line (H100) does not inherit it. On the same market-by-market test, Euronext Growth Paris (Capital B) and Deutsche Börse's Open Market / Scale segment (Bitcoin Group SE) each have to be checked as themselves rather than assumed either way, and US OTC quotations are not covered — so Metaplanet's MTPLF quotation would not be ISA-eligible even where a broker offers it, while the Tokyo 3350 line in principle would be. Verify every venue against HMRC's current list before you rely on this: the list changes, and our venue classification here is inference from listing segment, not a checked HMRC determination for each company.

Recognition is not availability. HMRC recognising an exchange does not mean your platform offers dealing on it. In practice UK brokers offer the Nasdaq and NYSE lines with a W-8BEN in place; very few offer direct Tokyo, Hong Kong, Euronext Growth or Canadian Securities Exchange dealing, and many restrict low-priced or repeatedly reverse-split US micro-caps. Several names on this page — PowerCompute at 1,794,135 shares outstanding, for instance — are nano-caps where spread and liquidity risk are severe.

No compensation scheme covers a market loss, and no ombudsman will fix one. The Financial Services Compensation Scheme is narrower than people assume, but it is not nothing. It does pay protected investment claims, up to £85,000 per person per firm, where an FCA-authorised firm has failed and owes you a civil liability — bad advice, misrepresentation, or a failure to return client assets when a platform collapses. What it does not do is make good a fall in market value. It is not a backstop for a coin falling, a premium collapsing, a preferred dividend being skipped, a lender liquidating pledged collateral, or a company restructuring its obligations — those are investment outcomes, not firm failures. The Financial Ombudsman Service considers complaints about how a UK-regulated firm treated you; it does not make good the loss on a US-listed share whose treasury went down. Where these securities are listed overseas, UK protections are further removed still. If the coin falls, or the premium collapses, or the company restructures, the loss is yours. The FCA's own material on cryptoassets is at fca.org.uk.

Sources and the limits of this snapshot

Every holdings figure above is attributed to a document, a date and an evidence grade. Where we could only find a company press release, a third-party tracker or conflicting reports, we have said so in the row rather than smoothing it into a clean number. Where a figure could not be verified at all — Upexi's current solana position, TAO Synergies' and GameSquare's current holdings, Solana Company's holdings, Galaxy's net own-account bitcoin, Bitcoin Group SE's and H100's counts, DDC's as-at date — we have left the gap visible rather than filling it.

Known open questions at the snapshot date, stated so you can close them yourself:

• Metaplanet's August 2026 coin movements were unresolved. A reported movement of roughly 5,014 BTC before results was denied as a sale by the CEO, and 1,000 BTC moved to Coinbase Prime on 25 August 2026 was classified as a deposit rather than a confirmed sale, because Coinbase Prime provides both custody and financing. The next holdings disclosure settles it.
• Boyaa's H1 2026 interim results were approved at a board meeting on 20 August 2026 and were due within days of our cut-off. Our primary-verified figure is 4,093 BTC at 31 March 2026.
• Hyperliquid Strategies' first audited holdings, for the year ended 30 June 2026, were scheduled for 27 August 2026 — one day after this snapshot.
• Whether the Evernorth SPAC merger has closed and XRPN has begun trading was not confirmed.
• Keel's 1,861 BTC at 7 August 2026 is press-reported from an exhibit we have linked but did not read in full.
• Monthly production disclosures from MARA, Riot and CleanSpark appear to have stopped appearing on EDGAR for July 2026; check company investor-relations pages directly if monthly figures matter to you.

All of this decays fast. MARA is the proof: within five weeks of the 30 June balance sheet date the encumbered share of its holdings went from 26% to over half. Every figure on this page should be re-derived from the filings for the September 2026 quarter, which companies publish in late October and early November 2026.

What this page deliberately does not do: forecast any price, rank these companies, describe any of them as an opportunity, or suggest an allocation. It is a description of what a set of public documents said on one particular day about a fast-moving, speculative and heavily promoted corner of the market. Stated corporate targets quoted anywhere above — Capital B's 15,000 BTC by 2027, DDC's 5,000 BTC by end-2026, Metaplanet's 100,000 and 210,000 BTC, BitMine's 5% of ether supply — are company promotional claims, not commitments, and several are arithmetically hard to reconcile with a sub-1.0 mNAV, because the financing mechanism that would deliver them is the one that has stopped working. If you are considering any of this with money that matters, take regulated advice.

Related guides on digital assets and tax

The companion pages on treasury companies are each linked once, in context above: the mechanism explainer, mNAV, what goes wrong, the UK-listed directory and the UK digital assets primer. Elsewhere on the site:

• Crypto ETNs and the UK ISA rules
• Crypto CGT in the UK
• Crypto tax calculator
• Complete UK CGT guide
• UK ETF guide
• Investment scam checklist
• Investing hub