Methodology

Accuracy gates

Five automated gates run against the committed data on every build and every refresh; any failure stops publication rather than shipping a suspect table:

• Row-sum invariant — the individual rows must sum to the tracked category total (within 1 BTC) and the row count must match the tracked company count exactly; the published block (top rows by BTC held) must equal the re-derived top-N count and sum, and the sunset date must be present in the data.
• Provenance — every editor-pinned holdings value must carry its primary-source citation and as-of date.
• No aggregator naming — no third-party aggregator may appear as a source in any public field.
• Percentage sanity — the published percentage of the 21 million supply must equal the recomputed value from the row data.
• Exclusion registry — no removed issuer may reappear.

The same discipline applies to the page copy: every count, total, and percentage rendered on the tracker (including the summary paragraph and FAQ answers) is derived from the data at build time. There are no hand-typed figures to go stale.

Issuers report differently — we follow the filing

"BTC held" is not a standardized disclosure. Issuers genuinely differ in what their reported figure includes, and this tracker follows each issuer's own balance-sheet presentation rather than forcing a uniform definition:

• Pledged, lent and receivable Bitcoin are counted as the issuer reports them. Bitcoin that an issuer has lent out, pledged as collateral, or recognizes as a receivable is counted as that issuer reports it. Where an issuer includes such bitcoin in its reported holdings, we include it; where the issuer derecognizes it, we do not. Because issuers account for these arrangements differently, each affected row states its basis in its footnote, citing the filing. We do not reclassify an issuer's balance sheet against its own accounting (ruling of 2026-09-24).
• Consolidation conventions differ: where a subsidiary is separately listed, its holdings appear on its own row and are excluded from the parent's row to avoid double-counting — the affected rows carry a footnote (marked with an asterisk in the table).
• Some issuers disclose composition details such as wrapped Bitcoin or lent balances; where a figure's basis is ambiguous across the issuer's own documents, the filed financial statements outrank presentations and decks.

Where a known holder discloses no unit count at all — only fiat carrying values — we do not infer one from third-party prices, blended multi-asset carrying values, or a cost basis. One narrow, defined exception exists: a unit count may be derived only when the issuer's own document states both a Bitcoin-isolated carrying value on a fair-value basis and the issuer's own valuation price for the same date, so that a reader can reproduce the arithmetic from that single document without any external price feed. Such a row is labelled as derived, with both inputs, their location in the document, the arithmetic, and any impairment caveat shown on the row itself — it is never presented as a filed unit count.

Encumbrances an issuer discloses — Bitcoin or other crypto pledged as collateral, held as restricted assets, or lent out — are reported on the row as the issuer states them, in the issuer's own terms (asset class and carrying value), and are not converted into a unit count unless the issuer states one. Whether such bitcoin sits inside or outside the row's total follows the issuer's own classification, per the rule above: for example, one issuer's filed total includes bitcoin it has lent out and carries as a loan receivable, while another derecognizes bitcoin posted as derivative collateral and reports it outside its holdings — both are shown as filed, each with its basis on the row.

Issuer announcements before a filing. When an issuer states a new holdings figure in its own official press release and no filing yet states it — for example a mid-quarter sale to fund a buyback that the issuer does not report on Form 8-K — the row may be pinned to that announcement. Such a row is cited to the issuer's release, classed as tier 3, and carries a visible "issuer announcement — not yet filed" label; it is never presented as a filed figure. The filing-upgrade watch then checks the issuer's EDGAR record every day, and when a filing that states the figure appears the announcement pin is replaced automatically by the filed figure and the filing citation. An announcement pin cannot go stale silently: if the filing's figure cannot be read unambiguously, the watch raises an alert for a human to re-pin instead.

Inclusion criteria

The main table lists publicly traded companies with a positive, disclosed Bitcoin position on their own balance sheet. Private companies, governments, ETFs and funds, and DeFi protocols are tracked as category aggregates (shown in the chart above the table) but do not get table rows. A row whose remaining value cannot be substantiated against issuer disclosure is excluded from the table.

Published rows. The public table shows the 50 largest holders by BTC held — a row count, not a holdings floor — out of the 181 public companies the dataset tracks. Every tracked row keeps updating and the ranking is live, so companies move into and out of the published set as their holdings change. Rows below the line are tracked but not public; the exports carry the cited rows within the published set.

Why a row count rather than a BTC floor. A fixed holdings floor would cut the table at a rank that shifts with every holdings change, adding and dropping rows from week to week. A fixed row count is stable: the set was sized (ruling 2026-09-09) so that, on the ruling date, the line fell some way below the 1,000-BTC mark, leaving a buffer of smaller balance sheets under it so that ordinary movement near the line does not churn the table. The published set is a row count only. It is not a materiality threshold and makes no claim about which holdings sizes it covers — where the line falls is a property of the data on each snapshot. As of the 2026-10-26 snapshot, the smallest published holding is 527 BTC.

Sunset for uncited rows. Any row still without a per-row primary-source citation on 2026-12-31 is removed from the public table (66 days from the 2026-10-26 snapshot). Until then those rows carry the "unverified — pre-citation" label and their shared vintage of 2026-08-14.

Announced-but-unclosed vehicles are not counted: Bitcoin that never reached a public company's balance sheet (for example, a terminated SPAC combination) is carried nowhere on this tracker.

What this tracker does not include yet

Capital-structure metrics — debt outstanding and preferred equity — are not shown on the tracker today. They are part of a planned expansion; when they ship, preferred equity will be reported as its own line, separate from debt, because the two carry different obligations and conflating them misstates leverage. Until those columns exist on the page, we make no capital-structure claims here.

Cost basis appears for issuers that disclose it and is otherwise blank — it is never estimated.

Questions or discrepancies

If a figure here disagrees with an issuer's own filing, use the "suggest an update" form on the tracker page. Corrections that survive verification against the primary source are applied and published in the Reconciliation Ledger.